Industrial equipment monitoring task tree automatic construction method based on heterogeneous data sources

Abstract

The application discloses an industrial equipment manufacturing supervision task tree automatic construction method based on heterogeneous data sources, relates to the technical field of industrial equipment manufacturing supervision and information management, and comprises the following steps: mapping manufacturing supervision data from multiple heterogeneous data sources into a manufacturing supervision semantic entity set; detecting semantic conflicts according to the manufacturing supervision semantic entity set, and generating corresponding manufacturing supervision task nodes when conflicts are detected; determining parent-child hierarchical relationships between manufacturing supervision task nodes according to evolution paths of the manufacturing supervision semantic entity set in time and process dimensions, and constructing a hierarchical manufacturing supervision task tree; and during manufacturing supervision task execution, reconstructing the manufacturing supervision task tree in a structural level according to newly acquired manufacturing supervision data; the application solves the problems of inconsistent manufacturing supervision data semantics, task structure dependence on manual maintenance and difficulty in timely reflecting actual manufacturing supervision conditions.

Description

Technical Field

[0001] This invention relates to the field of industrial equipment manufacturing supervision and information management technology, and more specifically, to a method for automatically constructing an industrial equipment manufacturing supervision task tree based on heterogeneous data sources. Background Technology

[0002] As the scale of industrial equipment manufacturing continues to expand, the process of industrial equipment supervision typically requires processing data from multiple business systems simultaneously, such as supervision planning systems, task execution record systems, problem management systems, and personnel management systems. These systems differ significantly in data structure, data granularity, and data semantics, resulting in multiple heterogeneous data sources.

[0003] In existing technologies, the management methods for supervising the manufacture of industrial equipment mainly employ the following two approaches: On the one hand, data integration or data platform technology can be used to centrally store and display multi-source data. However, such technology mainly focuses on data aggregation and querying, and lacks the ability to automatically understand the business logic of the supervision business. The supervision tasks still need to be manually broken down and configured. On the other hand, the supervision tasks are managed through workflow or project management systems. These systems usually rely on manual pre-definition of task structures or process nodes, and cannot automatically generate or adjust the task hierarchy structure according to changes in supervision data from heterogeneous data sources.

[0004] Because the existing technologies mentioned above have failed to effectively solve the problem of semantic inconsistency in heterogeneous supervision data, the supervision task structure relies on manual maintenance, resulting in problems such as low task generation efficiency, rigid task structure, and difficulty in timely reflecting the actual situation of on-site supervision, which in turn affects the efficiency and quality of supervision work. Summary of the Invention

[0005] To overcome the aforementioned deficiencies of the prior art, embodiments of the present invention provide an automatic construction method for industrial equipment supervision task trees based on heterogeneous data sources, which solves the problems of semantic inconsistency in heterogeneous supervision data, reliance on manual maintenance of task structures, and difficulty in timely reflecting the actual situation of on-site supervision.

[0006] To achieve the above objectives, the present invention provides the following technical solution: This application provides an automatic construction method for an industrial equipment supervision task tree based on heterogeneous data sources. The method includes: mapping supervision data from multiple heterogeneous data sources into a supervision semantic entity set; detecting semantic conflicts based on the supervision semantic entity set and generating corresponding supervision task nodes when a conflict is detected; determining the parent-child hierarchical relationship between each supervision task node based on the evolution path of the supervision semantic entity set in the time and process dimensions, and constructing a hierarchical supervision task tree; and reconstructing the supervision task tree at the structural level based on newly acquired supervision data during the execution of the supervision task.

[0007] In one embodiment, the step of acquiring supervision data includes: determining a heterogeneous data source set that includes at least supervision plans, supervision execution records, and problem record data sources; identifying the data access methods and data structure characteristics of each heterogeneous data source; collecting the original supervision data and attaching a data source identifier; and generating structured supervision data with a data source identifier after structured processing.

[0008] In one embodiment, the mapping to a supervision semantic entity set includes: constructing a unified semantic description system and a corresponding set of standard semantic attributes based on industrial equipment supervision data; extracting source fields from heterogeneous data sources and constructing field probability distributions; using the optimal transmission model and the Wasserstein distance to measure the difference between the distribution and the standard semantic attribute distribution; determining field matching relationships based on the minimum transmission cost; generating a semantic mapping rule set according to the field matching relationships; performing semantic transformation and aggregation on heterogeneous data; generating supervision objects, processes, and problem semantic entities to constitute the supervision semantic entity set.

