Intelligent industrial data clustering and grouping method, system and device and medium

Abstract

The invention relates to an intelligent industrial data clustering and grouping method, system and device and a medium, and the method comprises the steps: carrying out the standardization preprocessing of collected original industrial data, obtaining standard industrial data, and constructing a high-dimensional feature space according to the standard industrial data; performing industrial feature extraction and dimension reduction on the high-dimensional feature space to obtain a low-dimensional feature vector; performing preliminary grouping on the low-dimensional feature vectors according to a preset clustering algorithm to obtain a first clustering result; performing secondary adjustment on the first clustering result to obtain a second clustering result; and performing industrial scene simulation verification on the second clustering result to obtain a verification result, and performing optimization adjustment on the second clustering result according to the verification result to obtain a target clustering combination. According to the method, a more reasonable clustering result can be obtained, and the accuracy and applicability of the clustering effect are improved.

Description

Technical Field

[0001] The present invention relates to the technical field of industrial data clustering, and in particular to an intelligent industrial data clustering grouping method, system, device and medium. Background Art

[0002] With the rapid development of industrial automation and intelligent manufacturing, a large amount of production data and equipment operation data are continuously generated. In order to extract valuable information from these massive data, automated data processing technology, especially data clustering analysis methods, has become increasingly important. However, the application of data clustering technology in industrial environments faces multiple challenges. Industrial data usually contains multiple features, forming a high-dimensional feature space, which poses great challenges to data analysis, especially in feature selection and dimensionality reduction. For clustering analysis of industrial data, it is necessary to select a suitable clustering algorithm and perform parameter tuning to obtain the best clustering effect, but existing methods often lack effective guiding principles. Summary of the invention

[0003] The main purpose of the present invention is to provide a method, system, device and medium for intelligent industrial data clustering and grouping, which can obtain more reasonable clustering results and improve the accuracy and applicability of clustering effects.

[0004] To achieve the above object, the present invention provides an intelligent industrial data clustering and grouping method, comprising: Performing standardized preprocessing on the collected raw industrial data to obtain standard industrial data, and constructing a high-dimensional feature space based on the standard industrial data; Performing industrial feature extraction and dimensionality reduction on the high-dimensional feature space to obtain a low-dimensional feature vector; Preliminarily grouping the low-dimensional feature vectors according to a preset clustering algorithm to obtain a first clustering result; Perform a secondary adjustment on the first clustering result to obtain a second clustering result; The second clustering result is verified by industrial scenario simulation to obtain a verification result, and the second clustering result is optimized and adjusted according to the verification result to obtain a target clustering combination.

[0005] Furthermore, the collected raw industrial data is subjected to standardization preprocessing to obtain standard industrial data, and a high-dimensional feature space is constructed based on the standard industrial data, including: Performing data cleaning on the original industrial data, removing outliers and missing values ​​to obtain cleaned data, and normalizing the cleaned data to unify data of different dimensions to the same scale to obtain normalized industrial data; Performing standardization processing on the normalized industrial data to obtain standardized industrial data; According to the standardized industrial data, each data point is mapped using a preset feature mapping function to obtain a high-dimensional feature vector; Performing feature matrix synthesis on all the high-dimensional feature vector groups to obtain an initial feature space; Performing sparsity analysis on the initial feature space, calculating correlation coefficients between features, and obtaining a feature correlation matrix; The initial feature space is optimized according to the feature correlation matrix to obtain the high-dimensional feature space.

[0006] Furthermore, the industrial feature extraction and dimensionality reduction of the high-dimensional feature space to obtain a low-dimensional feature vector includes: Iteratively centralize the high-dimensional feature space to obtain a centralized industrial data matrix; Calculate a time-series weighted covariance matrix based on the centralized industrial data matrix, wherein weight coefficients are dynamically allocated according to the time sequence of the data using an exponential decay function; Performing eigenvalue decomposition on the time series weighted covariance matrix to obtain an eigenvalue sequence and a corresponding eigenvector group; Arrange the eigenvector group in descending order according to the eigenvalue sequence, and select eigenvectors to construct an eigenvector matrix in combination with a preset variance threshold during the sorting process; Performing a tensor product operation on the centralized industrial data matrix and the feature vector matrix to obtain an initial industrial feature vector; Performing nonlinear dimensionality reduction processing on the initial industrial feature vector to obtain a nonlinear vector; According to the distribution characteristics of the nonlinear vector, the nonlinear vector is normalized to obtain a standard feature vector; The standard feature vector is subjected to multi-level feature screening in combination with the variance threshold to obtain the low-dimensional industrial feature vector.

