Robot process data storage method and device

Abstract

The embodiment of the invention provides a robot process data storage method and device. The method comprises the steps that a geometric model file, a process parameter file and an industrial sensor real-time data flow of an industrial robot are collected through a standardized interface; a resolver based on the STEP standard is adopted to carry out feature extraction on the geometric model file, a manufacturing feature entity is obtained, entity extraction is carried out on the technological parameter file according to a set BiLSTM-CRF model, the technological parameter entity subjected to entity extraction is converted into a tree structure according to an XML Schema template, and the manufacturing feature entity is obtained; performing logic constraint on the tree-shaped process parameter structure to obtain a process parameter entity; the method comprises the following steps: constructing a knowledge graph according to a manufacturing feature entity, a process parameter entity and industrial robot equipment, performing time window division on an industrial sensor real-time data stream, calculating data statistical features in each time window, and mapping the data statistical features and parameter nodes in the knowledge graph to obtain a graph database; the process data use efficiency of the industrial robot can be improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of data processing, in particular to a robot process data storage method and device. BACKGROUND

[0002] Industrial robots are increasingly widely used in the field of intelligent manufacturing, and the storage and management of their process data have become a key link to ensure efficient and stable production. However, current robot process data storage faces many challenges, especially in terms of data size, diversity, and processing efficiency.

[0003] On the one hand, with the increasing automation of production lines, robot process data is showing an explosive growth trend. These data not only include geometric topology data, process rule data, but also cover various types of process parameter data, with high heterogeneity and complexity. Traditional local storage methods are difficult to cope with such a large amount of data, and it is difficult to achieve efficient sharing and reuse of data.

[0004] On the other hand, process data generated by robots of different brands and models have different formats, making it difficult to manage and analyze the data uniformly. In addition, the real-time requirements of process data are increasingly high, and how to ensure data security while achieving fast access and efficient processing of data has become a technical problem to be solved.

[0005] Therefore, there is an urgent need for a robot process data storage method to solve the many problems faced by current robot process data storage, to achieve efficient storage, management, and sharing of massive heterogeneous process data, and to improve the efficiency of using process data of industrial robots. SUMMARY

[0006] To solve the problems in the prior art, the present application provides a robot process data storage method and device, which can improve the efficiency of using process data of industrial robots.

[0007] In order to solve at least one of the above problems, the present application provides the following technical solutions: In a first aspect, the present application provides a robot process data storage method, comprising: acquiring multi-source heterogeneous raw data of industrial robots through a standardized interface, wherein the multi-source heterogeneous raw data includes geometric model files, process parameter documents, and real-time data streams of industrial sensors; According to the preset Poisson reconstruction algorithm, the point cloud data in the geometric model file is boundary reconstructed to determine the corresponding standard geometric model file, a parser based on the STEP standard is used to extract features from the standard geometric model file to determine the corresponding manufacturing feature entity, an initial process parameter entity is determined by performing entity extraction on the process parameter document according to a set BiLSTM-CRF model, an XML Schema template is defined, the initial process parameter entity is converted into a tree structure according to the XML Schema template to determine the corresponding tree-shaped process parameter structure, and the tree-shaped process parameter structure is logically constrained according to a preset process condition constraint rule and a preset quality evaluation index to determine the process parameter entity. According to the manufacturing feature entity, the process parameter entity, and the industrial robot device, a knowledge graph is constructed, a plurality of time windows are determined by dividing a time window of the industrial sensor real-time data stream according to a process stage, and a graph database is determined by mapping the data statistical features in each time window to the parameter nodes of the corresponding stage in the knowledge graph, the graph database being used to store robot process data.

[0008] Further, the boundary reconstruction of the point cloud data in the geometric model file according to the preset Poisson reconstruction algorithm to determine the corresponding standard geometric model file comprises: The point cloud data is subjected to normal vector estimation and voxelization processing according to the preset Poisson reconstruction algorithm, and the point cloud data subjected to the voxelization processing is subjected to data fitting according to an implicit function fitting algorithm to determine the corresponding isosurface. The isosurface is subjected to network model extraction according to a marching cubes algorithm to determine the corresponding standard geometric model file.

[0009] Further, before the entity extraction on the process parameter document according to the set BiLSTM-CRF model to determine the corresponding initial process parameter entity, comprising: A process text corpus library adapted to the field is constructed, and the process text corpus library is subjected to process parameter entity annotation according to a hierarchical annotation strategy to determine a process text training set; An initial BiLSTM-CRF model is subjected to model training according to the process text training set to determine the corresponding BiLSTM-CRF model, wherein the BiLSTM layer of the initial BiLSTM-CRF model is provided with a bidirectional gating mechanism, the hidden states of the forward layer and the backward layer are fused through learnable weights, and the CRF layer of the initial BiLSTM-CRF model constrains the label transition rule.

[0010] Further, the definition of the XML Schema template comprises: define a hierarchical relationship of process parameters, wherein the hierarchical relationship of process parameters comprises a process type, a parameter group, and a parameter value; define an XML Schema template according to the hierarchical relationship of process parameters, wherein the XML Schema template comprises a three-layer structure of a process type classification node, a parameter group node, and a parameter value node, wherein the parameter group node is logically grouped according to equipment attributes, material attributes, and environmental attributes, and the parameter value node is bound to a measurement unit and a valid value range attribute.

[0011] Further, the logical constraint of the tree-shaped process parameter structure according to the preset process condition constraint rule and the preset quality evaluation index to determine the corresponding process parameter entity comprises: parsing a conditional statement of the process parameter document according to a preset rule engine, and converting the parsed conditional statement into an XPath process condition constraint rule; constructing a parameter-quality association matrix according to a preset quality evaluation index, and logically constraining the tree-shaped process parameter structure according to the XPath process condition constraint rule and the parameter-quality association matrix to determine the corresponding process parameter entity.

[0012] Further, the construction of the knowledge graph according to the manufacturing feature entity, the process parameter entity, and the industrial robot equipment comprises: creating a feature-parameter relationship edge according to a necessity level and a parameter combination constraint, and linking the manufacturing feature entity to the process parameter entity according to the feature-parameter relationship edge; creating a parameter-equipment relationship edge according to an adaptation score, and linking the process parameter entity to an industrial robot equipment node according to the parameter-equipment relationship edge to determine the corresponding knowledge graph.

[0013] Further, the time window division of the industrial sensor real-time data stream according to the process stage to determine a plurality of time windows, and the relationship mapping of the data statistical features in each time window to the parameter node of the corresponding stage in the knowledge graph to determine the corresponding graph database, comprises: adopting Apache Flink to process the sensor data stream in real time, dividing window data blocks according to the process stage through a window function, extracting statistical features of the sensor data in each window, and determining a corresponding process state vector according to the statistical features; associating the process state vector with the corresponding process node in the knowledge graph to determine the corresponding graph database.

[0014] In a second aspect, the present application provides a robot process data storage device, comprising: The data collection module is configured to collect multi-source heterogeneous original data of the industrial robot through a standardized interface, wherein the multi-source heterogeneous original data comprises a geometric model file, a process parameter document, and an industrial sensor real-time data stream. The manufacturing and process entity determination module is configured to perform boundary reconstruction on point cloud data in the geometric model file according to a preset Poisson reconstruction algorithm, to determine a corresponding standard geometric model file, to perform feature extraction on the standard geometric model file by using a parser based on a STEP standard, to determine a corresponding manufacturing feature entity, to perform entity extraction on the process parameter document according to a set BiLSTM-CRF model, to determine a corresponding initial process parameter entity, to define an XML Schema template, to convert the initial process parameter entity into a tree structure according to the XML Schema template, to determine a corresponding tree-shaped process parameter structure, and to perform logical constraint on the tree-shaped process parameter structure according to a preset process condition constraint rule and a preset quality evaluation index, to determine a corresponding process parameter entity. The graph database determination module is configured to construct a knowledge graph according to the manufacturing feature entity, the process parameter entity, and the industrial robot device, to divide the industrial sensor real-time data stream into a plurality of time windows according to a process stage, to determine a corresponding plurality of time windows, to perform relationship mapping between a data statistical feature in each of the time windows and a parameter node at a corresponding stage in the knowledge graph, and to determine a corresponding graph database, wherein the graph database is configured to store robot process data.