[0009] In one embodiment, a unified semantic description system is constructed based on industrial equipment manufacturing supervision data, including: standardizing the supervision data to obtain a structured dataset; extracting candidate attributes and their hierarchical relationships according to three dimensions: supervision object, process information, and problem information; classifying the candidate attributes to obtain semantic attribute categories, and establishing association relationships between attributes within and across categories; setting business constraint rules for semantic attributes, including mandatory fields, uniqueness, value rules, and dependency conditions; and constructing a unified semantic description system based on the semantic attribute categories, association relationships, and constraint rules.

[0010] In one embodiment, detecting semantic conflicts based on the supervised semantic entity set and generating corresponding supervised task nodes when a conflict is detected includes: converting fields from various heterogeneous data sources and attributes from the supervised semantic entity set into multi-dimensional feature vectors, calculating the similarity of each dimension and weighting and fusing them to obtain a comprehensive similarity score; dynamically adjusting the weights of each dimension based on historical mapping records, optimizing the candidate mapping list, and associating the mappings to the hierarchical structure of unified semantic entities according to the field hierarchy, generating a cross-data source semantic entity set, performing semantic consistency detection accordingly, and generating corresponding supervised task nodes when a semantic conflict is detected.

[0011] In one embodiment, semantic consistency detection includes: representing a cross-data source semantic entity set as a graph structure, where nodes represent semantic entities or fields, and edges represent parent-child hierarchical relationships between nodes; performing feature encoding on each node to form a node feature set, and calculating a graph matching matrix based on the node features; based on the graph matching matrix, performing subgraph isomorphism and graph matching analysis on subgraphs from different data sources, comparing the structural consistency of the same semantic entity and its child nodes in different data sources, and when the structural consistency is lower than a preset threshold, marking the corresponding node as a potential semantic conflict node and labeling the conflict type, thereby generating a corresponding supervision task node.

[0012] In one embodiment, based on the evolution path of the supervision semantic entity set in the time and process dimensions, the parent-child hierarchical relationship between each supervision task node is determined, and a hierarchical supervision task tree is constructed. This includes: extracting the evolution path in the time and process dimensions based on the supervision semantic entity set, including the process execution order and dependency relationship; determining the parent-child hierarchical relationship between each supervision task node through causal priority, setting the node corresponding to the top-level supervision object as the root node, recursively constructing a multi-level sub-task node structure, and forming a hierarchical supervision task tree.

[0013] In one embodiment, the parent-child hierarchical relationship between each supervision task node is determined by causal priority, including: defining the executable preconditions for each process and extracting the actual execution triggering status of each process from the supervision data; when each process is triggered, determining whether all its preconditions are met; if all are met, determining that the process providing the preconditions is causally prioritized over the current process, and establishing a causal priority association; performing time consistency verification on the parent-child nodes of the causal priority association, distinguishing between normal execution paths and delayed execution paths; when a process is triggered when the preconditions are not met, identifying it as a causal abnormal path and classifying it; and generating the parent-child hierarchical relationship between supervision task nodes based on the causal priority association and path type.

[0014] In one embodiment, during the execution of the supervision task, the supervision task tree is restructured at the structural level based on the newly acquired supervision data. This includes: dividing the supervision task tree into multiple subtree regions, each with a unique identifier and lock status; acquiring the latest supervision data and mapping it to the supervision semantic entity set; identifying newly added or changed nodes, locking the corresponding subtree region, and updating local task nodes only within that region, including adjusting parent-child hierarchical relationships, updating execution status, inserting new nodes, or modifying node attributes; and unlocking the updated subtree region and integrating it into the original task tree to form a restructured full tree structure.

[0015] As can be seen from the above technical solutions, the embodiments of this application have the following advantages: By mapping monitoring information from multi-source heterogeneous data to a standardized semantic entity set, and comprehensively considering the dual evolution paths of time and process dimensions, the parent-child hierarchical relationship of task nodes is accurately determined through causal priority. Dynamic incremental reconstruction is achieved during task execution, with local adjustments made only for newly added or changed nodes, thereby significantly improving the construction accuracy and execution rationality of the monitoring task tree. At the same time, by combining locking mechanisms, precondition judgment, abnormal path marking, and log tracking, the stability, traceability, and data consistency of the task tree structure can be guaranteed, effectively solving the problems of low global reconstruction efficiency, incorrect parent-child relationship, or uncontrollable abnormal execution in traditional methods. Thus, in large-scale, multi-source, and dynamically updated industrial monitoring scenarios, management intelligence, automation, and reliability are significantly improved. Attached Figure Description

[0016] Figure 1 This is a schematic diagram of the automatic construction method for industrial equipment manufacturing supervision task tree based on heterogeneous data sources provided in the embodiments of this application.