[0007] Furthermore, the low-dimensional feature vectors are preliminarily grouped according to a preset clustering algorithm to obtain a first clustering result, including: Calculating the Euclidean distance between the low-dimensional feature vectors to obtain a distance matrix, applying a hierarchical grouping method to the distance matrix to generate a grouping tree structure; The grouping tree structure is cut according to a preset dynamic cutting method to obtain an initial grouping center; Based on the initial grouping center, the low-dimensional feature vectors are grouped according to a preset iterative mean grouping method to obtain an initial grouping result; Performing a silhouette coefficient analysis on the initial grouping result to obtain a grouping quality evaluation index; The initial grouping result is iteratively optimized according to the grouping quality evaluation index to obtain the first clustering result.

[0008] Further, the second adjustment of the first clustering result to obtain the second clustering result includes: Performing intra-cluster sample calculation on each cluster in the first clustering result to obtain intra-cluster aggregation degree and inter-cluster separation degree; The intra-cluster aggregation degree and the inter-cluster separation degree are evaluated according to a preset clustering threshold to obtain a cluster to be adjusted, and the first clustering result is clustered according to the cluster to be adjusted to obtain an unadjusted cluster; Re-clustering the clusters to be adjusted according to a preset local clustering algorithm to obtain a local clustering result; Merging the local clustering result with the unadjusted clustering cluster to form an adjusted clustering result; Calculating the membership of the adjusted clustering result according to a preset mean clustering algorithm to obtain a sample membership matrix; Performing sample allocation on the adjusted clustering result according to the sample membership matrix to obtain an allocated clustering result; Performing boundary sample identification on the distribution clustering result to obtain boundary samples and non-boundary samples; The boundary samples are locally optimized according to a preset nearest neighbor algorithm to obtain local boundary samples; the local boundary samples are merged with the non-boundary samples to obtain the second clustering result.

[0009] Furthermore, the second clustering result is verified by industrial scenario simulation to obtain a verification result, and the second clustering result is optimized and adjusted according to the verification result to obtain a target clustering combination, including: Performing numerical simulation on the second clustering result according to a preset industrial scene parameter set to obtain multiple groups of simulation data; Calculating a performance indicator matrix of each cluster group of the second clustering result according to the multiple groups of simulation data to obtain a cluster group performance evaluation result; Scoring the cluster group performance evaluation results according to a preset multi-dimensional scoring function to obtain a comprehensive score for each cluster group; Sorting the cluster groups in descending order according to the comprehensive scores of the cluster groups to obtain a cluster group sorting result; Sorting the cluster group sorting results according to a preset crossover mutation algorithm to generate multiple candidate cluster combination schemes; Performing industrial scenario numerical simulations on a plurality of candidate cluster combination schemes respectively to obtain a plurality of sets of simulation performance data; Calculating the comprehensive performance index vector of each candidate cluster combination scheme according to the multiple sets of simulation performance data to obtain the candidate scheme evaluation result; Cluster selection is performed according to the candidate scheme evaluation results to obtain the target cluster combination.

[0010] The present invention also provides an intelligent industrial data clustering and grouping system, which is applied to any of the above intelligent industrial data clustering and grouping methods, including: An acquisition module, which is used to perform standardized preprocessing on the collected raw industrial data to obtain standard industrial data, and to construct a high-dimensional feature space based on the standard industrial data; An analysis module, the analysis module is used to extract industrial features and reduce the dimension of the high-dimensional feature space to obtain a low-dimensional feature vector; An association module, the association module is used to preliminarily group the low-dimensional feature vectors according to a preset clustering algorithm to obtain a first clustering result; A processing module, the processing module is used to perform a second adjustment on the first clustering result to obtain a second clustering result; A control module is used to perform industrial scenario simulation verification on the second clustering result to obtain a verification result, and optimize and adjust the second clustering result according to the verification result to obtain a target clustering combination.

[0011] The present invention also provides an intelligent industrial data clustering and grouping device, comprising: Memory, used to store programs; The processor is used to execute the program to implement each step of any one of the above-mentioned methods for intelligent industrial data clustering and grouping.

[0012] The present invention also provides a storage medium storing computer instructions, wherein the computer instructions are used to enable a computer to perform any of the above methods.

