A knowledge-driven automatic generation method for a part machining process and related products

Abstract

The present application relates to the field of intelligent manufacturing and digital process planning, and particularly relates to a knowledge-driven automatic generation method for part machining process and related products, which comprises constructing a knowledge graph for machining process, coupling the knowledge graph with a knowledge big model to form a feedback closed loop, extracting process knowledge units from multi-source data, quantitatively scoring and credibility screening the process knowledge units based on a preset knowledge evaluation algorithm to obtain a high-reliability process knowledge dataset, performing fine-grained attribute relationship extraction on the process knowledge in the process knowledge dataset, and dynamically updating the structured extraction results to the knowledge graph to realize incremental updating of knowledge. The present application solves the technical bottleneck in traditional process planning through the synergistic innovation of the four core links of efficient and automatic knowledge acquisition, structured and refined process knowledge expression, and dynamic expansion and real-time updating capability.

Description

Technical Field

[0001] This invention relates to the field of intelligent manufacturing and digital process planning, specifically to a knowledge-driven automatic generation method for part processing technology and related products. It belongs to the cross-disciplinary technologies of computer-aided process design (CAPP), knowledge engineering, artificial intelligence and manufacturing informatization, and is applicable to process planning, optimization and decision support in the processing of complex parts. Background Technology

[0002] In the context of intelligent transformation in manufacturing, the automatic generation technology of parts processing procedures serves as a core support for improving production efficiency and ensuring quality. Its technological maturity directly impacts the agile response capability of manufacturing systems. However, existing process knowledge sources are scattered and inconsistent in format, with varying data structures from different sources, including natural language descriptions, structured reports, sensor time-series data, and image data. Personnel typically need to manually clean, annotate, and organize various texts, which is time-consuming, error-prone, and makes it difficult to form a unified knowledge base that can be directly calculated and utilized.

[0003] Existing methods struggle to effectively assess the quality and reliability of these knowledge units, leading to frequent reliance on manual review or repeated trials in downstream process decisions, impacting production efficiency and cost control. While rule-based or machine learning-based process decision models have been applied to parameter recommendation in recent years, these models are often disconnected from domain ontology or expert systems, lacking a deep understanding of process constraints and contextual semantics. Furthermore, their knowledge bases are mostly statically maintained, unable to incorporate new materials, cutting tools, or process experiment results in real time, failing to meet ever-changing production demands. Although some research has organized processing information into knowledge graphs, the maintenance of nodes and relationships largely relies on manual rules or periodic updates, making it difficult to reflect new developments in the production environment promptly. In terms of knowledge retrieval, traditional keyword- or relation-based query methods are inefficient and have limited ability to match complex semantic combinations. Summary of the Invention

[0004] The technical problem to be solved by this invention is that existing knowledge management and decision support systems for parts processing technology generally suffer from difficulties in knowledge acquisition, insufficient quality assurance, weak dynamic evolution, and low retrieval efficiency. The purpose is to provide a knowledge-driven automatic generation method and related products for parts processing technology, which can automatically extract highly reliable knowledge from massive multi-source data, combine structured graphs and deep semantic reasoning, and provide a fully automated solution with dynamic expansion and real-time update capabilities.

[0005] This invention is achieved through the following technical solution:

[0006] A knowledge-driven automatic generation method for part machining processes includes:

[0007] A knowledge graph oriented towards processing technology is constructed, and the knowledge graph is coupled with a large knowledge model to form a feedback loop. The knowledge graph provides the large knowledge model with structured process constraints and semantic background. The large knowledge model makes inference decisions based on process constraints and feeds back the decision results to update the knowledge graph.

[0008] Process knowledge units are extracted from multi-source data, and quantitative scoring and credibility screening of process knowledge units are performed based on a preset knowledge evaluation algorithm to obtain a highly reliable process knowledge dataset.

[0009] Fine-grained attribute relationships are extracted from the process knowledge dataset, and the extracted results are structured and dynamically updated to the knowledge graph to achieve incremental knowledge updates.

[0010] Optionally, the steps for constructing a knowledge graph oriented towards processing technology include:

[0011] The steps to establish a machining process method chain include: introducing a process matching engine, establishing a two-way coupling mechanism driven by part requirements and enterprise resource capabilities, the two-way coupling mechanism derives a candidate set of machining methods from top to bottom based on part characteristics, and filters and replaces the candidate set from bottom to top based on enterprise resource capabilities to generate a machining process method chain that meets machining requirements and resource constraints.

[0012] The machining process method chain is mapped to a graph database. The machining parts, machining surfaces, operations, tools, fixtures and materials in the machining process method chain are used as instance nodes in the graph. The machining sequence, usage relationship or applicability relationship between nodes are used as edges in the graph for semantic connection to construct a knowledge graph.

[0013] Optionally, the steps for coupling knowledge graphs with large knowledge models to form a feedback loop include:

[0014] Vectorize the entities and relations in the knowledge graph and build a vector index library;

[0015] When a processing requirement is received, the subgraph related to the processing requirement is retrieved from the vector index library;

[0016] The subgraph and processing requirements are input into the knowledge model for reasoning, generating decision recommendations with confidence scores.

[0017] The credibility of newly generated knowledge during the reasoning process is evaluated, and knowledge that meets the preset credibility threshold after evaluation is incrementally updated to the knowledge graph and vector index library.

[0018] Optionally, the steps of extracting process knowledge units from multi-source data and performing quantitative scoring and credibility screening include:

[0019] Text data from multi-source data is preprocessed, and a BiLSTM-CNN-CRF architecture entity recognition model is used to extract process knowledge entities to obtain process knowledge entities.