[0017] Figure 2 A line graph illustrating the structural consistency of potentially semantically conflicting nodes provided in this application embodiment. Detailed Implementation

[0018] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of the present invention.

[0019] Reference Figure 1 As shown in the diagram, the automatic construction method for industrial equipment manufacturing supervision task tree based on heterogeneous data sources provided by this invention includes the following steps: S1. Obtain industrial equipment manufacturing supervision data from multiple heterogeneous data sources, construct a unified semantic description system for the industrial equipment manufacturing supervision data, and map the supervision objects, process information and problem information in different heterogeneous data sources into a unified set of supervision semantic entities.

[0020] In this embodiment, industrial equipment manufacturing supervision data from multiple heterogeneous data sources is acquired, including: A set of heterogeneous data sources for construction supervision data collection is determined, including construction supervision plan data sources, construction supervision execution record data sources, and problem record data sources; For different data sources in the heterogeneous data source set, the corresponding data access methods and data structure characteristics are identified respectively. The data structure characteristics include field names, field types and field hierarchical relationships. Based on the data access method, raw monitoring data is collected from each heterogeneous data source, and corresponding data source identification information is attached to the collected data. The original manufacturing supervision data is initially processed into structured data to generate structured industrial equipment manufacturing supervision data with data source identifiers.

[0021] Among these, heterogeneous data sources differ in data structure or data semantics.

[0022] Furthermore, a unified semantic description system is constructed for the industrial equipment manufacturing supervision data, mapping the supervision objects, process information, and problem information from different heterogeneous data sources into a unified set of supervision semantic entities, including: Based on industrial equipment manufacturing supervision data, a unified semantic description system is constructed to describe industrial equipment manufacturing supervision business. The unified semantic description system includes semantic elements for describing the supervision object, process information and problem information. For each semantic element, a corresponding standard semantic attribute set is generated. The standard semantic attribute set includes at least the attribute name, attribute code, data type, value range, unit information, and constraint rules. The standard semantic attribute set is formed based on the correlation between attributes in historical monitoring data. Based on the standard semantic attribute set, fields are extracted from the data tables in the heterogeneous data source to form a source field set; and a field sample set is constructed according to the data records corresponding to the fields. The field sample set is preprocessed, including missing value completion and data type standardization. Specifically, missing values ​​in numeric fields are filled using the mean; missing values ​​in enumerated fields are filled using the mode; and data of different formats are uniformly converted.

[0023] Field probability distributions are constructed based on the preprocessed field sample set. For continuous fields, the kernel density estimation method is used to generate continuous probability density distributions, and for discrete fields, the frequency statistics method is used to generate discrete probability distributions. Based on the probability distribution of the field and the standard distribution corresponding to the standard semantic attribute, an optimal transmission model between the source field and the standard semantic attribute is constructed. The difference between the field values ​​is used as the transmission cost to construct a cost matrix. The elements in the cost matrix represent the Euclidean distance between the source field sample points and the standard semantic attribute sample points. The optimal transmission cost between the source field distribution and the standard semantic attribute distribution is solved based on the optimal transmission model. The overall transmission cost is minimized by solving the transmission matrix, and the Wasserstein Distance is used to measure the distribution difference. Wasserstein Distance is a distance metric used to measure the degree of difference between two probability distributions. It reflects the similarity of two distributions in their overall form by calculating the minimum “transfer cost” required to transform one distribution into another.

[0024] A similarity matrix between the field and the standard semantic attributes is constructed based on the transmission cost, where the similarity is obtained by normalizing the transmission cost; The normalization transformation is as follows:

[0025] In the formula, This represents the distributional similarity between fields, used to characterize the degree of matching between the source field and standard semantic attributes. The value range is (0,1]. The closer the value is to 1, the more similar the semantic distribution of the two fields is. D is the transmission cost between the source field distribution and the standard semantic attribute distribution.

[0026] A field matching optimization model is constructed based on the similarity matrix, wherein the optimization objective is to minimize the overall matching cost, and a linear assignment algorithm is used to determine the correspondence between the source field and the standard semantic attribute. Perform consistency checks on the obtained field matching relationships, including: data type consistency checks and whether the field values ​​fall within the range of standard attribute values, and remove field matching relationships that do not meet the constraints; The data type consistency check involves reading the data type of the source field and the target data type corresponding to the standard semantic attribute, and determining whether they belong to the same data type category. When the data type categories are different and there is a preset type conversion rule, the corresponding data type conversion is performed on the source field; when there is no corresponding type conversion rule, the matching relationship of the field is marked as an abnormal matching relationship and removed.