[0015] In a third aspect, the present application provides an electronic device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the steps of the robot process data storage method when executing the program.

[0016] In a fourth aspect, the present application provides a computer readable storage medium, which stores a computer program, wherein the computer program is executable on a processor to implement the steps of the robot process data storage method.

[0017] In a fifth aspect, the present application provides a computer program product, comprising a computer program / instruction, wherein the computer program / instruction is executable on a processor to implement the steps of the robot process data storage method.

[0018] According to the technical solution, the robot process data storage method and device are provided, the geometric model file of the industrial robot, the process parameter document and the industrial sensor real-time data stream are collected through the standardized interface, the feature extraction is performed on the geometric model file by using the parser based on the STEP standard to obtain the manufacturing feature entity, the entity extraction is performed on the process parameter document according to the set BiLSTM-CRF model, the process parameter entity after the entity extraction is converted into the tree structure according to the XML Schema template, the process parameter entity is obtained by performing the logical constraint on the tree process parameter structure, the knowledge graph is constructed according to the manufacturing feature entity, the process parameter entity and the industrial robot equipment, the industrial sensor real-time data stream is divided into time windows, the data statistical features in each time window are calculated and mapped with the parameter nodes in the knowledge graph to obtain the graph database, and therefore the process data use efficiency of the industrial robot can be improved. BRIEF DESCRIPTION OF DRAWINGS

[0019] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, a brief introduction will be given below to the drawings needed to be used in the embodiments or prior art description. Obviously, the drawings in the following description are some embodiments of the present application, and other drawings can also be obtained by those skilled in the art without creative effort on the basis of these drawings.

[0020] Figure 1 One of the flowcharts of the robot process data storage method in the embodiments of the present application; Figure 2 The structural diagram of the robot process data storage device in the embodiments of the present application; Figure 3 The structural diagram of the electronic device in the embodiments of the present application.

[0021] Reference signs: Electronic device 9600, central processor 9100, memory 9140, communication module 9110, input unit 9120, audio processor 9130, display 9160, power supply 9170, buffer memory 9141, application / function storage unit 9142, data storage unit 9143, driver program storage unit 9144, antenna 9111, speaker 9131, microphone 9132. DETAILED DESCRIPTION

[0022] To make the purposes, technical solutions, and advantages of the embodiments of the present application clearer, the technical solutions in the embodiments of the present application will be described clearly and completely below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are some but not all of the embodiments of the present application. Based on the embodiments in the present application, all other embodiments obtained by a person of ordinary skill in the art without creative work fall within the protection scope of the present application.

[0023] The acquisition, storage, use, processing, and the like of data in the technical solutions of the present application comply with relevant provisions of national laws and regulations.

[0024] Considering that with the improvement of the automation degree of a production line, robot process data presents high heterogeneity and complexity, the present application provides a robot process data storage method and device, which collects geometric model files, process parameter documents, and industrial sensor real-time data streams of an industrial robot through a standardized interface; adopts a parser based on a STEP standard to perform feature extraction on the geometric model files to obtain manufacturing feature entities, performs entity extraction on the process parameter documents according to a set BiLSTM-CRF model, converts the process parameter entities after entity extraction into a tree structure according to an XML Schema template, and performs logical constraint on the tree-shaped process parameter structure to obtain process parameter entities; constructs a knowledge graph according to the manufacturing feature entities, the process parameter entities, and an industrial robot device, divides industrial sensor real-time data streams into time windows, calculates data statistical features in each time window, and maps the data statistical features with parameter nodes in the knowledge graph to obtain a graph database, thereby improving the use efficiency of process data of the industrial robot.

[0025] To improve the use efficiency of process data of an industrial robot, the present application provides an embodiment of a robot process data storage method, referring to Figure 1 , which specifically includes the following contents: Step S101: collecting multi-source heterogeneous original data of an industrial robot through a standardized interface, wherein the multi-source heterogeneous original data includes geometric model files, process parameter documents, and industrial sensor real-time data streams. Optionally, in the present embodiment, a unified interface specification is designed to realize the collection of data from different sources. These interfaces can be compatible with multiple data formats and transmission protocols to ensure the integrity and consistency of the data.

[0026] Specifically, the multi-source heterogeneous original data includes; Geometric model files: geometric model files exported from CAD software (such as SolidWorks, CATIA, etc.), containing three-dimensional geometric information of a part, such as point cloud data, triangular mesh, and the like.

[0027] Process Parameter Documents: These contain the process parameters required during robotic machining, such as welding current, speed, grinding force, etc. These documents exist in the form of text, tables, etc.

[0028] Industrial Sensor Real-Time Data Streams: Real-time data streams obtained from industrial sensors such as force sensors, displacement sensors, temperature sensors, etc. These data reflect dynamic information during the machining process.

[0029] Through standardized interfaces, the unified collection of data from different sources is achieved, providing a foundation for subsequent data processing and analysis.

[0030] Step S102: According to the preset Poisson reconstruction algorithm, the boundary reconstruction of the point cloud data in the geometric model file is performed, and the corresponding standard geometric model file is determined. A parser based on the STEP standard is used to extract features from the standard geometric model file, and the corresponding manufacturing feature entity is determined. According to the set BiLSTM-CRF model, the entity extraction of the process parameter document is performed, and the corresponding initial process parameter entity is determined. Define an XML Schema template, convert the initial process parameter entity into a tree structure according to the XML Schema template, determine the corresponding tree-shaped process parameter structure, and perform logical constraints on the tree-shaped process parameter structure according to the preset process condition constraint rule and the preset quality evaluation index, and determine the corresponding process parameter entity. Optionally, in the present embodiment, the above data is subjected to feature extraction and entity extraction, laying a foundation for efficient heterogeneous data query in subsequent graph database construction.

[0031] Optionally, in the present embodiment, the geometric model file is subjected to entity extraction to obtain the manufacturing feature entity.

[0032] Specifically, for non-STEP format CAD models (such as triangular meshes) in the geometric model file, first generate B-rep expression through Poisson reconstruction algorithm, and then perform feature extraction.

[0033] Poisson reconstruction algorithm converts discrete point cloud data into continuous geometric surface through mathematical modeling and optimization technology, thereby generating standard geometric model file. Poisson reconstruction algorithm can effectively handle the incompleteness and noise of point cloud data, and generate high-quality geometric model.

[0034] For the geometric model file preprocessed by the Poisson reconstruction algorithm, a parser based on the STEP AP242 standard is used to parse the standard geometric model file and extract the geometric feature information, including feature position, size parameter. (Such as boundary representation (B-rep) topological surface, edge, vertex data, etc.). Next, manufacturing feature objects are generated for each machining feature (such as hole, groove, chamfer). The parser based on the STEP standard ensures the standardization and consistency of the geometric model, providing accurate geometric information for subsequent process planning.

[0035] Optionally, in this embodiment, entity extraction is performed on the process parameter document to obtain process parameter entities.

[0036] Specifically, BiLSTM (Bidirectional Long Short-Term Memory Network) and CRF (Conditional Random Field) models are used for entity extraction of process parameter documents. BiLSTM can capture bidirectional dependencies in text, while CRF can label and classify extracted entities to determine initial process parameter entities.

[0037] Specifically, an XML Schema template is defined to convert the initial process parameter entities into a tree structure. XML Schema is a specification for defining the structure and content of XML documents, and through the template, complex process parameter documents can be converted into structured tree data. The tree structure converted by the XML Schema template clearly represents the hierarchical relationship and dependency relationship between process parameters.

[0038] Specifically, according to the preset process condition constraint rules and quality evaluation indexes, logical constraints are performed on the tree-shaped process parameter structure to ensure the rationality and effectiveness of the process parameters. For example, the welding current cannot exceed a certain threshold, and the grinding speed must be within a certain range, etc.