[0013] The present invention provides the following beneficial effects: By standardizing and preprocessing the original industrial data, the missing value and noise problems in the data can be effectively solved, and the accuracy and reliability of subsequent analysis can be improved. The principal component analysis method is used to reduce the dimensionality of the high-dimensional feature space, which can effectively reduce the data dimension and reduce the processing complexity while retaining key feature information. Preliminary grouping is performed through the preset clustering algorithm, and secondary adjustment is performed in combination with the classification rules to obtain more reasonable clustering results, which improves the accuracy and applicability of the clustering effect. The second clustering result is verified by industrial scenario simulation and optimized according to the verification results to ensure the effectiveness and practicality of the clustering results in actual industrial scenarios. BRIEF DESCRIPTION OF THE DRAWINGS

[0014] Figure 1 The present invention provides a flow chart of a method for intelligent industrial data clustering and grouping; Figure 2 It is a system structure diagram for providing an intelligent industrial data clustering grouping for the present invention; Figure 3 The present invention provides a device structure diagram for intelligent industrial data clustering and grouping.

[0015] The realization of the purpose, functional features and advantages of the present invention will be further explained in conjunction with embodiments and with reference to the accompanying drawings. DETAILED DESCRIPTION

[0016] In order to make the purpose, technical solution and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention.

[0017] The present invention is further described below in conjunction with the accompanying drawings and specific implementation methods.

[0018] Reference Figure 1 As shown, the present invention provides an intelligent industrial data clustering and grouping method, comprising: Step S1: performing standardized preprocessing on the collected original industrial data to obtain standard industrial data, and constructing a high-dimensional feature space based on the standard industrial data; Step S2: extracting industrial features and reducing the dimension of the high-dimensional feature space to obtain a low-dimensional feature vector; Step S3: Preliminarily grouping the low-dimensional feature vectors according to a preset clustering algorithm to obtain a first clustering result; Step S4: performing a second adjustment on the first clustering result to obtain a second clustering result; Step S5: Perform industrial scenario simulation verification on the second clustering result to obtain a verification result, and optimize and adjust the second clustering result based on the verification result to obtain a target clustering combination.

[0019] As described in the above steps, the detailed process is as follows: Step S1: Obtain raw industrial data from various industrial equipment, sensors or systems, remove outliers, fill in missing values, and unify the data format.

[0020] Normalize or standardize the data so that it falls within a certain standard range to eliminate dimensional differences.

[0021] Extract industry-related features from the standardized data and construct a high-dimensional feature space. Features can include temperature, pressure, flow, and other parameters. Step S2: Calculate the covariance matrix of the high-dimensional feature space, perform eigenvalue decomposition on the covariance matrix, and obtain eigenvectors and eigenvalues.

[0022] Select the principal components that can explain most of the data variance, sort them according to the size of the eigenvalues, and select the first k principal components.

[0023] The original high-dimensional feature data is projected onto the selected principal components to obtain a low-dimensional feature vector.

[0024] Step S3: Select a suitable clustering algorithm, such as K-means, DBSCAN or hierarchical clustering.

[0025] Set the parameters of the clustering algorithm, such as the number of clusters k for K-means, the minimum number of samples and neighborhood radius for DBSCAN, etc.

[0026] Input the low-dimensional feature vector into the clustering algorithm, perform clustering operations, and obtain preliminary grouping results, i.e. the first clustering results Step S4: According to the requirements of the industrial application scenario, set the classification rules, such as the maximum or minimum number of samples for each cluster, the threshold of specific features, etc.

[0027] According to the classification rules, the first clustering result is adjusted.

[0028] If the number of samples in some clusters is too small, similar clusters can be merged.

[0029] If some clusters have too many samples, they are further split based on specific features.

[0030] Redistribute the samples that do not meet the rules.

[0031] Step S5: simulating the application effect of the second clustering result in an industrial scenario.

[0032] Use industrial simulation systems or actual production environments for verification and evaluate the rationality and effectiveness of clustering results. Indicators may include production efficiency, failure rate, energy consumption, etc.

[0033] According to the simulation verification results, the second clustering result is optimized and adjusted, and the parameters or classification rules of the clustering algorithm are adjusted if necessary.

[0034] Repeated simulation verification and adjustment are carried out until the expected results are achieved, and the optimized target clustering combination is obtained and applied in actual production.