[0027] A semantic mapping rule set is generated based on the verified field matching relationship, wherein the semantic mapping rule set includes field mapping relationship, field transformation rule and field standardization rule; Based on the semantic mapping rule set, heterogeneous data is semantically transformed to generate unified semantic expression data. The unified semantic expression data is aggregated to generate semantic entities of the supervised object, semantic entities of the process, and semantic entities of the problem, and a set of semantic entities of the supervised object is constructed based on the entity identifier.

[0028] It should be noted that, compared to conventional heterogeneous data semantic mapping techniques that rely on field name matching, rule matching, or cosine similarity calculation, the above-mentioned technical solution constructs a unified semantic description system and introduces an optimal transmission model based on Wasserstein Distance to measure the overall difference between the source field distribution and the standard semantic attribute distribution. This achieves a transformation from "field text matching" to "data distribution semantic matching," effectively reducing mismatch problems caused by inconsistent field naming, abbreviation differences, heterogeneous data sources, and inconsistent field semantic expressions, thus improving the accuracy and robustness of heterogeneous supervision data field mapping. Simultaneously, by uniformly converting supervision objects, process information, and problem information from different heterogeneous data sources into standard semantic entities and constructing relationships between entities, it effectively breaks down data barriers between multi-source supervision data, improving the unified management capabilities, semantic association capabilities, and feasibility of subsequent intelligent analysis and task generation for industrial equipment supervision data.

[0029] Furthermore, based on industrial equipment manufacturing supervision data, a unified semantic description system for describing industrial equipment manufacturing supervision operations is constructed, which also includes: The industrial equipment manufacturing supervision data is standardized to form a structured manufacturing supervision dataset that can be used for semantic analysis. The standardization process includes field name cleaning, data type conversion, and missing value annotation. Based on the structured supervision dataset, candidate attributes and their hierarchical relationships for each field are extracted according to three dimensions: supervision object, process information, and problem information. Candidate attributes are classified according to functional characteristics and business relevance to form semantic attribute categories. The semantic attribute categories include semantic attribute categories for supervised objects, semantic attribute categories for process information, and semantic attribute categories for problem information. For each category, an attribute name, data type, and value range are defined. Within each semantic attribute category, establish parent-child and dependency relationships between attributes, and establish association relationships across categories, such as business associations between supervised objects and processes, and between processes and problems, to form a preliminary semantic hierarchy structure; Set business constraint rules for semantic attribute categories and attributes. The constraint rules include required attributes, uniqueness constraints, value rules, and dependency conditions. Among them, required attributes specify which attributes in the semantic attribute category must provide values ​​during data mapping; otherwise, the system will prompt for missing values ​​or block the mapping. Uniqueness constraints ensure that certain attributes have unique values ​​throughout the entire monitoring dataset, such as equipment numbers or process codes, to avoid duplicate or conflicting mappings. Value rules define the allowed value range, enumeration options, or data formats for attributes; for example, the problem level can only be "high, medium, or low" to ensure data consistency and standardization. Dependency conditions specify the business logic relationships between attributes; for example, if the process status is "completed," a completion time field must be provided to maintain the logical correctness and integrity between attributes.

[0030] Based on semantic attribute categories, semantic hierarchy structure, and constraint rules, a unified semantic description system is constructed. The unified semantic description system includes semantic element categories of supervised objects, process information, and problem information.

[0031] S2, based on the supervision semantic entity set, detects the semantic consistency of the same supervision object in different heterogeneous data sources. When a semantic conflict is detected, a corresponding supervision task node is generated.

[0032] In this embodiment, based on the supervision semantic entity set, the semantic consistency of the same supervision object in different heterogeneous data sources is detected. When a semantic conflict is detected, a corresponding supervision task node is generated, including: The attributes of each heterogeneous data source field and the unified supervised semantic entity set are transformed into multi-dimensional feature vectors. The multi-dimensional feature vectors include field name text features, field data type encoding, field value range and unit features, and field hierarchy and context features. For the generated multidimensional feature vector, the similarity of each dimension is calculated, including text feature similarity, type feature consistency, and value range feature matching degree. Among them, text feature similarity uses cosine similarity to calculate the similarity between field names and target semantic attribute names; type feature consistency uses distance metric to determine the consistency between source field type and semantic attribute type; value range feature matching degree is calculated by normalizing the value range of the source field after unit normalization and comparing it with the allowed value range of the target semantic attribute. The higher the coverage ratio, the higher the matching degree score.