[0039] After logical constraints, the tree-shaped process parameter structure is finally determined as process parameter entities, which will be used for subsequent process planning and optimization.

[0040] Step S103: Construct a knowledge graph according to the manufacturing feature entities, the process parameter entities, and the industrial robot device, divide the time window of the industrial sensor real-time data stream according to the process stage, determine a plurality of corresponding time windows, map the data statistical characteristics in each time window to the parameter nodes of the corresponding stage in the knowledge graph, and determine a corresponding graph database, wherein the graph database is used to store robot process data.

[0041] Optionally, in this embodiment, this step is the construction process of the graph database.

[0042] Specifically, a knowledge graph is constructed based on manufacturing feature entities, process parameter entities, and industrial robot devices. The knowledge graph is a structured semantic knowledge base that integrates and correlates data from different sources through the representation of entities, relationships, and attributes. In the knowledge graph, manufacturing feature entities, process parameter entities, and industrial robot devices are nodes, and the relationships between them are edges, representing the complex relationships between them through a graph structure.

[0043] Specifically, based on the above knowledge graph, the industrial sensor real-time data stream is divided into time windows according to the process stage. The division of time windows can divide continuous data streams into multiple discrete time windows according to the characteristics of the process stage, such as the welding stage, grinding stage, etc. In each time window, the statistical characteristics of the data are calculated, such as mean, variance, maximum, minimum, etc. These statistical characteristics can reflect the overall characteristics of the data in the time window.

[0044] The data statistical characteristics in each time window are mapped to the corresponding parameter nodes in the knowledge graph. Through mapping, real-time data can be associated with knowledge in the knowledge graph, thereby realizing semantic understanding and analysis of real-time data. The mapped data is stored in a graph database, forming a complete robot process data storage system. The data in the graph database not only includes static process parameters and geometric models, but also includes dynamic real-time data, providing comprehensive data support for subsequent process optimization and quality control.

[0045] Through the above steps, the collection, processing, and integration of industrial robot multi-source heterogeneous raw data are realized. Geometric model files are processed through the Poisson reconstruction algorithm and STEP standard parser to generate standard geometric models and manufacturing feature entities; process parameter documents are processed through the BiLSTM-CRF model and XML Schema template to generate structured tree-shaped process parameter structures, and process parameter entities are determined through logical constraints; industrial sensor real-time data streams are associated with knowledge in the knowledge graph through time window division and relationship mapping, and finally stored in the graph database. This process not only improves the automation and efficiency of data processing, but also provides a rich knowledge base and comprehensive data support for subsequent process planning, optimization, and quality control.

[0046] This example demonstrates how the present embodiment processes and integrates multi-source heterogeneous data to construct a graph database to improve the efficiency of using process data of industrial robots.

[0047] As can be known from the above description, the robot process data storage method provided by the embodiments of the present application can collect the geometric model file of the industrial robot, the process parameter document and the industrial sensor real-time data stream through the standardized interface; the features of the geometric model file are extracted by using the parser based on the STEP standard to obtain the manufacturing feature entity, the entity extraction is performed on the process parameter document according to the set BiLSTM-CRF model, the process parameter entity after the entity extraction is converted into the tree structure according to the XML Schema template, and the process parameter entity is obtained by performing the logical constraint on the tree-shaped process parameter structure; the knowledge graph is constructed according to the manufacturing feature entity, the process parameter entity and the industrial robot equipment, the industrial sensor real-time data stream is divided into time windows, the data statistical features in each time window are calculated, and the parameter nodes in the knowledge graph are mapped to obtain the graph database, so that the use efficiency of the process data of the industrial robot can be improved.

[0048] In an embodiment of the robot process data storage method of the present application, the following contents can also be specifically included: Step S201: estimating the normal vector of the point cloud data according to a preset Poisson reconstruction algorithm, and performing voxelization processing on the point cloud data, fitting the data of the point cloud data after the voxelization processing according to a hidden function fitting algorithm, and determining the corresponding isosurface; Step S202: extracting the network model of the isosurface according to a moving cube algorithm, and determining the corresponding standard geometric model file.

[0049] Optionally, in the embodiment, the Poisson reconstruction algorithm is an algorithm for reconstructing a three-dimensional geometric model from point cloud data. It converts discrete point cloud data into continuous geometric surfaces through mathematical modeling and optimization techniques. In the point cloud data, the normal vector of each point is very important information in the reconstruction process. The normal vector estimation calculates the normal vector of each point by locally fitting a plane or using the geometric information of the neighborhood points. The Poisson reconstruction algorithm uses these normal vectors to better understand the geometric structure of the point cloud, thereby generating more accurate geometric surfaces.

[0050] Specifically, the normal vector estimation is to calculate the neighborhood of each point (such as the set of points within the K-nearest neighbor or radius search range), fit a local plane using principal component analysis (PCA), and determine the direction of the normal vector of the point. The consistency of the direction of the normal vector is corrected by a minimum spanning tree (MST) or a global optimization method (such as Gaussian cuts), to ensure that all normal vectors are oriented outward from the surface. Accurate normal vectors can represent the local geometric features (such as curvature, orientation) of the point cloud surface, provide correct gradient information for the hidden function, and avoid holes or distortions in the reconstructed surface.

[0051] Voxelization is the process of converting point cloud data into a voxel grid, which divides the point cloud space into a uniform 3D voxel grid. Each voxel contains spatial position and attributes (such as density, normal vector). By counting the distribution of points in each voxel (such as mean or weighted average), the discrete point cloud is converted into a continuous voxel representation. The voxel resolution needs to be set according to the point cloud density and the reconstruction accuracy requirements. Preferably, an octree structure is used to optimize storage and computing efficiency. Voxelization converts unordered point cloud into structured data, reduces noise influence, and provides uniform sampling space for implicit function fitting.

[0052] Specifically, the implicit function fitting algorithm is applied to the point cloud data after voxelization processing. The algorithm finds a best implicit function to fit the point cloud data by minimizing the error between the point cloud data and the implicit function. This implicit function can describe the geometric shape of the point cloud data and represent the reconstructed geometric model through the isosurface.

[0053] The isosurface is the set of points in the implicit function where all scalar values are equal. Through the implicit function fitting algorithm, an isosurface can be determined, which represents the surface of the reconstructed geometric model. The selection of the isosurface is based on the zero isosurface of the implicit function, because the zero isosurface can well represent the boundary of the geometric model.

[0054] Specifically, the Marching Cubes Algorithm searches for the intersection points of the isosurface in the voxel grid and generates triangular meshes according to these intersection points, thereby extracting the surface of the geometric model.

[0055] The Marching Cubes Algorithm is applied with the isosurface determined by the implicit function fitting algorithm as input. The algorithm searches for the intersection points of the isosurface in the voxel grid and generates triangular meshes according to these intersection points. These triangular meshes can accurately represent the shape of the isosurface, thereby extracting the standard geometric model file. The extracted standard geometric model file (such as STL file) can be widely used in subsequent process planning, simulation and machining, etc., providing an accurate geometric model basis for the application of industrial robots.

[0056] Through step S202, the embodiment realizes the conversion from point cloud data to standard geometric model file, laying the foundation for subsequent extraction of manufacturing feature entities.

[0057] In an embodiment of the robot process data storage method of the present application, the following contents can also be specifically included: Step S301: Construct a field-adapted process text corpus, and perform process parameter entity annotation on the process text corpus using a hierarchical annotation strategy to determine a corresponding process text training set; Step S302: model training is performed on the initial BiLSTM-CRF model according to the process text training set to determine the corresponding BiLSTM-CRF model, wherein the BiLSTM layer of the initial BiLSTM-CRF model is provided with a bidirectional gating mechanism, and the hidden states of the forward layer and the backward layer are fused through learnable weights, and the CRF layer of the initial BiLSTM-CRF model restricts the label transition rule.

[0058] Optionally, in the embodiment, a field-adapted process text corpus is constructed.