[0020] For the extracted knowledge entities, an existence verification is performed in the knowledge graph. If the entity exists in the knowledge graph, a multi-hop path retrieval is initiated for the corresponding knowledge entity. If the entity does not exist in the knowledge graph, it is stored in a temporary cache queue.

[0021] A comprehensive path score is calculated based on multi-hop paths, and an adaptive screening threshold is determined based on the statistical distribution of the comprehensive path score to classify the reliability level of knowledge entities.

[0022] Optionally, in the input layer of the entity recognition model, the Shannon entropy calculated based on the label distribution of the local text context is concatenated with the embedding vector corresponding to the text to form an enhanced input feature, enabling the model's convolutional layer to adaptively focus on high-entropy regions during feature extraction.

[0023] Optionally, methods for calculating the overall path score include:

[0024] A three-factor linear combination model is used to weight and score multi-hop paths to obtain a comprehensive path score; the three factors include:

[0025] The path length factor is calculated based on the logarithm of the number of entity nodes in the path.

[0026] Termination node authority factor is quantified by comprehensively evaluating the global importance and local connectivity density of the terminal node in the path.

[0027] The relational coherence factor is obtained by calculating the semantic similarity between the embedding vectors of adjacent relationships on the path.

[0028] Optionally, the steps for classifying the reliability levels of knowledge entities include:

[0029] The path score of multiple paths associated with a knowledge entity is subjected to minimization normalization.

[0030] The weighted aggregation of multiple path scores after normalization is performed to calculate the final reliability score of the knowledge entity.

[0031] Based on the statistical distribution of the final reliability scores of all knowledge entities, an improved box plot method is used to determine the adaptive threshold.

[0032] Knowledge entities are classified into high reliability, low reliability, or unreliable levels based on the comparison results of the final reliability score with the third quartile and the adaptive threshold. Knowledge entities with a final reliability score greater than the adaptive threshold are considered high reliable, knowledge entities with a final reliability score less than or equal to the third quartile are considered unreliable, and the remaining knowledge entities are considered low reliable.

[0033] Optionally, multi-source data may include at least one or more of the following: process documents, standards and specifications, historical cases, structured reports, and sensor data;

[0034] The steps for preprocessing text data from multi-source datasets include:

[0035] A binary classifier based on BERT model fine-tuning is used to identify and label noise samples in text data through supervised learning methods;

[0036] Based on the identification and annotation results, data identified as noise samples are filtered out to obtain cleaned text.

[0037] The cleaned text is segmented into words to generate sequence labeling samples for subsequent extraction of process knowledge units.

[0038] Optionally, the steps for dynamically updating the knowledge graph include:

[0039] For knowledge entities classified as highly reliable, a Span-Based framework and a deep learning classifier are used to extract the attributes and relationships of the knowledge entities in a fine-grained manner.

[0040] The extracted attributes and relationships are encapsulated in a structured manner and automatically mapped to the corresponding ontology structure of the knowledge graph to realize the logical association of knowledge.

[0041] After the knowledge graph completes the mapping update, incremental vector embedding recalculation is triggered for newly added knowledge entities and relationships, and the retrieval index is updated synchronously.

[0042] A computer program product includes a computer program / instructions that, when executed by a processor, implement a knowledge-driven automatic generation method for part manufacturing processes as described above.

[0043] Compared with the prior art, the present invention has the following advantages and beneficial effects:

[0044] This invention solves the technical bottlenecks in traditional process planning through the collaborative innovation of four core links: efficient and automated knowledge acquisition, structured and refined process knowledge expression, and dynamic expansion and real-time updating capabilities.

[0045] By deeply coupling the structured graph with the pre-trained knowledge model, the system can utilize both the ontology constraints and semantic context provided by the graph, and leverage the deep understanding and reasoning capabilities of the large model in complex production scenarios. Through real-time feedback and incremental update mechanisms, the processing method chain is continuously verified and optimized to ensure the feasibility and accuracy of the process solution in a dynamic manufacturing environment.

[0046] Through large-scale knowledge extraction and pre-screening modules, the system automatically retrieves, preprocesses, filters, and extracts information from massive amounts of raw data from multiple sources. Combined with a reliable quantitative scoring algorithm, it can quickly produce highly reliable process knowledge units, significantly reducing the workload of manual cleaning, labeling, and sorting, and significantly improving the efficiency of knowledge acquisition.

[0047] A domain knowledge graph is constructed based on the machining process ontology. A Span-Based framework and a deep learning classifier are then used to extract fine-grained attributes and relationships from high-confidence entities. This enables precise, structured representations of core elements such as process parameters, tool characteristics, and operational logic, fully satisfying the dual requirements of semantic richness and logical consistency. Leveraging incremental vector embedding recomputation and vector retrieval index updates, the system can immediately incorporate newly added knowledge into the retrieval and reasoning process, ensuring that new knowledge is instantly detected and efficiently invoked in the subsequent "perception" stage, achieving continuous self-evolution of the knowledge graph.

[0048] This system features a modular and interface-based design, enabling seamless integration with new data sources, algorithm models, and ontology definitions. As application scenarios expand and feedback continues, its knowledge model will continuously improve itself, forming an efficient, scalable, and self-reinforcing intelligent manufacturing knowledge management and decision support platform. Attached Figure Description

[0049] The accompanying drawings illustrate exemplary embodiments of the present invention and, together with the description thereof, serve to explain the principles of the invention. These drawings are included to provide a further understanding of the invention and are incorporated in and constitute a part of this specification, but do not constitute a limitation on the embodiments of the present invention.

[0050] Figure 1 This is a flowchart illustrating a knowledge-driven automatic generation method for part processing technology according to the present invention.