[0033] The similarity scores of each dimension are weighted according to preset weights to generate a comprehensive similarity score; A preliminary list of candidate maps is generated by filtering high-confidence maps, where high confidence means that the overall similarity score is greater than a preset confidence threshold; By introducing historical mapping records and dynamically adjusting the weights of similar fields, the initial mapping list is optimized to obtain the optimized candidate mapping. The process involves calculating the similarity between the current source field to be mapped and fields in historical mapping records based on their name, type, and value range. This identifies the set of mapping records in historical mappings that are most similar to the current field (i.e., have a similarity greater than a set similarity threshold). Weights are dynamically adjusted: if historical mapping records show that a certain dimension plays a dominant role in successful mapping records (i.e., the average contribution rate of that dimension in successful historical mappings exceeds a set threshold), the weight of that dimension is increased by a preset proportion; if historical records show that a certain dimension is prone to mismatches (i.e., the contribution rate of that dimension in failed mapping records exceeds a threshold), the weight of that dimension is decreased by a preset proportion. The adjusted weights are used to calculate the weighted comprehensive score of the similarity of each dimension of the current field, forming the field mapping confidence score. This dynamic weight adjustment ensures that the comprehensive score of the current field mapping is closer to the historical successful pattern, improving mapping accuracy.

[0034] Based on the field hierarchy, the candidate mappings are associated with the hierarchical structure of the unified supervision semantic entity to generate a complete cross-data source semantic entity set; Semantic consistency detection is performed based on semantic entity sets from cross-data source.

[0035] It's important to note that a cross-data source semantic entity set refers to a unified semantic representation set formed by mapping and associating fields from different heterogeneous data sources with a unified supervision semantic entity, and organizing them according to their semantic attributes and parent-child hierarchy. It not only preserves the original characteristics of each data source field (such as name, type, value range, and hierarchical relationship) but also aggregates the corresponding fields of the same semantic entity in different data sources, forming a complete cross-source semantic view. By generating this semantic entity set, centralized management, unified comparison, and semantic consistency detection of the same supervision object across different data sources can be achieved. This enables rapid identification of potential semantic conflicts, guides the generation of supervision task nodes, optimizes mapping accuracy, and provides a standardized framework for cross-data source data fusion, quality control, and subsequent analysis, significantly improving the visualization, controllability, and automation of data management.

[0036] Furthermore, semantic consistency detection based on cross-data source semantic entity sets also includes: Based on the cross-data source semantic entity set, its mapping fields and corresponding unified supervision semantic entities are represented as a graph structure. The graph structure includes nodes and edges, where nodes represent semantic entities or fields, and edges represent the parent-child hierarchical relationship between nodes, thus forming a cross-data source semantic entity set graph. Each node in the cross-data source semantic entity set graph is feature-encoded, including field name, field data type, field value range and field hierarchical position, and these features are mapped into vector representations to form a node feature set; Based on the node feature set, similarity is calculated for nodes from different data sources in the cross-data source semantic entity set graph, generating a graph matching matrix, where each element represents the similarity score between two nodes; Based on the graph matching matrix, subgraph isomorphism and graph matching analysis are performed on the semantic entity set graph across data sources to compare the structural consistency of the same semantic entity and its child nodes in different data sources and identify structural differences or potential semantic conflicts. The subgraph isomorphism or graph matching analysis based on the graph matching matrix refers to constructing a graph matching matrix by analyzing the nodes and their parent-child hierarchical relationships from different data sources in a cross-data source semantic entity graph. Then, structural matching analysis is performed on the subgraphs corresponding to the same semantic entity and its child nodes based on this matrix. Specifically, subgraph isomorphism identifies parts in different data sources where the number of nodes, hierarchical relationships, and dependency structures are completely consistent, while approximate graph matching measures structural similarity with slight differences. This analysis quantifies the degree of structural consistency of the same semantic entity across different data sources, identifying which child nodes have hierarchical differences, are missing, or have extra nodes, thus accurately identifying potential semantic conflicts. When structural consistency is below a preset threshold, relevant nodes are marked as potential conflicts and further classified according to conflict type (such as value range mismatch, type inconsistency, hierarchical logical conflict, etc.), providing a clear basis for generating supervision task nodes. This method ensures that not only the feature matching of individual fields is considered, but also the hierarchical and dependency relationships between fields are preserved and analyzed, thereby improving the completeness and accuracy of semantic consistency detection.

[0037] like Figure 2 As shown, based on the structural consistency assessment results, node combinations with structural consistency below a preset threshold are selected as potential semantic conflict nodes, and conflict types are marked, including value range mismatch, type inconsistency, and hierarchical logical conflict. For each potential semantic conflict node, a corresponding supervision task node is generated.