[0059] A large amount of process text data is collected from the field of industrial robot manufacturing. The text data includes process parameter documents, operation manuals, technical reports, and process flow descriptions. These documents contain rich process parameter information, such as welding current, welding speed, grinding force, and feed speed. The cleaned text data is organized into a structured corpus for subsequent processing and analysis.

[0060] Optionally, in the embodiment, the hierarchical annotation strategy refers to annotating process parameter entities in the text at different levels. The strategy adopted in this scheme is a three-level annotation strategy, specifically: Lexical layer annotation: annotating basic language units such as words and phrases in the text, and identifying process parameter-related words such as "current" and "speed".

[0061] Syntax layer annotation: analyze the syntax structure of the text, and annotate the syntax components related to the process parameters, such as subject, predicate, and object, to determine the context relationship of the parameters.

[0062] Semantic layer annotation: annotate the semantic information in the text, and identify specific process parameter entities and their attributes, such as "welding current" and "100A" in "welding current is 100A".

[0063] After annotation, a labeled process text training set is formed, and each annotated entity contains its type, position, and attribute information. The construction of a field-adapted process text corpus provides a rich data foundation for subsequent model training, ensuring that the model can learn the specific language patterns and process parameter characteristics in the field of industrial robot manufacturing.

[0064] Optionally, in the embodiment, the initial BiLSTM-CRF model is set as follows: BiLSTM layer: BiLSTM (Bidirectional Long Short-Term Memory Network) can capture bidirectional dependencies in text. In the initial BiLSTM model, a bidirectional gating mechanism is set, i.e. the forward layer and the backward layer capture the forward and reverse information of the text respectively. The forward layer processes information from the beginning to the end of the text, and the backward layer processes information from the end to the beginning of the text. The hidden states of the two layers are fused through learnable weights to fully utilize the bidirectional information.

[0065] Specifically, the forward layer learns historical process parameter dependencies (such as "current value is affected by previous process speed"), and the backward layer captures future process constraints (such as "the subsequent polishing process requires the current grinding force to be ≤5N"). The contribution of forward and backward hidden states is dynamically adjusted through learnable weights (such as "higher backward layer weight" to emphasize key parameters at the end of the process).

[0066] CRF layer: CRF (Conditional Random Field) layer is used to constrain the label transition rules in the labeling task. CRF layer can consider the transition probability between labels to ensure the coherence and reasonableness of the labeling results.

[0067] For example, when labeling process parameter entities, the CRF layer can constrain that "welding current" must be followed by a numerical label such as "100A".

[0068] For another example, the "welding current" label cannot be directly followed by the "polishing process" label.

[0069] In an embodiment, the rule constraints can be generated by process association rule mining and applied to the CRF layer to constrain the label transition rules and avoid illegal sequences.

[0070] Optionally, in the present embodiment, the training process of the BiLSTM-CRF model is as follows: First, convert the process text training set into word vectors (such as Word2Vec or BERT embedding) and integrate domain-specific features (such as "grinding and polishing" and "spraying" vectors).

[0071] Then, jointly optimize the cross-entropy loss of BiLSTM and the sequence transition loss of CRF to ensure the consistency of parameter entity recognition and process sequence labeling. Adjust the weights through backpropagation to focus on optimizing the recognition ability of long-tail data (such as rare process parameters).

[0072] After the above training, the process entities extracted by the BiLSTM-CRF model can be directly used to construct a process knowledge graph, such as storing the "welding current-weld quality" relationship in the graph to support template recommendation based on subgraph retrieval.

[0073] It can be understood that the bidirectional gating mechanism of the BiLSTM layer fully utilizes the bidirectional information of the text, so that the model can better capture the context features of the process parameter entity. The label transition rule constraint of the CRF layer ensures the coherence and reasonableness of the labeling result, avoiding unreasonable labeling results. Ensure the professionalism and application value of the model in the field of industrial robot manufacturing.

[0074] Through step S302, the training of the BiLSTM-CRF model is realized, and a high-quality data basis is provided based on the bidirectional gating mechanism and the label transition rule constraint of the CRF layer.

[0075] In an embodiment of the robot process data storage method of the present application, the following contents can also be specifically included: Step S401: defining the hierarchical relationship of process parameters, wherein the hierarchical relationship of process parameters includes process type, parameter group, and parameter value; Step S402: defining an XML Schema template according to the hierarchical relationship of process parameters, wherein the XML Schema template includes a three-layer structure of process type classification nodes, parameter group nodes, and parameter value nodes, wherein the parameter group nodes are logically grouped according to equipment attributes, material attributes, and environmental attributes, and the parameter value nodes are bound to measurement units and valid value range attributes.

[0076] Optionally, in the present embodiment, the core of this step is to structure and layer the robot process parameters, and to establish a three-level logical system from macro to micro.

[0077] Specifically, the hierarchical structure of process parameters: Process type: top-level classification is divided according to processing task type, and process parameters are classified according to different process types. Each process type corresponds to a unique parameter system, for example, welding process, grinding process, and spraying process.

[0078] Parameter group: middle layer grouping according to parameter physical meaning or scope, under each process type, further grouping parameters into different groups. These groups are logically grouped according to equipment attributes, material attributes, and environmental attributes. For example, in the welding process, equipment attribute parameters may include welding current, welding voltage, etc.; material attribute parameters may include welding material type, welding material diameter, etc.; environmental attribute parameters may include environmental temperature, humidity, etc.

[0079] Parameter value: bottom layer defines specific values and constraint conditions, under each parameter group, defines the value of each parameter. These values can be specific numerical values or range values. For example, the value of welding current can be 100A to 200A, the value of welding voltage can be 20V to 30V, etc.

[0080] By defining the hierarchical relationship of process type, parameter group and parameter value, a clear hierarchy is established. This structure can help users better understand and manage process parameters, while also providing a basis for subsequent data processing and analysis.

[0081] It can be understood that the logical grouping of parameter groups reveals the coupling relationship between parameters. For example, material hardness (material attribute) and abrasive tool linear speed (device attribute) jointly affect the polishing quality, and the hierarchical relationship provides a framework for subsequent parameter collaborative optimization.

[0082] Optionally, in this embodiment, the hierarchical relationship is converted into a machine-readable XML Schema specification, and the parameter template is standardized by three-layer nodes.

[0083] Specifically, in the XML Schema template, a top-level node is defined to classify different process types. Under each process type node, parameter group nodes are defined. These nodes are logically grouped according to device attributes, material attributes, and environmental attributes. Under each parameter group node, specific parameter value nodes are defined. These nodes contain specific values or ranges of parameters, and bind measurement units and valid value range attributes.

[0084] It can be understood that the measurement unit attribute is bound to each parameter value node to ensure the consistency of parameter values. At the same time, the valid value range attribute is defined for each parameter value node to ensure that the parameter value is within a reasonable range.

[0085] By defining the XML Schema template, the process parameters are represented in a structured manner. This structured representation helps data storage, transmission and processing, improving data usability and maintainability.

[0086] Through step S402, this embodiment successfully realizes the structured management and representation of process parameters by defining the hierarchical relationship of process parameters and the XML Schema template.

[0087] In an embodiment of the robot process data storage method of the present application, the following contents can also be specifically included: Step S501: According to the preset rule engine, the process parameter document is parsed with conditional statements, and the parsed conditional statements are converted into XPath process condition constraint rules; Step S502: According to the preset quality evaluation index, a parameter-quality association matrix is constructed, and the tree-shaped process parameter structure is logically constrained according to the XPath process condition constraint rules and the parameter-quality association matrix to determine the corresponding process parameter entity.

[0088] Optionally, in this embodiment, a rule engine (such as Drools, Jess) is used to perform lexical analysis and syntax parsing on the structured conditional statements in the process parameter document, identifying the conditional statements therein. These conditional statements may exist in natural language form, such as "when the material thickness exceeds 10 mm, the welding current should be greater than 150 A". The parsed conditional statements are converted into XPath expressions, and the XPath rules support dynamic retrieval of the process parameter document in XML or JSON format, so as to be able to match and verify with the tree-shaped process parameter structure.