[0035] The present invention provides an intelligent industrial data clustering and grouping method, which can effectively solve the missing value and noise problems in the data by standardizing and preprocessing the original industrial data, and improve the accuracy and reliability of subsequent analysis. The principal component analysis method is used to reduce the dimensionality of the high-dimensional feature space, which can effectively reduce the data dimension and reduce the processing complexity while retaining key feature information. Preliminary grouping is performed by a preset clustering algorithm, and secondary adjustment is performed in combination with classification rules to obtain more reasonable clustering results, thereby improving the accuracy and applicability of the clustering effect. The second clustering result is verified by industrial scenario simulation and optimized according to the verification result, so as to ensure the effectiveness and practicality of the clustering result in the actual industrial scenario.

[0036] In one embodiment, the collected raw industrial data is subjected to standardized preprocessing to obtain standard industrial data, and a high-dimensional feature space is constructed based on the standard industrial data, including: The original industrial data is cleaned to remove outliers and missing values ​​to obtain cleaned data, and the cleaned data is normalized to unify data of different dimensions to the same scale to obtain normalized industrial data.

[0037] The normalized industrial data is standardized to obtain standardized industrial data.

[0038] According to the standardized industrial data, each data point is mapped using a preset feature mapping function (such as polynomial mapping, kernel function mapping, etc.) to obtain a high-dimensional feature vector.

[0039] All high-dimensional feature vectors are synthesized to form a feature matrix to obtain the initial feature space Perform sparsity analysis on the initial feature space, calculate the correlation coefficients between features (such as Pearson correlation coefficient, Spearman correlation coefficient, etc.), and construct a feature correlation matrix.

[0040] According to the feature correlation matrix, feature selection is performed to eliminate redundant and highly correlated features, retain representative and unrelated features, optimize the initial feature space, and combine the optimized feature sets to form the final high-dimensional feature space.

[0041] This embodiment ensures the accuracy and completeness of the data by removing outliers and missing values ​​through the data cleaning step, providing a reliable basis for subsequent data analysis. Normalization processing unifies data of different dimensions to the same scale, eliminates the impact of dimensional differences on data analysis, and improves the comparability and analysis effect of the data. Standardization processing makes the data conform to the standard normal distribution, helps to improve the stability of the data, and enhances the effect of data analysis and modeling. Through the feature mapping function, the standardized industrial data is mapped to the high-dimensional feature space, which enriches the feature information, improves the expression ability and prediction accuracy. Feature correlation analysis and sparsity analysis can effectively identify and eliminate redundant and highly correlated features, retain representative and unrelated features, and simplify the feature space.

[0042] In one embodiment, industrial feature extraction and dimensionality reduction are performed on the high-dimensional feature space to obtain a low-dimensional industrial feature vector, including: Iteratively centralize the high-dimensional feature space to obtain a centralized industrial data matrix; The time-series weighted covariance matrix is ​​calculated based on the centralized industrial data matrix, where the weight coefficients are dynamically allocated according to the time sequence of the data using an exponential decay function; Perform eigenvalue decomposition on the time series weighted covariance matrix to obtain the eigenvalue sequence and the corresponding eigenvector group; Arrange the eigenvector group in descending order according to the eigenvalue sequence, and select the eigenvectors to construct the eigenvector matrix in combination with the preset variance threshold during the sorting process; Perform tensor product operation on the centralized industrial data matrix and the eigenvector matrix to obtain the initial industrial eigenvector; Perform nonlinear dimension reduction processing on the initial industrial feature vector to obtain a nonlinear vector; According to the distribution characteristics of the nonlinear vector, the nonlinear vector is normalized to obtain the standard eigenvector; The standard feature vector is screened at multiple levels in combination with the variance threshold to obtain a low-dimensional industrial feature vector.

[0043] This embodiment can effectively eliminate the offset effect of data and improve the accuracy of subsequent feature extraction and dimensionality reduction by iteratively centralizing the high-dimensional feature space. The time-series weighted covariance matrix is ​​adopted, and the weight coefficient is dynamically allocated through the exponential decay function, which not only takes into account the time sequence of the data, but also enhances the sensitivity of the model to the changes in historical data, and improves the timeliness and reliability of feature extraction. By performing eigenvalue decomposition on the covariance matrix, and combining the preset variance threshold for descending arrangement and screening, the representativeness and effectiveness of the feature vector are ensured, and redundant information is effectively reduced. The tensor product operation of the centralized data matrix and the feature vector matrix can quickly obtain the initial industrial feature vector, thereby simplifying the calculation process. Through nonlinear dimensionality reduction and normalization processing, not only can the feature dimension be further reduced, but also the feature vector can have good distribution characteristics and standardized characteristics, and the generalization ability of the model is enhanced. Finally, through multi-level feature screening, combined with the variance threshold, the accuracy and effectiveness of the low-dimensional industrial feature vector are ensured, thereby improving the overall efficiency and performance of feature extraction.