[0051] Figure 2 This is a flowchart illustrating the process of constructing a knowledge graph oriented towards processing technology according to the present invention.

[0052] Figure 3 This is a schematic diagram of the process for quantitative scoring and credibility screening according to the present invention. Detailed Implementation

[0053] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are for illustrative purposes only and are not intended to limit the scope of the invention.

[0054] It should also be noted that, for ease of description, only the parts relevant to the present invention are shown in the accompanying drawings.

[0055] Where there is no conflict, the embodiments and features described in the present invention can be combined with each other. The present invention will now be described in detail with reference to the accompanying drawings and embodiments.

[0056] Example 1

[0057] like Figure 1 As shown, this embodiment provides a knowledge-driven automated generation system for the entire process of parts processing technology, aiming to solve the problems of scattered sources of knowledge for parts processing technology, inconsistent quality, low efficiency of manual processing, and the inability of isolated model decisions to evolve dynamically in the existing technology; specifically, it includes the following three steps:

[0058] Step 1: Construct a large-scale process knowledge model oriented towards the processing method chain.

[0059] A knowledge graph oriented towards the manufacturing process is constructed, and the knowledge graph is coupled with a large knowledge model to form a feedback loop. The knowledge graph provides the large knowledge model with structured process constraints and semantic background. The large knowledge model makes inference decisions based on process constraints and feeds back the decision results to update the knowledge graph.

[0060] A domain knowledge graph is constructed based on the manufacturing process ontology and deeply coupled with a large knowledge model. In this closed loop, the knowledge graph provides the model with structured process constraints and semantic background, while the pre-trained model undertakes deep understanding and reasoning decisions regarding the manufacturing context. The model's decision results are continuously updated through a feedback mechanism to update the knowledge graph, enabling feasibility verification, accuracy assurance, and continuous iterative optimization of the process method chain in a dynamic manufacturing environment.

[0061] The knowledge graph provides structured process constraints and semantic context, supporting the modeling of logical connections in the manufacturing process; the large knowledge model is responsible for deep understanding and intelligent reasoning of process requirements and the production environment, providing support for process decision-making. The system optimizes and verifies the processing method chain through a continuous feedback mechanism, ensuring the feasibility and accuracy of the process plan under dynamic manufacturing conditions.

[0062] Knowledge graphs store various entities (specific tools, processes, materials) in the field of machining processes and their precise relationships (applicable to, preceding processes, etc.) in the form of a network.

[0063] A knowledge-based big data model refers to a large-scale artificial intelligence model with powerful natural language understanding and logical reasoning capabilities. It can understand complex, unstructured process requirements and make inference decisions based on the constraints provided by the knowledge graph, such as planning process routes or recommending process parameters.

[0064] Feedback loop refers to the decision made by a knowledge model, after being verified, and the result (such as a new process combination that has been proven to be efficient) is used in turn to update and enrich the knowledge graph.

[0065] Step 2: Construct a highly reliable automatic extraction and screening mechanism for process knowledge.

[0066] Process knowledge units are extracted from multi-source data, and quantitative scoring and credibility screening of process knowledge units are performed based on a preset knowledge evaluation algorithm to obtain a highly reliable process knowledge dataset.

[0067] A large-scale retrieval of target information sources is carried out to collect massive amounts of raw data, which are then preprocessed. Based on knowledge filtering and information extraction technologies, the preprocessed raw data is processed and extracted to extract knowledge units with practical application value. Based on a preset knowledge evaluation algorithm, the extracted knowledge units are quantitatively scored and their credibility is screened to obtain a highly reliable knowledge dataset.

[0068] Multi-source data includes at least one or more of the following: process documents, standards and specifications, historical cases, structured reports, and sensor data. The system automatically retrieves and preprocesses massive amounts of raw data from these multi-source sources, extracts knowledge units with practical application value through filtering and information extraction techniques, and uses a reliability metric scoring algorithm to screen out highly reliable process knowledge, thereby providing solid data support for subsequent process decisions.

[0069] This step addresses the challenges of acquiring process knowledge and inconsistent quality in the manufacturing process by proposing a novel method that integrates multi-source heterogeneous information processing and reliable knowledge identification. This mechanism can automatically retrieve and efficiently preprocess massive amounts of raw data from various information sources, including process literature, standards and specifications, historical cases, and sensor data. Utilizing filtering and information extraction techniques based on natural language processing and semantic understanding, it accurately extracts knowledge units with practical application value. A reliability quantification and scoring algorithm is introduced to evaluate the reliability and prioritize the extracted knowledge units, selecting highly reliable process knowledge and significantly improving knowledge quality and usability.

[0070] Step 3: Construct a fine-grained attribute relationship extraction and dynamic graph update mechanism.

[0071] Fine-grained attribute relationships are extracted from the process knowledge dataset, and the extracted results are structured and dynamically updated to the knowledge graph to achieve incremental knowledge updates.

[0072] A fine-grained attribute and relation extraction and dynamic graph update mechanism is constructed. This involves locating the pre-screened high-confidence knowledge unit entities from step one, extracting span-based attributes and relations, constructing a fixed-length representation for each span, and using attribute and relation classifiers to identify whether it corresponds to entity attribute values ​​and semantic relationships between different entities. Structured encapsulation and domain graph mapping map the newly added nodes and edges to their corresponding ontology positions in the domain knowledge graph, completing semantic structuring and logical association. Dynamic indexing and large-scale knowledge model updates are performed incrementally to ensure that the latest supplemented knowledge can be retrieved and invoked instantly in the next closed loop.