[0089] Optionally, in this embodiment, a parameter-quality association matrix is constructed.

[0090] According to the preset quality evaluation indicators, a parameter-quality association matrix is constructed. This matrix describes the relationship strength between each process parameter and the quality evaluation indicators. For example, a row in the matrix may represent the relationship between the welding current and the welding quality, and another row may represent the relationship between the welding speed and the welding quality.

[0091] First, according to the process requirements (such as welding strength, grinding surface roughness), the quality targets are quantified to form an indicator set Q = {q1, q2, …}.

[0092] Then, a statistical method (such as Pearson correlation coefficient) is used to calculate the influence weight of each process parameter on the quality indicators. For example, the weight of the welding current on the weld strength is 0.7, and the weight of the feed speed on the roughness is 0.5. A matrix M[p_i][q_j] is constructed, where the element value represents the influence degree of the parameter p_i on the indicator q_j, and the numerical value is usually normalized to [0, 1].

[0093] Optionally, in this embodiment, the parameter nodes filtered out by the XPath rules are associated with the corresponding parameter rows in the matrix. For example, if the XPath locates to "current > 200A", the weight of the current on each quality indicator is extracted from the matrix to generate the constraint condition: "the increase of the current needs to be synchronized with the increase of the preheating temperature (weight 0.6) to ensure the strength".

[0094] Combined with algorithms such as NSGA-II, the quality indicators are used as optimization objectives to search for Pareto optimal solutions in the feasible region constrained by XPath. For example, under the premise of "current > 200A", the "maximum strength" and "minimum energy consumption" are balanced.

[0095] It can be understood that the matrix explicitly associates process parameters with quality targets, avoiding the blindness of empirical parameter adjustment. For example, in the spraying process, it is found through the matrix that the "atomization pressure" has a significant impact on the "coating uniformity" (weight 0.8), so this parameter is prioritized for optimization. When the XPath rule conflicts with the quality target (such as high current to improve strength but increase energy consumption), the system automatically recommends a compromise solution (such as selecting a combination of current 220A + preheating 100℃).

[0096] The final output parameter entity includes: Base value: parameter range defined by XPath rule (such as current ∈ [200, 250] A).

[0097] Optimization suggestion: associated parameter adjustment recommended by the matrix (such as "suggested preheating 80℃ when current is 200A").

[0098] Quality prediction: based on the weighted calculation of the matrix, the quality score of the current parameter combination is estimated (such as strength expected 90 points).

[0099] Through step S502, the embodiment successfully realizes the conversion of process parameters from static documents to dynamic optimization entities through rule analysis and matrix modeling.

[0100] In an embodiment of the robot process data storage method of the present application, the following content can also be included: Step S601: create feature-parameter relationship edges according to necessity level and parameter combination constraints, link the manufacturing feature entity to the process parameter entity according to the feature-parameter relationship edges; Step S602: create parameter-equipment relationship edges according to the adaptation score, link the process parameter entity to the industrial robot equipment node according to the parameter-equipment relationship edges, and determine the corresponding knowledge graph.

[0101] Optionally, in the embodiment, first, the necessity level of the association between the manufacturing feature entity (such as the geometric features of the workpiece, material properties, etc.) and the process parameter entity (such as cutting speed, feed speed, grinding force, etc.) is evaluated. The necessity level is divided into three levels: high, medium and low, which is used to quantify the degree of dependence between features and parameters. For example, in the grinding and polishing process, the correlation between surface curvature and grinding force may be marked as "high necessity", while the correlation between ambient temperature and feed speed may be "low necessity".

[0102] Then, the combination constraints between process parameters are modeled to ensure the rationality and feasibility of parameter combination. For example, in the welding process, the current and welding speed need to meet a certain proportion range to avoid weld defects.

[0103] Based on the necessity level and combination constraints, "feature-parameter relationship edges" are created in the knowledge graph, representing the association between manufacturing feature entities and process parameter entities.

[0104] For example, for a manufacturing feature entity "weld width", which has a high necessity level, there is a combination constraint between the associated process parameter entities "welding current" and "welding speed". Therefore, a relationship edge is created to represent the dependency between "weld width" and "welding current" and "welding speed".

[0105] By creating feature-parameter relationship edges, the dependency between manufacturing features and process parameters is clearly defined, making it easier to understand which parameters have a direct impact on a specific feature during process planning and optimization. In the knowledge graph, these relationship edges provide a foundation for intelligent decision-making.

[0106] Optionally, in this embodiment, the matching degree of process parameters with industrial robot equipment (such as brand and model, end effector, number of axes, etc.) is evaluated. The adaptation score is based on the following dimensions: Device capability: whether the robot load, precision, and working range support the parameter requirements (such as "grinding force ≤ robot maximum load").

[0107] Historical performance: statistics of the device's performance in executing similar parameters, such as yield rate and efficiency.

[0108] Expert experience: manually annotated device-parameter adaptation rules (such as "ABB robot is suitable for high-precision welding").

[0109] According to the adaptation score, relationship edges are created between process parameter entities and industrial robot equipment nodes, representing the adaptation relationship between process parameters and equipment.

[0110] For example, for a process parameter entity "welding current", which has a high adaptation score, a relationship edge is created between it and the industrial robot equipment node "welding robot A", indicating that the "welding current" parameter is suitable for use on "welding robot A".

[0111] Through the created relationship edges, process parameter entities are linked to industrial robot equipment nodes. This step ensures that in the knowledge graph, each process parameter entity has a clear association with its adapted industrial robot equipment.

[0112] Through the construction of the knowledge graph by these relationship edges, in the process execution stage, it can be quickly determined which equipment is most suitable for the current process parameters, thereby reducing the failure of process or quality problems caused by equipment mismatch.

[0113] Subsequently, real-time processing data collected through the process cloud service platform can dynamically adjust the properties of the relationship edges.

[0114] Through the step S602, the embodiment successfully realizes the association among the manufacturing feature entity, the process parameter entity and the industrial robot device, and improves the accuracy and efficiency of process planning and execution.

[0115] In an embodiment of the robot process data storage method of the present application, the following contents can also be specifically included: Step S701: Adopt Apache Flink real-time processing of sensor data stream, divide window data block according to process stage through window function, extract statistical features of sensor data in each window, and determine corresponding process state vector according to the statistical features; Step S702: Associate the process state vector with the corresponding process node in the knowledge graph, and determine the corresponding graph database.

[0116] Optionally, in the embodiment, the stream processing engine of Apache Flink is adopted to receive the sensor data stream (such as force / position signal, temperature, vibration, etc.) from the robot process site in real time. By defining a rolling window (Tumbling Window) or a sliding window (Sliding Window) based on time or the number of events, the continuous data stream is divided into discrete window data blocks. For example, the welding process can be divided into windows according to the stages of "arc striking-welding-arc recovery", and each window covers the data of a complete process stage.

[0117] Multi-dimensional feature calculation is performed on the original sensor data in each window, including: Time domain features, mean (reflecting parameter stability), variance (characterizing fluctuation degree), peak value (detecting abnormal impact).

[0118] Frequency domain features, dominant frequency components (such as vibration frequency in grinding) are extracted through FFT transformation.

[0119] High-order features, time series correlation based on sliding window (such as autoregressive coefficient) or process-specific composite index (such as welding heat input = current x voltage / speed).

[0120] The extracted statistical features are arranged in a fixed order to form a structured vector, called a process state vector.

[0121] For example, the state vector of a welding window can be expressed as: [Mean current = 150A, variance voltage = 2.1V 2 , temperature slope = 0.5°C / s,...] Optionally, in this embodiment, the process state vector generated in step S701 is associated with the corresponding process node in the knowledge graph. For example, the process state vector of a welding window can be associated with the "welding process" node in the knowledge graph.

[0122] The association process can be achieved by matching the features in the process state vector with the attributes of the nodes in the knowledge graph. For example, if the average current in the process state vector matches the current range of a certain welding process node in the knowledge graph, the vector is associated with the node.

