Production progress collaborative operation intelligent management system based on large model

Abstract

The invention discloses a production progress collaborative operation intelligent management system based on a large model, and relates to the technical field of artificial intelligence, and the system comprises a multi-source production data collection module, a cross-enterprise semantic adaptation module, a bidirectional collaboration module, a scene intention recognition and active collaboration module, and a large model decision engine module. And all the modules are linked based on the data stream and the decision stream. According to the invention, the large model decision engine module, the multi-source production data acquisition module, the cross-enterprise semantic adaptation module, the bidirectional collaboration module and the scene intention recognition and active collaboration module form bidirectional linkage of data flow and decision flow, so that intelligent closed-loop decision of production progress collaboration management is realized. According to the invention, the continuity of cross-module collaboration and the fitness of decision schemes can be improved, so that the problems of disjunction of decision support and business execution and insufficient practicability of scheduling schemes caused by multi-source data decentralized processing in a traditional production collaboration system can be solved.

Description

Technical Field

[0001] This invention relates to the field of artificial intelligence technology, and in particular to an intelligent management system for collaborative production schedule operations based on a large model. Background Technology

[0002] The intelligent management system for collaborative production progress can monitor key information such as production progress, material supply, and equipment status, provide early warnings of abnormal situations, optimize resource allocation and work scheduling, reduce communication costs and waiting losses, improve production collaboration efficiency and progress controllability, and achieve cost reduction, efficiency improvement, and refined production management.

[0003] Currently, intelligent management systems for collaborative production schedules suffer from several problems, including a disconnect between decision support and business execution, fragmented processing of multi-source data leading to insufficient practicality of scheduling schemes, limited data collection dimensions, semantic differences in cross-enterprise data causing distortion in collaborative information transmission, and a separation between quality and schedule management, as well as passive responses to collaborative needs. These issues severely impact the continuity of production collaboration, the accuracy of data interaction, and overall collaborative efficiency, making it difficult to guarantee the core objectives of on-time order delivery and cost reduction and efficiency improvement.

[0004] Therefore, we propose an intelligent management system for collaborative production schedule operations based on a large model to solve the above problems. Summary of the Invention

[0005] The main objective of this invention is to provide an intelligent management system for collaborative production schedule operations based on a large model, in order to solve the problems mentioned in the background above.

[0006] To achieve the above objectives, the technical solution adopted by the present invention is as follows: an intelligent management system for collaborative production progress based on a large model, the system comprising a multi-source production data acquisition module, a cross-enterprise semantic adaptation module, a two-way collaboration module, a scene intent recognition and proactive collaboration module, and a large model decision engine module, wherein each module is linked based on data flow and decision flow; The multi-source production data acquisition module is used to collect multi-source production data, preprocess the collected data, and output standardized production data. The cross-enterprise semantic adaptation module is used to receive the standardized production data, construct an industrial semantic graph based on a large model, perform semantic parsing and alignment on cross-enterprise heterogeneous system data, and output semantically aligned production data. The bidirectional collaboration module is used to receive the semantically aligned production data, establish a semantic association model of production progress and quality data, realize the linkage between quality anomaly tracing and progress adjustment, and output progress and quality linkage data. The scenario intent recognition and proactive collaboration module is used to receive the semantically aligned production data, analyze the interaction behavior and real-time operation scenario of the collaborative roles, identify potential collaborative needs and push the associated semantically aligned production data and decision support information, and output collaborative need data. The large model decision engine module is used to provide semantic reasoning support for the cross-enterprise semantic adaptation module, bidirectional collaboration module, and scene intent recognition and proactive collaboration module. At the same time, it integrates the standardized production data, semantically aligned production data, progress and quality linkage data, and collaboration requirement data to construct a multi-objective decision model. Based on system operation feedback, it optimizes the large model and outputs a production scheduling plan.

[0007] Preferably, the multi-source production data acquisition module is specifically used for: Collect order information, bill of materials, and process progress data from the enterprise's internal ERP, MES, and QMS systems; Collect material flow data, order delivery data, and logistics transportation data from upstream and downstream enterprises in the industrial chain; The collected text data such as process documents, sensor time-series data, and structured data such as orders and inventory are cleaned, deduplicated, and formatted. Data of different formats are transformed into a unified data format that can be processed by large models, forming and outputting standardized production data.

[0008] Preferably, the cross-enterprise semantic adaptation module is specifically used for: Based on large-scale model training, we construct an industrial semantic graph that includes data terms, semantic connotations, and relationships. Extract textual descriptions and data tags of production-related data from different industrial systems and map them to the corresponding nodes of the industrial semantic graph; The large model is used to calculate the association paths and similarities of differentiated representations in the semantic graph, and semantic association weights are generated. Based on the semantic association weights, the reliability of the data source, and the contextual features of the data representation, the semantic alignment confidence between cross-enterprise data representations is calculated. When the semantic alignment confidence level reaches a preset threshold, the semantic alignment is confirmed to be effective and the semantically aligned production data is output.