[0044] In one embodiment, the low-dimensional feature vectors are preliminarily grouped according to a preset clustering algorithm to obtain a first clustering result, including: Calculate the Euclidean distance between low-dimensional feature vectors to obtain a distance matrix, apply a hierarchical grouping method to the distance matrix, and generate a grouping tree structure; The grouping tree structure is cut according to a preset dynamic cutting method to obtain the initial grouping center; Based on the initial grouping center, the low-dimensional feature vectors are grouped according to the preset iterative mean grouping method to obtain the initial grouping result; Perform silhouette coefficient analysis on the initial grouping results to obtain grouping quality evaluation indicators; According to the grouping quality evaluation index, the initial grouping result is iteratively optimized to obtain the first clustering result.

[0045] This embodiment can effectively handle the complexity of high-dimensional data by preliminarily grouping low-dimensional feature vectors according to a preset clustering algorithm, making clustering more efficient and accurate. First, by calculating the Euclidean distance between low-dimensional feature vectors and generating a distance matrix, the quantification of the similarity of data points can be achieved, which helps to improve the accuracy of grouping. Secondly, the hierarchical grouping method is applied to the distance matrix to generate a grouping tree structure, which can more intuitively display the hierarchical relationship between data points and provide a solid foundation for subsequent dynamic cutting and iterative mean grouping. The initial grouping results are analyzed by contour coefficient, which can quantify the quality of the grouping results and provide a scientific basis for subsequent iterative optimization. By iteratively optimizing the initial grouping results based on the grouping quality evaluation index, not only can the accuracy and stability of the grouping be further improved, but also the local optimal problem can be effectively avoided, and the first clustering result finally obtained has higher reliability and practicality.

[0046] In one embodiment, performing a secondary adjustment on the first clustering result to obtain a second clustering result includes: For each cluster in the first clustering result, the intra-cluster sample calculation is performed to obtain the intra-cluster aggregation degree and the inter-cluster separation degree.

[0047] The intra-cluster aggregation degree and the inter-cluster separation degree are evaluated according to a preset clustering threshold to obtain a cluster to be adjusted, and the first clustering result is clustered according to the cluster to be adjusted to obtain an unadjusted cluster.

[0048] The clusters to be adjusted are re-clustered according to the preset local clustering algorithm to obtain local clustering results.

[0049] The local clustering results are merged with the unadjusted clustering clusters to form the adjusted clustering results.

[0050] The membership of the clustering results is calculated and adjusted according to the preset mean clustering algorithm to obtain the sample membership matrix.

[0051] The adjusted clustering results are allocated samples according to the sample membership matrix to obtain the allocated clustering results.

[0052] The boundary samples are identified on the distribution clustering results to obtain boundary samples and non-boundary samples.

[0053] The boundary samples are locally optimized according to the preset nearest neighbor algorithm to obtain local boundary samples.

[0054] The local boundary samples are merged with the non-boundary samples to obtain the second clustering result.

[0055] This embodiment can accurately identify the clustering clusters that need to be further optimized according to the evaluation of the intra-cluster aggregation degree and the inter-cluster separation degree by making a secondary adjustment to the first clustering result, thereby improving the accuracy and reliability of the clustering result. The clustering clusters to be adjusted are re-clustered by the local clustering algorithm, and the complex data structure can be processed more carefully, so that the local clustering results are more in line with the actual data distribution. The local clustering results are merged with the unadjusted clustering clusters, which effectively retains the existing reasonable clustering structure, optimizes the part that needs to be adjusted, and improves the rationality of the overall clustering results. The membership of the adjusted clustering results is calculated using the preset mean clustering algorithm, and the samples are allocated through the sample membership matrix to ensure the scientificity and accuracy of the sample allocation. The boundary sample identification of the allocated clustering results can effectively distinguish the boundary samples from the non-boundary samples, so that the processing of the boundary samples is more accurate. The boundary samples are locally optimized by the nearest neighbor algorithm, which further improves the classification accuracy of the boundary samples and reduces the possibility of misclassification. Finally, the local boundary samples are merged with the non-boundary samples, and the second clustering result is not only more accurate overall, but also better reflects the true distribution of the data, significantly improving the performance and effect of clustering.