[0123] It can be understood that when constructing the knowledge graph based on manufacturing feature entities, process parameter entities, and industrial robot device information, we have structured the process parameter entities and obtained a tree-like process parameter structure. Therefore, the root node of the process parameter entity in the knowledge graph is the process node, and the secondary nodes and leaf nodes are also associated with the process node.

[0124] Finally, the association results are written into a graph database such as Neo4j. The topology structure of the graph database supports full-link tracing, for example, quickly locating problems through the "current process → abnormal event → root cause" path.

[0125] By associating the process state vector with the process node in the knowledge graph, semantic understanding of real-time data is achieved, improving the interpretability and value of the data. The use of graph databases integrates real-time data with static knowledge, forming a complete knowledge system for subsequent queries and analysis. At the same time, the efficient storage and query capabilities of the graph database can quickly respond to complex query requests, improving the efficiency of data processing.

[0126] Through step S702, this embodiment successfully realizes real-time processing, feature extraction, semantic association, and efficient storage of industrial robot sensor data streams, and improves the efficiency of industrial robot process data through the construction of a graph database.

[0127] To improve the efficiency of industrial robot process data, the present application provides an embodiment of a robot process data storage device for implementing all or part of the contents of the robot process data storage method, as shown in Figure 2 , which specifically includes the following contents: A data acquisition module 10 is used to acquire multi-source heterogeneous raw data of an industrial robot through a standardized interface, wherein the multi-source heterogeneous raw data includes geometric model files, process parameter documents, and industrial sensor real-time data streams. The manufacturing and process entity determination module 20 is configured to perform boundary reconstruction on the point cloud data in the geometric model file according to a preset Poisson reconstruction algorithm, to determine a corresponding standard geometric model file, to perform feature extraction on the standard geometric model file by using a parser based on the STEP standard, to determine a corresponding manufacturing feature entity, to perform entity extraction on the process parameter document according to a set BiLSTM-CRF model, to determine a corresponding initial process parameter entity, to define an XML Schema template, to convert the initial process parameter entity into a tree structure according to the XML Schema template, to determine a corresponding tree-shaped process parameter structure, and to perform logical constraint on the tree-shaped process parameter structure according to a preset process condition constraint rule and a preset quality evaluation index, to determine a corresponding process parameter entity. The graph database determination module 30 is configured to construct a knowledge graph according to the manufacturing feature entity, the process parameter entity, and the industrial robot device, to divide the industrial sensor real-time data stream into a plurality of time windows according to a process stage, to determine a corresponding plurality of time windows, to perform relationship mapping between a data statistical feature in each time window and a parameter node at a corresponding stage in the knowledge graph, and to determine a corresponding graph database, which is configured to store robot process data.

[0128] As can be seen from the above description, the robot process data storage device provided by the embodiments of the present application can collect the geometric model file, the process parameter document, and the industrial sensor real-time data stream of the industrial robot through a standardized interface. The manufacturing feature entity is obtained by performing feature extraction on the geometric model file by using a parser based on the STEP standard. The process parameter entity is obtained by performing entity extraction on the process parameter document according to a set BiLSTM-CRF model, converting the process parameter entity after entity extraction into a tree structure according to an XML Schema template, and performing logical constraint on the tree-shaped process parameter structure. The graph database is obtained by constructing a knowledge graph according to the manufacturing feature entity, the process parameter entity, and the industrial robot device, dividing the industrial sensor real-time data stream into a plurality of time windows, performing relationship mapping between a data statistical feature in each time window and a parameter node at a corresponding stage in the knowledge graph, and determining a corresponding graph database. Therefore, the use efficiency of the process data of the industrial robot can be improved.

[0129] From the hardware level, in order to improve the use efficiency of the process data of the industrial robot, the present application provides an embodiment of an electronic device for implementing all or part of the robot process data storage method, which specifically includes the following contents: The system comprises a processor, memory, a communications interface, and a bus; wherein the processor, memory, and communications interface communicate with each other via the bus; the communications interface is used to realize information transmission between the robot process data storage method and core business systems, user terminals, and related databases and other related devices; the logic controller can be a desktop computer, tablet computer, or mobile terminal, etc., and this embodiment is not limited to these. In this embodiment, the logic controller can be implemented with reference to the embodiments of the robot process data storage method in the present embodiment, and the contents of the embodiments of the robot process data storage method are incorporated herein, and repeated details will not be described again.

[0130] It is understood that the user terminal may include smartphones, tablet computers, network set-top boxes, portable computers, desktop computers, personal digital assistants (PDAs), in-vehicle devices, smart wearable devices, etc. Among these, the smart wearable devices may include smart glasses, smartwatches, smart bracelets, etc.

[0131] In practical applications, some aspects of the robot process data storage method can be executed on the electronic device side as described above, or all operations can be completed in the client device. The choice can be made based on the processing power of the client device and the limitations of the user's usage scenario. This application does not impose any limitations on this. If all operations are completed in the client device, the client device may further include a processor.

[0132] The aforementioned client device may have a communication module (i.e., a communication unit) that can communicate with a remote server to achieve data transmission with the server. The server may include a server on the task scheduling center side; in other implementation scenarios, it may also include a server on an intermediate platform, such as a server on a third-party server platform that has a communication link with the task scheduling center server. The server may include a single computer device, a server cluster consisting of multiple servers, or a distributed server structure.

[0133] Figure 3 This is a schematic block diagram illustrating the system configuration of the electronic device 9600 according to an embodiment of this application. Figure 3 As shown, the electronic device 9600 may include a central processing unit 9100 and a memory 9140; the memory 9140 is coupled to the central processing unit 9100. It is worth noting that... Figure 3 This is an example; other types of structures can also be used to supplement or replace this structure to achieve telecommunications functions or other functions.

[0134] In an embodiment, the robot process data storage method function can be integrated into the central processor 9100. The central processor 9100 can be configured to control as follows: Step S101: Collecting multi-source heterogeneous original data of an industrial robot through a standardized interface, wherein the multi-source heterogeneous original data includes a geometric model file, a process parameter document, and an industrial sensor real-time data stream; Step S102: Reconstructing the boundary of the point cloud data in the geometric model file according to a preset Poisson reconstruction algorithm, determining a corresponding standard geometric model file, extracting features of the standard geometric model file using a parser based on the STEP standard, determining a corresponding manufacturing feature entity, extracting entities from the process parameter document according to a set BiLSTM-CRF model, determining a corresponding initial process parameter entity, defining an XML Schema template, converting the initial process parameter entity into a tree structure according to the XML Schema template, determining a corresponding tree-shaped process parameter structure, logically constraining the tree-shaped process parameter structure according to a preset process condition constraint rule and a preset quality evaluation index, and determining a corresponding process parameter entity; Step S103: Constructing a knowledge graph according to the manufacturing feature entity, the process parameter entity, and the industrial robot equipment, dividing the industrial sensor real-time data stream into a plurality of time windows according to the process stages, determining a corresponding plurality of time windows, mapping the data statistical features in each time window to the parameter nodes of the corresponding stages in the knowledge graph, and determining a corresponding graph database, wherein the graph database is used to store robot process data.

[0135] As can be seen from the above description, the electronic device provided by the embodiments of the present application collects the geometric model file, the process parameter document, and the industrial sensor real-time data stream of the industrial robot through a standardized interface; extracts features of the geometric model file using a parser based on the STEP standard to obtain a manufacturing feature entity, extracts entities from the process parameter document according to a set BiLSTM-CRF model, converts the process parameter entity after entity extraction into a tree structure according to an XML Schema template, and logically constrains the tree-shaped process parameter structure to obtain a process parameter entity; constructs a knowledge graph according to the manufacturing feature entity, the process parameter entity, and the industrial robot equipment, divides the industrial sensor real-time data stream into a plurality of time windows, maps the data statistical features in each time window to the parameter nodes in the knowledge graph, and obtains a graph database, thereby improving the use efficiency of the process data of the industrial robot.

[0136] In another embodiment, the robot process data storage method can be configured separately from the central processor 9100, for example, the robot process data storage method can be configured as a chip connected with the central processor 9100, and the functions of the robot process data storage method are realized through the control of the central processor.