[0038] It should be noted that by constructing a cross-data source semantic entity set and combining multi-dimensional feature matching and graph structure analysis, the semantic consistency of the same supervised object in heterogeneous data sources can be accurately detected. This can systematically and automatically identify potential conflicts and generate supervision task nodes, thereby significantly improving mapping accuracy, data management efficiency, and cross-source semantic controllability.

[0039] S3. Based on the data evolution path of the supervision semantic entity set in the time dimension and process dimension, determine the parent-child hierarchical relationship between each supervision task node and construct a hierarchical supervision task tree structure.

[0040] In this embodiment, based on the data evolution path of the supervision semantic entity set in the time and process dimensions, the parent-child hierarchical relationship between each supervision task node is determined, and a hierarchical supervision task tree structure is constructed, including: Based on the semantic entity set of supervision, the evolution order of each supervision object and process information is extracted according to the time dimension, including the planned start time, actual execution time, completion time and process execution order. At the same time, in the process dimension, the dependencies between processes and the logical order of preceding and following processes are extracted to form a two-dimensional evolution path of time and process. Based on the evolution path of time and process, the parent-child hierarchical relationship between the supervision task nodes is determined by causal priority; Based on the parent-child hierarchical relationship, the supervision task nodes are recursively organized hierarchically. The supervision task node corresponding to the top-level supervision object is set as the root node, and a multi-level sub-task node structure is constructed downwards in sequence to form a hierarchical supervision task tree structure. For each level node, the task node identifier, parent-child relationship, process information, planned and actual time, execution status, historical adjustment identifier, and causal priority value are recorded.

[0041] Furthermore, based on the dual-dimensional evolution path of time and process, the parent-child hierarchical relationship between supervision task nodes is determined through causal priority, including: The executable prerequisites for each type of process in the supervision semantic entity set are defined. The prerequisites include the completion status of the upstream process, the compliance status of the key quality inspection results, the readiness status of necessary materials or equipment resources, and the necessary supervision confirmation or approval status. The prerequisites are then associated and stored with the corresponding processes in a structured manner. The industrial equipment manufacturing supervision data is analyzed to extract the actual execution trigger status information of each supervised object at each process node. The execution trigger status information includes the process start, execution, completion or interruption status identifier, as well as the corresponding occurrence time information. Based on the execution trigger status information, at the execution trigger time of each process, it is determined whether the corresponding preconditions have been fully met, and a precondition satisfaction determination result corresponding to the execution behavior of that process is generated. If the result of the precondition satisfaction determination for each process is that all preconditions are satisfied, the process that provides the preconditions is determined to have priority over the current process in causal logic. Based on the causal logic priority relationship, establish the causal priority association between the supervision task nodes, including recording the parent node, child node, causal type and triggering condition; Consistency checks are performed on the time information of parent and child nodes in causal priority relationships. If the actual completion time of the parent node is earlier than the execution time of the child node, it is marked as a normal execution path; if the actual completion time of the parent node is later than the execution time of the child node, the causal parent-child relationship remains unchanged, but it is marked as a delayed execution path. When a certain process is detected to be triggered without meeting its prerequisites, it is identified as a causal abnormal path. The abnormal path is then classified into types, including rework path, order insertion path, or abnormal execution path, and the triggering process, resources, and time information are recorded. Based on causal priority association, normal execution path and causal abnormal path, a parent-child hierarchical relationship is generated between the supervision task nodes. The parent node corresponds to the supervision task node that is prioritized in causal logic, and the child node corresponds to the supervision task node that is triggered based on its causal conditions.

[0042] For example, the causal priority associations obtained through the above steps can determine which nodes should logically execute before other nodes; the normal execution path can determine whether the actual execution order conforms to logical causality; and the abnormal path markers indicate those nodes that execute even though they do not meet the preconditions. Combining these three types of information, we can clarify the logical parent node (the node that takes precedence causally) and child nodes (the nodes triggered based on their conditions) of each task node, thereby generating a parent-child hierarchical relationship that conforms to causal logic, providing a reliable basis for constructing a hierarchical supervision task tree.

[0043] It should be noted that the hierarchical task tree structure for supervision constructed using the above technical solution has significant advantages: it comprehensively considers both time and process dimensions of evolutionary paths, and determines the logical parent-child relationship between task nodes through causal priority, effectively solving the problems of incorrect parent-child relationships or incomplete task tree structures that may result from simply relying on the chronological order in traditional methods. Simultaneously, this solution introduces precondition satisfaction judgment, normal execution path marking, and abnormal path classification during node construction, enabling the task tree to not only reflect process dependency logic but also dynamically record abnormal and delayed execution situations, ensuring the integrity, traceability, and rationality of the task tree. Compared with conventional technical methods, this solution combines causal priority with time-process dual-dimensional evolutionary paths, uses a systematic approach to determine parent-child hierarchical relationships, and achieves automation, precision, and dynamic adjustability in task tree construction through structured preconditions and abnormal path analysis, thereby significantly improving the intelligence level and reliability of supervision task management.