[0073] Employing a Span-based information extraction framework, the system accurately locates the textual context of high-confidence entities and efficiently identifies the semantic relationships between entities and their attribute values ​​using a deep learning classifier, achieving a fine-grained, structured representation of process knowledge. Simultaneously, the extracted knowledge triples are automatically mapped to ontology nodes and edges in the domain knowledge graph, enabling complete organization and expression of knowledge through semantic association. Combined with incremental vector embedding recomputation and retrieval index update mechanisms, the system ensures that new knowledge can be instantly detected and efficiently invoked after each addition.

[0074] By achieving the three objectives mentioned above in a coordinated manner, this embodiment can significantly improve the efficiency and reliability of generating part processing solutions, reduce the cost of manual intervention, and continuously iterate and optimize itself in actual production, providing an efficient, scalable, and continuously self-reinforcing solution for automatically generating process knowledge for intelligent manufacturing.

[0075] Example 2

[0076] like Figure 2 As shown, this embodiment provides a specific implementation process for step one, which can be broken down into two main parts: first, the construction of the knowledge graph itself, and second, the operation mechanism of the feedback loop between it and the large knowledge model. That is, firstly, a high-quality process knowledge graph that conforms to theoretical logic and fits actual production capabilities is constructed through a two-way verification mechanism; then, based on this graph, a closed-loop workflow of "perception-decision-update" is established, enabling the system to respond to external demands, make intelligent decisions, and feed back the new knowledge generated in the decision-making process to the system in real time, achieving self-iteration.

[0077] Step 1.1: Establish the processing technology chain.

[0078] The process method chain is established by introducing a process matching engine and establishing a two-way coupling mechanism driven by part requirements and enterprise resource capabilities. The two-way coupling mechanism derives a candidate set of processing methods from top to bottom based on part characteristics and filters and replaces the candidate set from bottom to top based on enterprise resource capabilities to generate a process method chain that meets processing requirements and resource constraints.

[0079] To ensure the rationality and feasibility of the process method chain, the modeling process needs to comprehensively embed factors such as machining process rules, empirical constraints, and equipment feasibility. Specifically, this includes matching rules between machining features and machining methods, sequential logic rules between machining methods, usage restrictions on equipment / fixtures / tools, and compatibility constraints between process parameters. This is achieved by defining a part machining requirement function. With enterprise resource capability function , The geometric feature type to which the main machining surface belongs (such as hole, groove, plane, inclined plane, etc.); This includes the geometric and topological information of the part (such as shape features, connection relationships between adjacent surfaces, dimensional coordinates, etc.). For processing information related to the processing surface (such as the surface's location, orientation, accessibility, etc.); For user-defined machining technology requirements (such as accuracy, tolerance, surface roughness, heat treatment requirements, etc.).

[0080] A process matching engine is introduced to establish a two-way coupling mechanism driven by part requirements and combined with actual resource capabilities. This mechanism starts from the part characteristics from top to bottom, sequentially deriving the required processing features, operations, and process parameters, and generating a candidate set of processing methods. Simultaneously, it filters and replaces candidate processing methods from bottom to top based on the company's current resource capabilities, ensuring that the selected process path is feasible and executable. After completing the above two-way matching and multi-constraint filtering, the system finally outputs a set of processing method sequences that meet the processing requirements and are executable. Each of the processing methods All of these are related to the specific process, machining surface, and tool matching, and the setting of process parameters (by...). Decision), and implementable equipment and tooling resources (by...) The constraints, cutting tools, and process parameters are all matched one-to-one. The final generated machining method chain... It is based on and Optimization results under co-coupling.

[0081] Step 1.2: Map the processing method chain to the graph database.

[0082] The machining process method chain is mapped to a graph database. The machining parts, machining surfaces, operations, tools, fixtures and materials in the machining process method chain are used as instance nodes in the graph. The machining sequence, usage relationship or applicability relationship between nodes are used as edges in the graph for semantic connection to construct a knowledge graph.

[0083] A machining process knowledge graph is established. After constructing the machining method chain, the generated machining method chain is mapped to a graph database. Each machining entity, attribute, and its relationship is represented in a graph structure, thus constructing a standardized machining process knowledge graph. This knowledge graph is defined as... The nodes in the graph are composed of the core elements of the machining method chain, including the machined parts, machined surfaces, processes, cutting tools, fixtures, materials, and process parameters. Each machining method unit serves as an instance node in the graph, carrying multi-dimensional attribute information such as its associated geometric features, machining requirements, and equipment resources. Each process unit in the machining method chain serves as an instance node in the graph, and the constraints between nodes (such as "machining sequence," "fixture used," and "applicable cutting tools") are semantically connected as edges in the graph, forming a structured and queryable knowledge network.

[0084] Step 1.3: The feedback loop operation of knowledge graph and knowledge big model.

[0085] The entities and relationships in the knowledge graph are vectorized, transforming abstract process concepts into numerical vectors that can be mathematically operated on by computers, and a vector index library is constructed. When a processing requirement is received, a quick search is first performed in the vector index library to find the existing knowledge that is most semantically relevant to the current requirement, i.e., a subgraph. Then, the system inputs the subgraph and the processing requirement into the knowledge model for reasoning, generating decision suggestions with confidence scores.

[0086] The credibility of newly generated knowledge during the reasoning process is evaluated, and knowledge that meets the preset credibility threshold after evaluation is incrementally updated to the knowledge graph and vector index library.