[0137] As shown in FIG. 9, the electronic device 9600 can further include a communication module 9110, an input unit 9120, an audio processor 9130, a display 9160, and a power supply 9170. It is to be noted that the electronic device 9600 does not necessarily include all the components shown in FIG. 9; in addition, the electronic device 9600 can include components not shown in FIG. 9, which can be referred to the prior art. Figure 3 Figure 3 As shown in FIG. 9, the central processor 9100, which is sometimes referred to as a controller or operating control, can include a microprocessor or other processor device and / or logic device, which receives input and controls the operation of the various components of the electronic device 9600. Figure 3 As shown in FIG. 9, the central processor 9100, which is sometimes referred to as a controller or operating control, can include a microprocessor or other processor device and / or logic device, which receives input and controls the operation of the various components of the electronic device 9600.

[0138] Figure 3 The memory 9140, for example, can be one or more of a buffer, a flash memory, a hard drive, a removable media, a volatile memory, a non-volatile memory, or other suitable device. Information related to the failure described above can be stored, and in addition, programs for executing the related information can be stored. The central processor 9100 can execute the programs stored in the memory 9140 to achieve information storage or processing, etc.

[0139] The input unit 9120 provides input to the central processor 9100. The input unit 9120 is, for example, a key or touch input device. The power supply 9170 is used to provide power to the electronic device 9600. The display 9160 is used to display display objects such as images and text. The display can be, for example, an LCD display, but is not limited thereto.

[0140] The memory 9140 can be a solid state memory, for example, a read only memory (ROM), a random access memory (RAM), a SIM card, etc. It can also be a memory that saves information even when power is off, can be selectively erased, and is provided with more data, examples of which are sometimes referred to as EPROM, etc. The memory 9140 can also be some other type of device. The memory 9140 includes a buffer memory 9141 (sometimes referred to as a buffer). The memory 9140 can include an application / function storage section 9142 for storing application programs and function programs or for executing the flow of the operation of the electronic device 9600 by the central processor 9100.

[0141] The memory 9140 can be a solid state memory, for example, a read only memory (ROM), a random access memory (RAM), a SIM card, etc. It can also be a memory that saves information even when power is off, can be selectively erased, and is provided with more data, examples of which are sometimes referred to as EPROM, etc. The memory 9140 can also be some other type of device. The memory 9140 includes a buffer memory 9141 (sometimes referred to as a buffer). The memory 9140 can include an application / function storage section 9142 for storing application programs and function programs or for executing the flow of the operation of the electronic device 9600 by the central processor 9100. ​​

[0142] The memory 9140 can further include a data storage 9143 for storing data such as contacts, digital data, pictures, sounds, and / or any other data used by the electronic device. A driver storage 9144 of the memory 9140 can include various drivers of the electronic device for communication functions and / or for performing other functions of the electronic device (e.g., a messaging application, a phonebook application, etc.).

[0143] The communication module 9110 is a transmitter / receiver that transmits and receives signals via the antenna 9111. The communication module 9110 is coupled to the central processor 9100 to provide input signals and receive output signals, as in the case of a conventional mobile communication terminal.

[0144] Based on different communication technologies, a plurality of communication modules 9110, such as a cellular network module, a Bluetooth module, and / or a wireless local area network module, etc., can be provided in the same electronic device. The communication module 9110 is further coupled to the speaker 9131 and the microphone 9132 via the audio processor 9130 to provide audio output via the speaker 9131 and to receive audio input from the microphone 9132, thereby implementing a conventional telecommunication function. The audio processor 9130 can include any suitable buffer, decoder, amplifier, etc. In addition, the audio processor 9130 is coupled to the central processor 9100, thereby enabling recording on the local device via the microphone 9132 and enabling playing of stored sounds on the local device via the speaker 9131.

[0145] The embodiments of the present application further provide a computer readable storage medium capable of implementing all steps of the robot process data storage method in the above-mentioned embodiments, wherein the computer readable storage medium stores a computer program, and the computer program is executed by a processor to implement all steps of the robot process data storage method in the above-mentioned embodiments, for example, the processor executes the computer program to implement the following steps: Step S101: collecting industrial robot multi-source heterogeneous original data through a standardized interface, wherein the multi-source heterogeneous original data includes a geometric model file, a process parameter document, and an industrial sensor real-time data stream; Step S102: boundary reconstruction is performed on the point cloud data in the geometric model file according to a preset Poisson reconstruction algorithm, a corresponding standard geometric model file is determined, a feature extractor based on a STEP standard is used to extract features from the standard geometric model file, a corresponding manufacturing feature entity is determined, an entity extraction is performed on the process parameter document according to a set BiLSTM-CRF model, an initial process parameter entity is determined, an XML Schema template is defined, the initial process parameter entity is converted into a tree structure according to the XML Schema template, a corresponding tree-shaped process parameter structure is determined, and a process parameter entity is determined by performing logical constraints on the tree-shaped process parameter structure according to a preset process condition constraint rule and a preset quality evaluation index; Step S103: a knowledge graph is constructed according to the manufacturing feature entity, the process parameter entity and the industrial robot device, a plurality of time windows are determined by dividing a time window of the industrial sensor real-time data stream according to a process stage, a corresponding graph database is determined by performing relationship mapping between data statistical features in each time window and parameter nodes in a corresponding stage in the knowledge graph, and the graph database is used to store robot process data.

[0146] From the above description, it can be known that the computer readable storage medium provided by the embodiments of the present application collects the geometric model file, the process parameter document and the industrial sensor real-time data stream through a standardized interface; a feature extractor based on a STEP standard is used to extract features from the geometric model file to obtain a manufacturing feature entity, an entity extraction is performed on the process parameter document according to a set BiLSTM-CRF model, the process parameter entity after the entity extraction is converted into a tree structure according to an XML Schema template, and a process parameter entity is obtained by performing logical constraints on the tree-shaped process parameter structure; a knowledge graph is constructed according to the manufacturing feature entity, the process parameter entity and the industrial robot device, the industrial sensor real-time data stream is divided into time windows, data statistical features in each time window are mapped with parameter nodes in the knowledge graph to obtain a graph database, and thus the use efficiency of the process data of the industrial robot can be improved.

[0147] The embodiments of the present application also provide a computer program product capable of implementing all steps of the robot process data storage method in the above-mentioned embodiments, wherein the computer program / instruction is executed by a processor to implement the steps of the robot process data storage method, for example, the computer program / instruction implements the following steps: Step S101: collecting industrial robot multi-source heterogeneous original data through a standardized interface, wherein the multi-source heterogeneous original data includes a geometric model file, a process parameter document and an industrial sensor real-time data stream; Step S102: Reconstruct the boundary of the point cloud data in the geometric model file according to the preset Poisson reconstruction algorithm to determine the corresponding standard geometric model file. Use a parser based on the STEP standard to extract features from the standard geometric model file to determine the corresponding manufacturing feature entities. Extract entities from the process parameter document according to the set BiLSTM-CRF model to determine the corresponding initial process parameter entities. Define an XML Schema template. Convert the initial process parameter entities into a tree structure according to the XML Schema template to determine the corresponding tree process parameter structure. Logically constrain the tree process parameter structure according to preset process condition constraint rules and preset quality evaluation indicators to determine the corresponding process parameter entities. Step S103: Construct a knowledge graph based on the manufacturing feature entity, the process parameter entity, and the industrial robot equipment; divide the real-time data stream of the industrial sensor into time windows according to the process stage; determine multiple corresponding time windows; map the relationship between the statistical characteristics of the data in each time window and the parameter nodes of the corresponding stage in the knowledge graph; and determine the corresponding graph database, which is used to store robot process data.