[0056] In one embodiment, an industrial scenario simulation verification is performed on the second clustering result to obtain a verification result, and the second clustering result is optimized and adjusted according to the verification result to obtain a target clustering combination, including: The second clustering results are numerically simulated according to the preset industrial scenario parameter set to obtain multiple groups of simulation data.

[0057] The performance indicator matrix of each clustering group of the second clustering result is calculated according to the multiple groups of simulation data to obtain the clustering group performance evaluation result.

[0058] The cluster group performance evaluation results are scored according to the preset multi-dimensional scoring function to obtain a comprehensive score for each cluster group. The scoring function can be customized according to different industrial scenarios and needs, and may include multiple dimensions such as efficiency, cost, and stability.

[0059] The cluster groups are sorted in descending order according to their comprehensive scores to obtain the cluster group sorting results.

[0060] The cluster group sorting results are sorted according to the preset crossover mutation algorithm to generate multiple candidate cluster combination schemes. The crossover mutation algorithm is an optimization strategy based on the genetic algorithm. It generates new candidate schemes through crossover and mutation operations. These schemes may contain better combinations.

[0061] Numerical simulations of industrial scenarios are performed on multiple candidate clustering combination schemes to obtain multiple groups of simulation performance data.

[0062] The comprehensive performance index vector of each candidate clustering combination scheme is calculated according to multiple groups of simulated performance data to obtain the candidate scheme evaluation results.

[0063] Cluster selection is performed based on the candidate scheme evaluation results to obtain the target cluster combination.

[0064] This embodiment can effectively evaluate and optimize the clustering results based on actual application requirements by simulating and verifying the second clustering results in an industrial scenario, thereby improving the practicality and accuracy of the clustering results. First, by performing numerical simulations with a preset industrial scenario parameter set, the clustering results can be verified multiple times in a virtual environment, saving time and cost for actual operations. Secondly, the clustering groups are evaluated and scored in detail using a performance indicator matrix and a multi-dimensional scoring function to ensure that the performance of each clustering group is comprehensively and objectively analyzed. Then, a variety of candidate clustering combination schemes are generated through a crossover mutation algorithm, and they are simulated and verified again to ensure that the final selected clustering combination scheme has excellent performance in a variety of scenarios. Finally, cluster selection is performed based on the candidate scheme evaluation results, and the target clustering combination obtained not only has higher comprehensive performance, but also can better adapt to the needs of industrial scenarios. This process greatly improves the reliability and application value of the clustering results, and enhances the adaptability and operational efficiency of the system in different industrial environments.

[0065] Reference Figure 2 As shown, the present invention also provides a system for intelligent industrial data clustering and grouping, which is applied to any of the above intelligent industrial data clustering and grouping methods, including: The acquisition module is used to perform standardized preprocessing on the collected raw industrial data to obtain standard industrial data, and to construct a high-dimensional feature space based on the standard industrial data; An analysis module, which is used to extract industrial features and reduce the dimension of a high-dimensional feature space using a preset principal component analysis method to obtain a low-dimensional feature vector; An association module, the association module is used to apply a preset clustering algorithm to the low-dimensional feature vectors for preliminary grouping to obtain a first clustering result; A processing module, the processing module is used to make a second adjustment to the first clustering result in combination with a preset classification rule to obtain a second clustering result; The control module is used to perform industrial scenario simulation verification on the second clustering result to obtain a verification result, and optimize and adjust the second clustering result according to the verification result to obtain a target clustering combination.

[0066] The present invention provides an intelligent industrial data clustering and grouping system, which can effectively solve the missing value and noise problems in the data by standardizing and preprocessing the original industrial data, and improve the accuracy and reliability of subsequent analysis. The principal component analysis method is used to reduce the dimensionality of the high-dimensional feature space, which can effectively reduce the data dimension and reduce the processing complexity while retaining key feature information. Preliminary grouping is performed through a preset clustering algorithm, and secondary adjustment is performed in combination with classification rules to obtain a more reasonable clustering result, thereby improving the accuracy and applicability of the clustering effect. The second clustering result is verified by industrial scenario simulation, and optimized and adjusted according to the verification result, to ensure the effectiveness and practicality of the clustering result in the actual industrial scenario.