[0087] The graph processed in step 1.2 is embedded into the graph neural encoding module of nano-graphrag to generate low-dimensional vectors for each entity and relation, and a vector index library is built simultaneously to support millisecond-level nearest neighbor retrieval. Then, when the system receives processing requirements and real-time features from users or devices, it retrieves the most relevant subgraph to the current scene in the graph vector space through the search(query_vector, top_k) interface of a large language model (such as DeepSeek), achieving accurate "perception" of historical process experience and parameters. Next, the system embeds the retrieved subgraph along with the original query and inputs it into the inference engine of nano-graphrag. Using a graph-text hybrid model, it performs deep reasoning and decision-making on processing parameters, tool selection, or process optimization schemes, and outputs the optimal suggestion with a confidence score. Finally, for newly identified or verified entities and relationships during the reasoning process, DeepSeek's verify(entity_vector,threshold) API is called again to evaluate their credibility. Only new knowledge that reaches the preset threshold will be written back to the main graph through the nano-graphrag incremental update interface and the vector index will be updated in real time, thus completing the entire "perception-decision-update" closed loop and ensuring that the processing technology knowledge base can evolve dynamically and continuously provide the latest and most reliable support for downstream applications.

[0088] Building upon the framework that integrates the processing method chain with the large knowledge model, a systematic knowledge extraction and screening mechanism is needed to ensure the quality and usability of knowledge input. This mechanism not only ensures the accuracy and reliability of the injected knowledge but also provides a solid foundation for subsequent graph updates and model inference.

[0089] Based on the processing technology ontology, a structured domain knowledge graph is constructed to systematically express key process elements and their relationships within the processing workflow. The knowledge graph ensures accuracy, professionalism, and scalability through the systematic organization and semantic modeling of process engineers' domain knowledge. A pre-trained general-purpose language model is deeply integrated with the process domain knowledge graph. The knowledge graph provides the large model with process rules, constraint boundaries, and contextual information, enabling the model to follow professional process logic when reasoning and generating process solutions, thus improving its interpretability and controllability. During the process of generating process solutions, the system automatically refers to the entity and attribute relationships in the graph to verify and correct the selected processing path.

[0090] Example 3

[0091] like Figure 3As shown in the figure, this embodiment provides a specific implementation process for step two. First, valuable information is quickly identified and extracted from a large amount of literature. Then, a preliminary background check is performed on this information. Finally, an objective score is given to the reliability of each piece of information through a quantitative scoring system, and the information is graded accordingly to ensure that only the highest quality knowledge can be adopted by the system.

[0092] Step 2.1, Multi-source data processing.

[0093] A binary classifier based on BERT model fine-tuning is used to identify and label noise samples in text data through supervised learning. Based on the identification and labeling results, data identified as noise samples are filtered out to obtain cleaned text. The cleaned text is then segmented to generate sequence labeling samples for subsequent extraction of process knowledge units.

[0094] For the acquired large-scale processing text data, preprocessing is first performed—special characters, garbled text, and irrelevant formatting marks are removed. Then, a supervised learning method is used to label noisy samples: a binary classifier is fine-tuned on the BERT model, labeling noisy samples as "Noise" and the remaining normal text as "Text". The labeled text is then normalized and converted to UTF-8 encoding. Based on the category label of each token, tokens identified as noise are filtered out, and the remaining tokens are reassembled in their original order to form the cleaned text. Next, the text is segmented into WordPiece fragments using BERT's built-in word segmenter, and alignment adjustments are made to address label misalignment issues caused by word segmentation, ultimately generating a format like... The sequence label samples not only ensured the standardization and consistency of the text, but also provided high-quality data input for subsequent downstream tasks.

[0095] Step 2.2, knowledge entity extraction.

[0096] Text data from multi-source datasets is preprocessed, and a BiLSTM-CNN-CRF architecture entity recognition model is used to extract process knowledge entities. BiLSTM-CNN-CRF is a model combination used in natural language processing for sequence labeling tasks (such as named entity recognition). It effectively combines contextual information of the text (via BiLSTM), local key features (via CNN), and the inherent constraints of the label sequence (via CRF) to accurately identify predefined entity categories in the text.

[0097] An algorithm for extracting knowledge entities related to processing technology is constructed. After preprocessing the text on processing technology (including denoising, unified encoding, and word segmentation), a BiLSTM–CNN–CRF architecture is used for knowledge entity extraction. Entropy reflects the uncertainty of the model's classification of each token in the local context. Inputting entropy as a priori feature into the CNN reveals which locations are prone to ambiguity, thus giving extra attention and suppression during feature extraction.

[0098] Step 2.2.1, Model optimization.

[0099] Traditional CNN modules rely solely on fixed-size convolutional kernels to extract features from local context at the character or sub-word level, making it difficult to distinguish the importance of different positions. To further improve the accuracy of entity extraction, this step introduces Shannon entropy, which measures the "uncertainty" of the model in identifying the category of a word (or character) in text. The higher the entropy value, the more uncertain the model's judgment is about that position, and the more ambiguous the situation may be.

[0100] In the input layer of the entity recognition model, the Shannon entropy, calculated based on the label distribution of the local text context, is concatenated with the embedding vector corresponding to the text to form enhanced input features, enabling the model's convolutional layers to adaptively focus on high-entropy regions during feature extraction.

[0101] First, within a sliding window for each token, the label distribution is statistically analyzed and the Shannon entropy is calculated. This serves as the entropy prior. Then, this entropy value is appended to the corresponding embedding vector to form the enhanced input: ;

[0102] in, Indicates the first The original embedding vector of each token, with dimension . , The vector concatenation operation represents the scalar... After being expanded into a one-dimensional vector, it is compared with the original Dimensional vector embedding and parallelization yields the amplified version. A dimensional vector is represented as , The Shannon information entropy, representing the distribution of tags within the sliding window for the token, is calculated using the following formula: ;

[0103] in, Is it predicted as the first in this local window? The probability of a class. High entropy. This indicates that the label distribution at this location is not concentrated in the context, and the prediction is uncertain. A low entropy value means that the model is very confident about the label in the local context. After this processing, the convolutional layer can adaptively focus on high-entropy regions when extracting local features.