[0148] As described above, the computer program product provided in this application collects geometric model files, process parameter documents, and real-time data streams from industrial sensors via a standardized interface. It uses a STEP-based parser to extract features from the geometric model files, obtaining manufacturing feature entities. Based on a set BiLSTM-CRF model, it extracts entities from the process parameter documents. Using an XML Schema template, it converts the extracted process parameter entities into a tree structure and applies logical constraints to the tree structure to obtain process parameter entities. Based on the manufacturing feature entities, process parameter entities, and industrial robot equipment, it constructs a knowledge graph. It divides the real-time data streams from industrial sensors into time windows, calculates the statistical features of the data within each time window, and maps them to parameter nodes in the knowledge graph to obtain a graph database. This improves the efficiency of using process data from industrial robots.

[0149] Those skilled in the art will understand that embodiments of the present invention can be provided as methods, apparatus, or computer program products. Therefore, the present invention can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention can take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0150] The computer program instructions can also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer-implemented process such that the instructions which execute on the computer or other programmable apparatus provide steps for implementing the functions specified in the flowchart block or blocks. Figure 1 one or more flowcharts and / or blocks Figure 1 means for functionally implementing the steps listed in the flowchart block or blocks.

[0151] The computer program instructions can also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer-implemented process such that the instructions which execute on the computer or other programmable apparatus provide steps for implementing the functions specified in the flowchart block or blocks. Figure 1 one or more flowcharts and / or blocks Figure 1 means for functionally implementing the steps listed in the flowchart block or blocks.

[0152] The computer program instructions can also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer-implemented process such that the instructions which execute on the computer or other programmable apparatus provide steps for implementing the functions specified in the flowchart block or blocks. Figure 1 one or more flowcharts and / or blocks Figure 1 means for functionally implementing the steps listed in the flowchart block or blocks.

[0153] The principles and implementations of the present application have been described in the specific embodiments. The above description is only used to help understand the method of the present application and its core idea; meanwhile, for those skilled in the art, according to the idea of the present application, the specific implementation and application range can be changed; in summary, the content of the specification should not be understood as the limitation of the present application.

Claims

1. A method for storing robot process data, characterized in that, The method includes: The system collects multi-source heterogeneous raw data from industrial robots through a standardized interface. The multi-source heterogeneous raw data includes geometric model files, process parameter documents, and real-time data streams from industrial sensors. The point cloud data in the geometric model file is reconstructed using a preset Poisson reconstruction algorithm to determine the corresponding standard geometric model file. A parser based on the STEP standard is used to extract features from the standard geometric model file to determine the corresponding manufacturing feature entities. The process parameter document is extracted using a set BiLSTM-CRF model to determine the corresponding initial process parameter entities. An XML Schema template is defined, and the initial process parameter entities are converted into a tree structure based on the XML Schema template to determine the corresponding tree process parameter structure. Logical constraints are applied to the tree process parameter structure based on preset process condition constraint rules and preset quality evaluation indicators to determine the corresponding process parameter entities. A knowledge graph is constructed based on the manufacturing feature entity, the process parameter entity, and the industrial robot equipment. The real-time data stream of the industrial sensor is divided into time windows according to the process stage to determine multiple corresponding time windows. The relationship mapping between the data statistical characteristics in each time window and the parameter nodes of the corresponding stage in the knowledge graph is performed to determine the corresponding graph database. The graph database is used to store robot process data.

2. The robot process data storage method according to claim 1, characterized in that, The step of reconstructing the boundary of the point cloud data in the geometric model file according to the preset Poisson reconstruction algorithm to determine the corresponding standard geometric model file includes: The point cloud data is normal vector estimated and voxelized according to the preset Poisson reconstruction algorithm. The voxelized point cloud data is then fitted according to the implicit function fitting algorithm to determine the corresponding isosurface. The network model of the isosurface is extracted using the moving cube algorithm to determine the corresponding standard geometric model file.

3. The robot process data storage method according to claim 1, characterized in that, Before performing entity extraction on the process parameter document based on the set BiLSTM-CRF model to determine the corresponding initial process parameter entities, the following steps are included: Construct a domain-adaptive process text corpus, and use a hierarchical annotation strategy to annotate the process text corpus with process parameter entities to determine the corresponding process text training set. The initial BiLSTM-CRF model is trained based on the process text training set to determine the corresponding BiLSTM-CRF model. The BiLSTM layer of the initial BiLSTM-CRF model is equipped with a bidirectional gating mechanism, and the hidden states of the forward and backward layers are fused through learnable weights. The CRF layer of the initial BiLSTM-CRF model constrains the label transfer rules.

4. The robot process data storage method according to claim 1, characterized in that, The defined XML Schema template includes: Define the hierarchical relationship of process parameters, wherein the hierarchical relationship of process parameters includes process type, parameter group and parameter value; An XML Schema template is defined based on the hierarchical relationship of the process parameters. The XML Schema template includes a three-layer structure of process type classification nodes, parameter group nodes, and parameter value nodes. The parameter group nodes are logically grouped according to equipment attributes, material attributes, and environmental attributes. The parameter value nodes are bound to measurement units and valid value range attributes.

5. The robot process data storage method according to claim 1, characterized in that, The step of logically constraining the tree-like process parameter structure according to preset process condition constraint rules and preset quality evaluation indicators to determine the corresponding process parameter entities includes: The process parameter document is parsed according to the preset rule engine, and the parsed condition statements are converted into XPath process condition constraint rules. A parameter-quality correlation matrix is ​​constructed based on preset quality evaluation indicators. Logical constraints are applied to the tree-like process parameter structure based on the XPath process condition constraint rules and the parameter-quality correlation matrix to determine the corresponding process parameter entities.

6. The robot process data storage method according to claim 1, characterized in that, The construction of a knowledge graph based on the manufacturing feature entity, the process parameter entity, and the industrial robot equipment includes: Create feature-parameter relationship edges based on necessity level and parameter combination constraints, and link the manufacturing feature entity to the process parameter entity based on the feature-parameter relationship edges; Based on the adaptation score, parameter-device relationship edges are created, and the process parameter entities are linked to industrial robot device nodes based on the parameter-device relationship edges to determine the corresponding knowledge graph.

7. The robot process data storage method according to claim 1, characterized in that, The process involves dividing the real-time data stream from the industrial sensors into time windows based on the process stages, determining multiple corresponding time windows, and mapping the statistical characteristics of the data within each time window to the parameter nodes of the corresponding stages in the knowledge graph to determine the corresponding graph database, including: Apache Flink is used to process sensor data streams in real time. Window functions are used to divide the data into window blocks according to the process stage. Statistical features are extracted from the sensor data in each window, and the corresponding process state vector is determined based on the statistical features. The process state vector is associated with the corresponding process node in the knowledge graph to determine the corresponding graph database.

8. A robot process data storage device, characterized in that, The device includes: The data acquisition module is used to acquire multi-source heterogeneous raw data of industrial robots through a standardized interface. The multi-source heterogeneous raw data includes geometric model files, process parameter documents, and real-time data streams from industrial sensors. The manufacturing and process entity determination module is used to perform boundary reconstruction on the point cloud data in the geometric model file according to the preset Poisson reconstruction algorithm, determine the corresponding standard geometric model file, perform feature extraction on the standard geometric model file using a parser based on the STEP standard, determine the corresponding manufacturing feature entities, extract entities from the process parameter document according to the set BiLSTM-CRF model, determine the corresponding initial process parameter entities, define an XML Schema template, convert the initial process parameter entities into a tree structure according to the XML Schema template, determine the corresponding tree process parameter structure, and apply logical constraints to the tree process parameter structure according to preset process condition constraint rules and preset quality evaluation indicators to determine the corresponding process parameter entities. The graph database determination module is used to construct a knowledge graph based on the manufacturing feature entity, the process parameter entity, and the industrial robot equipment; to divide the real-time data stream of the industrial sensor into time windows according to the process stage; to determine multiple corresponding time windows; and to map the relationship between the statistical characteristics of the data in each time window and the parameter nodes of the corresponding stage in the knowledge graph to determine the corresponding graph database, which is used to store robot process data.

9. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the program, it implements the steps of the robot process data storage method according to any one of claims 1 to 7.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When executed by a processor, the computer program implements the steps of the robot process data storage method according to any one of claims 1 to 7.

Images