[0067] Reference Figure 3 As shown, the present invention also provides an intelligent industrial data clustering and grouping device, comprising: Memory, used to store programs; The processor is used to execute the program to implement each step of any one of the above-mentioned intelligent industrial data clustering and grouping methods.

[0068] In this embodiment, the processor and the memory may be connected via a bus or other means. The memory may include a volatile memory, such as a random access memory; the memory may also include a non-volatile memory, such as a read-only memory, a flash memory, a hard disk, or a solid-state drive. The processor may be a general-purpose processor, such as a central processing unit, a digital signal processor, an application-specific integrated circuit, or one or more integrated circuits configured to implement the embodiments of the present invention.

[0069] The present invention also provides a storage medium storing computer instructions, wherein the computer instructions are used to enable a computer to perform any of the above methods.

[0070] It should be noted that technicians in the relevant technical field can clearly understand that for the convenience and conciseness of description, the specific working process of the system and each module described above can refer to the corresponding process in the aforementioned method embodiment, and will not be repeated here.

[0071] The above description is only a preferred embodiment of the present invention, and does not limit the patent scope of the present invention. Any equivalent structure or equivalent process transformation made by using the contents of the present invention specification and drawings, or directly or indirectly applied in other related technical fields, are also included in the patent protection scope of the present invention.

Claims

1. An intelligent industrial data clustering and grouping method, characterized in that: include: Performing standardized preprocessing on the collected raw industrial data to obtain standard industrial data, and constructing a high-dimensional feature space based on the standard industrial data; Performing industrial feature extraction and dimensionality reduction on the high-dimensional feature space to obtain a low-dimensional feature vector; Preliminarily grouping the low-dimensional feature vectors according to a preset clustering algorithm to obtain a first clustering result; Perform a secondary adjustment on the first clustering result to obtain a second clustering result; The second clustering result is verified by industrial scenario simulation to obtain a verification result, and the second clustering result is optimized and adjusted according to the verification result to obtain a target clustering combination.

2. The intelligent industrial data clustering and grouping method according to claim 1 is characterized in that: The collected raw industrial data is subjected to standardization preprocessing to obtain standard industrial data, and a high-dimensional feature space is constructed based on the standard industrial data, including: Performing data cleaning on the original industrial data, removing outliers and missing values ​​to obtain cleaned data, and normalizing the cleaned data to unify data of different dimensions to the same scale to obtain normalized industrial data; Performing standardization processing on the normalized industrial data to obtain standardized industrial data; According to the standardized industrial data, each data point is mapped using a preset feature mapping function to obtain a high-dimensional feature vector; Performing feature matrix synthesis on all the high-dimensional feature vector groups to obtain an initial feature space; Performing sparsity analysis on the initial feature space, calculating correlation coefficients between features, and obtaining a feature correlation matrix; The initial feature space is optimized according to the feature correlation matrix to obtain the high-dimensional feature space.

3. The intelligent industrial data clustering and grouping method according to claim 1 is characterized in that: The step of extracting industrial features and reducing the dimension of the high-dimensional feature space to obtain a low-dimensional feature vector includes: Iteratively centralize the high-dimensional feature space to obtain a centralized industrial data matrix; Calculate a time-series weighted covariance matrix based on the centralized industrial data matrix, wherein weight coefficients are dynamically allocated according to the time sequence of the data using an exponential decay function; Performing eigenvalue decomposition on the time series weighted covariance matrix to obtain an eigenvalue sequence and a corresponding eigenvector group; Arrange the eigenvector group in descending order according to the eigenvalue sequence, and select eigenvectors to construct an eigenvector matrix in combination with a preset variance threshold during the sorting process; Performing a tensor product operation on the centralized industrial data matrix and the feature vector matrix to obtain an initial industrial feature vector; Performing nonlinear dimensionality reduction processing on the initial industrial feature vector to obtain a nonlinear vector; According to the distribution characteristics of the nonlinear vector, the nonlinear vector is normalized to obtain a standard feature vector; The standard feature vector is subjected to multi-level feature screening in combination with the variance threshold to obtain the low-dimensional industrial feature vector.