[0104] Word-level embeddings are concatenated with character features derived from a CNN and then fed into a bidirectional LSTM to extract entities related to processing technology. The BiLSTM can simultaneously utilize forward and backward context to generate a hidden state sequence that considers both left- and right-side dependencies. Based on the BiLSTM hidden state sequence, a CRF layer performs joint decoding of the entire sentence through a globally optimal annotation path search, ensuring the consistency between the boundaries of processing technology entities and the label sequence.

[0105] Step 2.3, Existence verification.

[0106] For the extracted knowledge entities, an existence verification is performed in the knowledge graph. If the entity exists in the knowledge graph, a multi-hop path retrieval is initiated for the corresponding knowledge entity. If the entity does not exist in the knowledge graph, it is stored in a temporary cache queue.

[0107] After extracting the processing technology knowledge units from step 2.2, an existence verification of the knowledge unit graph is required. This involves retrieving the type definition library of the knowledge graph using SPARQL queries to check if the entity belongs to a predefined process concept category. If the entity exists in the graph, it is marked as a "registered trusted entity" and proceeds to the next analysis step; if it does not exist, it is temporarily classified as an "entity to be verified," requiring manual review or supplementary knowledge acquisition. This step ensures that the screening process is based on a verified knowledge system, avoiding interference from invalid entities.

[0108] Step 2.4, Path comprehensive scoring.

[0109] A comprehensive path score is calculated based on multi-hop paths, and an adaptive screening threshold is determined according to the statistical distribution of the comprehensive path score to classify the reliability level of knowledge entities. For entities that have been confirmed to exist, a multi-hop path discovery mechanism based on breadth-first search (BFS) is initiated, saving the order of traversed entity nodes, recording the semantic relationship types between nodes, and calculating in real time based on path length, node importance, and relationship density.

[0110] The method for calculating the overall path score includes: using a three-factor linear combination model to weight the scores of multi-hop paths to obtain the overall path score; the three factors include:

[0111] The path length factor, calculated based on the logarithm of the number of entity nodes in the path, is given by the following formula: ,in, It is the number of entity nodes in the path (e.g., from...). correspond ).

[0112] Termination node authority refers to the core position and influence of the terminal node of a certain association path in the entire knowledge network. This indicator is quantified by comprehensively evaluating the global importance and local connection density of the node. The formula for calculating the termination node authority factor, which is quantified by comprehensively evaluating the global importance and local connection density of the terminal node of the path, is as follows: , ,in, For the global importance of nodes, denoted as the local connectivity density of the node.

[0113] Relational coherence refers to the semantic similarity of adjacent relations calculated using relation type embedding vectors, ensuring the consistency of the path's process logic. The relational coherence factor is obtained by calculating the semantic similarity between the embedding vectors of adjacent relations along the path. The calculation formula is as follows: ,in, This represents the similarity between adjacent pairs. This is the attenuation factor.

[0114] The design path comprehensive scoring model adopts a three-factor linear combination model to form an interpretable weighted scoring system: ,in .

[0115] The path score of multiple paths associated with a knowledge entity is subjected to minimization normalization. , where P is the set of entity association paths.

[0116] The contribution of each path is clearly quantified, and the final reliability score of the knowledge entity is obtained by aggregation calculation; .

[0117] Step 2.5, Reliability level classification.

[0118] Based on the statistical distribution of the final reliability scores of all knowledge entities, an improved box plot method is used to determine the adaptive threshold. Among them, the screening boundary is adaptively determined using statistical methods. It is the first quartile. It is the third quartile. It is the interquartile range, and 1.5 is an empirical coefficient used to control the threshold and the degree of deviation of the main data distribution.

[0119] Based on the comparison results of the final reliability score with the third quartile and the adaptive threshold, a multi-level threshold design is adopted to classify knowledge entities into high reliability, low reliability, or unreliable levels. Among them, the final reliability score is greater than the adaptive threshold. The knowledge entity is considered highly reliable, with a final reliability score less than or equal to the third quartile. The knowledge entities of ) are unreliable, the rest The knowledge entity is of low reliability.

[0120] This embodiment constructs a process knowledge extraction and pre-screening module based on information filtering and knowledge extraction. It performs unified preprocessing on massive amounts of raw data such as text, reports, and sensor time series, and uses a preset quantitative scoring algorithm to score and screen the extracted knowledge units based on their credibility, automatically generating a highly reliable process knowledge set. This technology overcomes the inefficiency of traditional manual cleaning and semi-automatic annotation, significantly improving the efficiency and quality of cross-source process knowledge acquisition.

[0121] Example 4

[0122] This embodiment provides the specific process of step three, which is that after the system selects highly reliable knowledge, it performs in-depth analysis and seamlessly and dynamically integrates it into the existing knowledge system to achieve the system's self-evolution and capability enhancement.

[0123] A fine-grained knowledge extraction and dynamic graph update mechanism is introduced. This mechanism revolves around highly reliable entities, accurately locating their contextual information in the original text and extracting their attributes and relationships through a Span-Based model. The extracted structured knowledge units can be automatically mapped to the ontology structure of the existing graph, forming semantic connectivity and logical expansion. Simultaneously, after the graph is updated, it can be fed back to the larger model in real time, achieving synchronous improvement and a self-reinforcing closed loop between the knowledge base and model capabilities.