[0044] S4. During the execution of the supervision task, when new supervision data is obtained, the task nodes are restructured at the structural level based on the supervision task tree.

[0045] In this embodiment, during the execution of the supervision task, when new supervision data is obtained, the task nodes are restructured at the structural level based on the supervision task tree, including: The supervision task tree is divided into multiple subtree regions according to hierarchy or logical region. Each region contains several supervision task nodes and their parent-child hierarchical relationship. Each region is assigned a unique identifier and lock state so that specific regions can be operated independently without affecting other regions during subsequent update process. During the execution of the supervision task, the latest supervision data from heterogeneous data sources is acquired in real time or periodically. The latest supervision data includes the supervision object, process information, problem information, and corresponding execution status and time information. The latest supervision data is mapped to the supervision semantic entity set of the supervision task tree, and the supervision semantic entity set includes supervision object entities, process entities and problem entities; For new processes or new monitoring objects discovered during the mapping process, corresponding temporary task nodes are generated; Compare the mapped supervisory semantic entities with the original task tree nodes to identify newly added or changed nodes; The subtree region that identifies newly added or changed nodes will be locked to ensure that the region is not affected by other concurrent operations during the update process; Within the locked area, only the subtree containing the newly added or changed node is updated locally. The local task node update includes adjusting the parent-child hierarchy, updating the execution status, inserting new nodes, or modifying node attributes. Adjusting the parent-child hierarchy means redetermining the parent and child nodes of each node based on the newly added process dependencies, preconditions, and causal logic to ensure that the tree structure is logically continuous and free of loops. Updating the execution status means updating the current execution status of the node (such as pending, in execution, completed, or abnormal) after adjusting the parent-child relationship to accurately reflect its actual execution status in the supervision task sequence. Inserting new nodes or modifying node attributes means adding the newly mapped supervision objects or process nodes to the task tree and updating their attribute information, including planned start / completion time, actual execution time, process type, execution status, historical adjustment identifier, and associated supervision semantic entity information, to ensure that the new node is consistent with the original node in terms of parent-child relationship, causal priority, and execution status, thereby forming a logically complete, traceable, and dynamically updatable task tree structure.

[0046] Unlock the updated subtree regions and reintegrate all updated subtree regions into the original task tree to form a reconstructed full tree structure.

[0047] It should be noted that the above-mentioned technical solutions enable dynamic incremental updates of the supervision task tree when acquiring new supervision data. Only the subtrees containing newly added or changed nodes are locally adjusted, thus avoiding the performance overhead and structural oscillations caused by full tree reconstruction. At the same time, by adjusting the parent-child hierarchy, updating the execution status, and inserting new nodes, the causal logic, process dependencies, and execution order of task nodes are kept consistent. In addition, by combining locking mechanisms, subtree integrity verification, and logging, the stability of the task tree structure, the traceability of data, and the manageability of abnormal situations can be guaranteed. This improves the system response efficiency and the reliability of supervision management in large-scale, multi-source, and dynamically updated industrial supervision scenarios.

[0048] The above embodiments can be implemented, in whole or in part, by software, hardware, firmware, or any other combination thereof. When implemented using software, the above embodiments can be implemented, in whole or in part, in the form of a computer program product.

[0049] Those skilled in the art will recognize that the modules and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0050] In addition, the functional modules in the various embodiments of this application can be integrated into one processing module, or each module can exist physically separately, or two or more modules can be integrated into one module.

[0051] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the technical scope disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

[0052] In conclusion, the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A method for automatic construction of an industrial equipment manufacturing task tree based on heterogeneous data sources, characterized in that, include: Map the supervision data from multiple heterogeneous data sources to a set of supervision semantic entities; Semantic conflicts are detected based on the set of supervised semantic entities, and corresponding supervised task nodes are generated when a conflict is detected. Based on the evolution path of the supervision semantic entity set in the time and process dimensions, the parent-child hierarchical relationship between each supervision task node is determined, and a hierarchical supervision task tree is constructed. During the execution of the supervision task, the supervision task tree is restructured at the structural level based on the newly acquired supervision data.

2. The method of claim 1, wherein the method further comprises: The steps for obtaining the monitoring data include: Identify a heterogeneous set of data sources that at least include supervision plans, supervision execution records, and problem records; For each heterogeneous data source, its data access method and data structure characteristics are identified. The original supervision data is collected and data source identifiers are attached. After structuring, structured supervision data with data source identifiers is generated.