4. The intelligent industrial data clustering and grouping method according to claim 1, characterized in that: The low-dimensional feature vectors are preliminarily grouped according to a preset clustering algorithm to obtain a first clustering result, including: Calculating the Euclidean distance between the low-dimensional feature vectors to obtain a distance matrix, applying a hierarchical grouping method to the distance matrix to generate a grouping tree structure; Cutting the grouping tree structure according to a preset dynamic cutting method to obtain an initial grouping center; Based on the initial grouping center, the low-dimensional feature vectors are grouped according to a preset iterative mean grouping method to obtain an initial grouping result; Performing a silhouette coefficient analysis on the initial grouping result to obtain a grouping quality evaluation index; The initial grouping result is iteratively optimized according to the grouping quality evaluation index to obtain the first clustering result.

5. The intelligent industrial data clustering and grouping method according to claim 1, characterized in that: The second step of performing a secondary adjustment on the first clustering result to obtain a second clustering result includes: Performing intra-cluster sample calculation on each cluster in the first clustering result to obtain intra-cluster aggregation degree and inter-cluster separation degree; The intra-cluster aggregation degree and the inter-cluster separation degree are evaluated according to a preset clustering threshold to obtain a cluster to be adjusted, and the first clustering result is clustered according to the cluster to be adjusted to obtain an unadjusted cluster; Re-clustering the clusters to be adjusted according to a preset local clustering algorithm to obtain a local clustering result; Merging the local clustering result with the unadjusted clustering cluster to form an adjusted clustering result; Calculating the membership of the adjusted clustering result according to a preset mean clustering algorithm to obtain a sample membership matrix; Performing sample allocation on the adjusted clustering result according to the sample membership matrix to obtain an allocated clustering result; Performing boundary sample identification on the distribution clustering result to obtain boundary samples and non-boundary samples; The boundary samples are locally optimized according to a preset nearest neighbor algorithm to obtain local boundary samples; the local boundary samples are merged with the non-boundary samples to obtain the second clustering result.

6. The intelligent industrial data clustering and grouping method according to claim 1, characterized in that: The performing industrial scenario simulation verification on the second clustering result to obtain a verification result, and optimizing and adjusting the second clustering result according to the verification result to obtain a target clustering combination, includes: Performing numerical simulation on the second clustering result according to a preset industrial scene parameter set to obtain multiple groups of simulation data; Calculating a performance indicator matrix of each cluster group of the second clustering result according to the multiple groups of simulation data to obtain a cluster group performance evaluation result; Scoring the cluster group performance evaluation results according to a preset multi-dimensional scoring function to obtain a comprehensive score for each cluster group; Sorting the cluster groups in descending order according to the comprehensive scores of the cluster groups to obtain a cluster group sorting result; Sorting the cluster group sorting results according to a preset crossover mutation algorithm to generate multiple candidate cluster combination schemes; Performing industrial scenario numerical simulations on a plurality of candidate cluster combination schemes respectively to obtain a plurality of sets of simulation performance data; Calculating the comprehensive performance index vector of each candidate cluster combination scheme according to the multiple sets of simulation performance data to obtain the candidate scheme evaluation result; Cluster selection is performed according to the candidate scheme evaluation results to obtain the target cluster combination.

7. An intelligent industrial data clustering and grouping system, characterized in that: The method for intelligent industrial data clustering and grouping applied to any one of claims 1 to 6 above comprises: An acquisition module, which is used to perform standardized preprocessing on the collected raw industrial data to obtain standard industrial data, and to construct a high-dimensional feature space based on the standard industrial data; An analysis module, the analysis module is used to extract industrial features and reduce the dimension of the high-dimensional feature space to obtain a low-dimensional feature vector; An association module, the association module is used to preliminarily group the low-dimensional feature vectors according to a preset clustering algorithm to obtain a first clustering result; A processing module, the processing module is used to perform a second adjustment on the first clustering result to obtain a second clustering result; A control module is used to perform industrial scenario simulation verification on the second clustering result to obtain a verification result, and optimize and adjust the second clustering result according to the verification result to obtain a target clustering combination.

8. An intelligent industrial data clustering and grouping device, characterized in that: include: Memory, used to store programs; The processor is used to execute the program to implement each step of the method for intelligent industrial data clustering and grouping as described in any one of claims 1 to 6.

9. A storage medium, characterized in that: Computer instructions are stored, and the computer instructions are used to make a computer execute the method according to any one of claims 1 to 6.

Images