[0124] The specific steps include:

[0125] For knowledge entities classified as highly reliable, a Span-Based framework and a deep learning classifier are used to extract the attributes and relationships of the knowledge entities in a fine-grained manner. The context is located by tracing back to the original text, and a fixed-length representation is constructed for candidate word fragments based on the Span-Based framework. A deep attribute classifier and a relation classifier are used to identify the attribute values ​​of entities and the semantic relationships between entities, respectively.

[0126] The extracted attributes and relationships are encapsulated in a structured manner and automatically mapped to the corresponding ontology structure of the knowledge graph to realize the logical association of knowledge; the obtained triples (subject-verb-object) are encapsulated according to the specification and automatically mapped to the corresponding ontology nodes and edges of the domain knowledge graph to realize the refined structuring of process knowledge and the logical association of ontology.

[0127] After the knowledge graph completes the mapping update, it triggers the incremental vector embedding recalculation for the newly added knowledge entities and relationships, and synchronously updates the retrieval index. After the incremental writing of graph nodes and edges is completed, it automatically triggers the vectorization engine to recalculate the low-dimensional embeddings for the newly added entities and relationships, and performs incremental updates in the vector retrieval index to ensure that the newly supplemented process knowledge can be retrieved immediately and efficiently invoked, realizing the dynamic expansion and continuous self-strengthening of the knowledge graph.

[0128] Example Five

[0129] This embodiment provides a specific example.

[0130] 1. Automatic knowledge learning and update.

[0131] The system discovers a new type of drill bit called "ABC-123" from a newly published technical article, which performs excellently when machining TC4 titanium alloy.

[0132] The system executes an internal process: extracts the knowledge entity of "ABC-123 drill bit", evaluates it as a high reliability level by analyzing its association path with existing knowledge. Then, it performs fine-grained extraction on this entity to clarify its attributes and relationships, such as <ABC-123, applicable material, TC4 titanium alloy>. Finally, these new knowledge are dynamically updated to the knowledge graph and its retrieval index.

[0133] 2. Intelligent process task execution.

[0134] After the engineer submits a process planning request for "TC4 titanium alloy parts", when the system retrieves the solution, it retrieves "TABC-123" as highly relevant knowledge.

[0135] The system finally generates an optimized process plan, which clearly recommends "using ABC-123 drill bit for the finishing process" and gives a confidence score.

[0136] Example Six

[0137] A knowledge-driven automatic generation terminal for a part processing process includes a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, it implements the above-mentioned knowledge-driven automatic generation method for the part processing process.

[0138] The memory can be used to store software programs and modules. The processor executes various functional applications and data processing of the terminal by running the software programs and modules stored in the memory. The memory mainly includes a program storage area and a data storage area. Among them, the program storage area can store an operating system, execution programs required for at least one function, etc.

[0139] The storage data area can store data created based on the use of the terminal. Furthermore, the memory can include high-speed random access memory, and may also include non-volatile memory, such as at least one disk storage device, flash memory, or other volatile solid-state storage devices.

[0140] A computer-readable storage medium storing a computer program that, when executed by a processor, implements the aforementioned knowledge-driven automatic generation method for part processing technology.

[0141] Without loss of generality, computer-readable media can include computer storage media and communication media. Computer storage media includes volatile and non-volatile, removable and non-removable media implemented using any method or technology for storing information such as computer-readable instruction data structures, program modules, or other data. Computer storage media includes RAM, ROM, EPROM, EEPROM, flash memory or other solid-state storage technologies, CD-ROM, DVD or other optical storage, magnetic tape cassettes, magnetic tape, disk storage, or other magnetic storage devices. Of course, those skilled in the art will recognize that computer storage media are not limited to the above-mentioned types. The aforementioned system memories and mass storage devices can be collectively referred to as memory.

[0142] A computer program product includes a computer program / instructions that, when executed by a processor, implement the knowledge-driven automatic generation method for part processing technology as described above.

[0143] Computer program products include computer programs or instruction sets used to perform specific tasks or achieve specific functions. These programs or instructions are designed to be executed by a processor to implement a series of predefined steps or operations. The program product may be stored in various forms of computer storage media, such as memory, hard disks, solid-state drives, optical discs, or other forms of digital storage devices. It may exist in the form of compiled binary code or in the form of scripts or bytecode that can be executed by an interpreter. Through carefully designed algorithms and logical instructions, the program product enables the processor to process data in a specific order and manner, performing various functions such as data analysis, user interaction, and device control.

[0144] In the description of this specification, the references to terms such as "one embodiment / mode," "some embodiments / modes," "example," "specific example," or "some examples," etc., indicate that a specific feature, structure, material, or characteristic described in connection with that embodiment / mode or example is included in at least one embodiment / mode or example of this application. In this specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment / mode or example. Moreover, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments / modes or examples. Furthermore, without contradiction, those skilled in the art can combine and integrate the different embodiments / modes or examples described in this specification, as well as the features of different embodiments / modes or examples.

[0145] Furthermore, the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of technical features indicated. Thus, a feature defined as "first" or "second" may explicitly or implicitly include at least one of that feature. In the description of this application, "multiple" means at least two, such as two, three, etc., unless otherwise explicitly specified.

[0146] Those skilled in the art should understand that the above embodiments are merely for illustrating the present invention and are not intended to limit the scope of the invention. Those skilled in the art can make other changes or modifications based on the above invention, and these changes or modifications still fall within the scope of the present invention.