3. The method of claim 1, wherein the method further comprises: The mapping is a set of supervised semantic entities, including: Based on industrial equipment manufacturing supervision data, a unified semantic description system and a corresponding set of standard semantic attributes are constructed. Source fields are extracted from heterogeneous data sources and a field probability distribution is constructed. The optimal transmission model is used to measure the difference between this distribution and the standard semantic attribute distribution using Wasserstein distance. Field matching relationships are determined based on the minimum transmission cost. Based on the field matching relationship, a semantic mapping rule set is generated, heterogeneous data is semantically transformed and aggregated, and semantic entities of supervision objects, processes and problems are generated, forming a set of supervision semantic entities.

4. The method of claim 3, wherein the method further comprises: The aforementioned construction of a unified semantic description system based on industrial equipment manufacturing supervision data includes: The monitoring data is standardized to obtain a structured dataset; Candidate attributes and their hierarchical relationships are extracted based on three dimensions: the object under supervision, process information, and problem information. Candidate attributes are classified to obtain semantic attribute categories, and the relationships between attributes are established within and across categories; Set business constraint rules for semantic attributes, including requirements for required fields, uniqueness, value retrieval rules, and dependency conditions; Based on the semantic attribute categories, relationships, and constraint rules, a unified semantic description system is constructed.

5. The method of claim 1, wherein the method further comprises: The step of detecting semantic conflicts based on the set of supervised semantic entities and generating corresponding supervised task nodes when a conflict is detected includes: The fields of each heterogeneous data source and the attributes of the supervised semantic entity set are transformed into multi-dimensional feature vectors. The similarity of each dimension is calculated and weighted and fused to obtain a comprehensive similarity score. Based on historical mapping records, the weights of each dimension are dynamically adjusted to optimize the candidate mapping list. The mappings are then associated with the hierarchical structure of a unified semantic entity according to the field hierarchy, generating a cross-data source semantic entity set. Semantic consistency detection is performed accordingly, and a corresponding supervision task node is generated when a semantic conflict is detected.

6. The method of claim 5, wherein the method further comprises: The semantic consistency detection includes: The semantic entity set across data sources is represented as a graph structure, where nodes represent semantic entities or fields, and edges represent parent-child hierarchical relationships between nodes; Each node is feature-encoded to form a node feature set, and the graph matching matrix is ​​calculated based on the node features; Based on the graph matching matrix, subgraph isomorphism and graph matching analysis are performed on subgraphs from different data sources. The structural consistency of the same semantic entity and its child nodes in different data sources is compared. When the structural consistency is lower than a preset threshold, the corresponding node is marked as a potential semantic conflict node and the conflict type is labeled. Accordingly, the corresponding supervision task node is generated.

7. The method of claim 1, wherein the method further comprises: Based on the evolution path of the supervision semantic entity set in the time and process dimensions, the parent-child hierarchical relationship between each supervision task node is determined, and a hierarchical supervision task tree is constructed, including: Based on the semantic entity set of supervision, the evolution path in the time dimension and process dimension is extracted, including the process execution order and dependency relationship; The parent-child hierarchical relationship between each supervision task node is determined by causal priority. The node corresponding to the top-level supervision object is set as the root node. A multi-level sub-task node structure is recursively constructed to form a hierarchical supervision task tree.

8. The method of claim 7, wherein the method further comprises: The method of determining the parent-child hierarchy between each supervision task node through causal priority includes: Define the executable prerequisites for each process and extract the actual execution trigger status of each process from the monitoring data; When each process is triggered, it is determined whether all its prerequisites are met. If all conditions are met, the process that provides the preconditions is determined to have causal priority over the current process, and a causal priority relationship is established. Perform time consistency checks on parent-child nodes with causal priority relationships to distinguish between normal execution paths and delayed execution paths; When a process is triggered when the prerequisites are not met, it is identified as a causal abnormal path and classified accordingly. Based on causal priority association and path type, a parent-child hierarchical relationship is generated between the supervision task nodes.

9. The method of claim 1, wherein the method further comprises: The process of reconstructing the supervision task tree at the structural level based on newly acquired supervision data during the execution of the supervision task includes: The supervision task tree is divided into multiple subtree regions, each with a unique identifier and lock status. Obtain the latest construction supervision data and map it to the construction supervision semantic entity set; Identify newly added or changed nodes, lock the corresponding subtree area, and only update local task nodes within that area, including adjusting parent-child hierarchy relationships, updating execution status, inserting new nodes, or modifying node attributes; Unlock the updated subtree area and integrate it into the original task tree to form a reconstructed full tree structure.

Images