Claims

1. A knowledge-driven automatic generation method of a part machining process, characterized in that, The method comprises the following steps: constructing a knowledge graph for a machining process, and coupling the knowledge graph with a knowledge large model to form a feedback loop, wherein the knowledge graph provides structured process constraints and semantic background for the knowledge large model, the knowledge large model makes reasoning and decision-making according to the process constraints, and the decision-making results are fed back to update the knowledge graph; extracting process knowledge units from multi-source data, and quantitatively scoring and screening the process knowledge units based on a preset knowledge evaluation algorithm to obtain a high-reliability process knowledge dataset; performing fine-grained attribute relationship extraction on the process knowledge in the process knowledge dataset, and dynamically updating the extracted results to the knowledge graph to realize incremental updating of the knowledge; wherein the steps of extracting process knowledge units from multi-source data and performing quantitative scoring and reliability screening comprise: performing preprocessing on text data in the multi-source data, and using an entity recognition model with a BiLSTM-CNN-CRF architecture to extract machining process knowledge entities, thereby obtaining process knowledge entities; verifying the existence of the extracted knowledge entities in the knowledge graph, if the knowledge entities exist in the knowledge graph, initiating multi-hop path retrieval for the corresponding knowledge entities, if the knowledge entities do not exist in the knowledge graph, storing them in a temporary cache queue; calculating a path comprehensive score based on the multi-hop path, and determining an adaptive screening threshold according to the statistical distribution of the path comprehensive score to divide the knowledge entities into reliability levels; the method for calculating the path comprehensive score comprises: using a three-factor linear combination model to weight the multi-hop path to obtain the path comprehensive score; the three factors include: a path length factor calculated based on the logarithm of the number of entity nodes in the path; a termination node authority factor quantified by comprehensively evaluating the global importance and local connection density of the end nodes of the path; a relationship coherence factor obtained by calculating the semantic similarity between embedded vectors of adjacent relationships in the path.

2. The knowledge-driven automatic generation method of part machining process according to claim 1, characterized in that, The steps of constructing the knowledge graph for the machining process comprise: establishing a machining process method chain, and the steps of establishing the machining process method chain comprise: introducing a process matching engine, establishing a bidirectional coupling mechanism driven by part requirements and enterprise resource capabilities, the bidirectional coupling mechanism generates a candidate set of machining methods from top to bottom according to part features, and filters and replaces the candidate set from bottom to top according to enterprise resource capabilities, to generate a machining process method chain that meets the machining requirements and resource constraints; mapping the machining process method chain to a graph database, wherein the machining parts, machining surfaces, processes, tools, fixtures and materials in the machining process method chain are taken as instance nodes in the graph, and the machining sequence, use relationship or application relationship between the nodes are taken as edges in the graph for semantic connection, to construct the knowledge graph.

3. The knowledge-driven automatic generation method of part machining process according to claim 1, characterized in that, The steps of coupling the knowledge graph with the knowledge large model to form the feedback loop comprise: vectorizing the entities and relationships in the knowledge graph, and constructing a vector index library; when receiving a machining requirement, retrieving a subgraph related to the machining requirement in the vector index library; inputting the subgraph and the machining requirement into the knowledge large model for reasoning to generate a decision-making suggestion with a confidence score; The newly generated knowledge in the reasoning process is evaluated for credibility, and the knowledge satisfying a preset credibility threshold after evaluation is incrementally updated to the knowledge graph and the vector index library.

4. The knowledge-driven automatic generation method of part machining process according to claim 1, characterized in that, In the input layer of the entity recognition model, the Shannon entropy calculated according to the label distribution of the local context of the text is spliced with the embedding vector corresponding to the text to form an enhanced input feature, so that the convolutional layer of the model can adaptively focus on the high-entropy area during feature extraction.

5. The knowledge-driven automatic generation method of part machining process according to claim 1, characterized in that, The step of dividing the knowledge entities into reliability levels comprises: The path comprehensive scores of the multiple paths associated with a knowledge entity are normalized by maximum and minimum processing; The multiple path scores after normalization processing are weighted and aggregated to calculate the final reliability score of the knowledge entity; Based on the statistical distribution of the final reliability scores of all knowledge entities, an improved box plot method is used to determine an adaptive threshold; According to the comparison result of the final reliability score, the third quartile and the adaptive threshold, the knowledge entity is divided into high reliability, low reliability or unreliable levels, wherein the knowledge entity with a final reliability score greater than the adaptive threshold is high reliability, the knowledge entity with a final reliability score less than or equal to the third quartile is unreliable, and the remaining knowledge entity is low reliability.

6. The knowledge-driven automatic generation method of part machining process according to claim 1, characterized in that, The multi-source data at least includes one or more of the following: process literature, standard specification, historical case, structured report and sensor data; The step of preprocessing the text data in the multi-source data comprises: A binary classifier based on BERT model fine-tuning is used to identify and label noise samples in the text data by a supervised learning method; According to the identification and labeling results, the data identified as noise samples are filtered out to obtain cleaned text; The cleaned text is segmented to generate sequence labeling samples for subsequent extraction of process knowledge units.

7. The knowledge-driven automatic generation method of part machining process according to claim 5, characterized in that, The step of dynamically updating the knowledge graph comprises: For the knowledge entities divided into high reliability, a Span-Based framework and a deep learning classifier are used to perform fine-grained extraction of attributes and relationships of the knowledge entities; The extracted attributes and relationships are structured and encapsulated, and automatically mapped to the corresponding ontology structure of the knowledge graph to realize logical association of knowledge; After the mapping update of the knowledge graph is completed, the newly added knowledge entities and relationships trigger incremental vector embedding recalculation, and the search index is updated synchronously.

8. A computer program product comprising computer programs / instructions, characterized in that, The computer program / instructions are executed by the processor to implement the knowledge-driven automatic generation method of the part machining process according to any one of claims 1-7.

Images