[0009] Preferably, the bidirectional collaborative module is specifically used for: By establishing semantic associations between production schedule nodes and corresponding quality inspection results, material qualification certificates and equipment operating parameters through a large model, a schedule and quality semantic association model is formed. When a quality anomaly is detected, the production progress node corresponding to the anomaly is located based on the progress and quality semantic association model, and the material batch information, operation execution records and equipment operating parameters associated with the node are traced along the semantic association link. Based on the traceability results, a schedule adjustment plan is generated, which includes suggestions for process adjustments and resource allocation plans, and outputs schedule and quality linkage data.

[0010] Preferably, the scene intent recognition and proactive collaboration module is specifically used for: Collect historical query records, operation execution records, and real-time operation behaviors of collaborative roles to build user behavior profiles and scenario feature libraries; Based on the reasoning capabilities of large models, combined with user behavior profiles and real-time operation scenarios, potential collaborative needs of collaborative roles are identified. Based on the identified potential collaboration needs, relevant information is extracted from semantically aligned production data to generate decision support information, which is then pushed out through system pop-ups, mobile notifications, and industrial tablet prompts to output collaboration need data.

[0011] Preferably, the large model decision engine module is specifically used to provide semantic reasoning support for: By calling pre-trained large-scale models in the industrial field, inference support is provided for the semantic parsing process of the cross-enterprise semantic adaptation module, and the accuracy of semantic mapping is verified. Provide reasoning verification for the progress and quality correlation logic of the two-way collaborative module to ensure the correctness of the anomaly tracing path; It provides semantic reasoning support for the requirement identification process of the scene intent recognition and proactive collaboration module.

[0012] Preferably, the large model decision engine module is specifically used in constructing a multi-objective decision model for: It aggregates semantically aligned production data from the cross-enterprise semantic adaptation module, progress and quality linkage data from the bidirectional collaboration module, and collaboration requirement data from the scene intent recognition and proactive collaboration module. A multi-objective decision-making model is constructed by taking production schedule coordination, resource allocation optimization, and quality risk control as decision-making objectives. A production scheduling scheme is generated and output based on the multi-objective decision model.

[0013] Preferably, the cross-enterprise semantic adaptation module, when extracting textual descriptions and data tags of production-related data from different industrial systems and mapping them to corresponding nodes in the industrial semantic graph, is specifically used for: The text is segmented and tagged with parts of speech to extract core semantic keywords; The core semantic keywords are matched with the term nodes in the semantic graph to determine the initial mapping relationship; Based on the attribute information of the data tags, including data type and business scenario, the initial mapping relationship is corrected to obtain the final mapping node.

[0014] Preferably, when the bidirectional collaboration module generates a schedule adjustment plan containing process adjustment suggestions and resource allocation plans based on the traceability results, it is specifically used for: Based on the traceability results, and considering the severity of the quality anomaly, the schedule criticality of the process, and the flexibility of resource substitution, the impact of the quality anomaly on the production schedule is quantified. Based on the magnitude of the impact, corresponding level of process adjustment suggestions and resource allocation plans are generated.

[0015] Preferably, the scene intent recognition and proactive collaboration module is specifically used for: extracting related information and generating decision support information. Determine push channels based on user behavior profiles; Based on the scene feature library, semantically aligned production data that is highly relevant to the current scene is matched, including the inventory status of related materials and the equipment readiness status of the next process. The extracted correlation information and the generated decision support information are packaged together, pushed out through the determined push channels, and the push time and user viewing status are recorded.

[0016] The present invention has the following beneficial effects: 1. This invention establishes a two-way linkage between data flow and decision flow through a large-model decision engine module, a multi-source production data acquisition module, a cross-enterprise semantic adaptation module, a two-way collaboration module, and a scenario intent recognition and proactive collaboration module. This provides semantic reasoning support for each core module and integrates multiple types of production-related data to construct a multi-objective decision model, thereby realizing intelligent closed-loop decision-making for collaborative production progress management. Compared with existing technologies, this invention can improve the coherence of cross-module collaboration and the fit of decision schemes. Therefore, it can solve the problems of disconnect between decision support and business execution and insufficient practicality of scheduling schemes caused by the decentralized processing of multi-source data in traditional production collaboration systems.

[0017] 2. This invention covers data from multiple internal systems and the entire supply chain within an enterprise through a multi-source production data acquisition module. Combined with a cross-enterprise semantic adaptation module based on keyword mapping, similarity calculation, and confidence verification mechanisms using an industrial semantic graph, it achieves standardized processing of multi-dimensional production data and semantic alignment of heterogeneous cross-enterprise data. Compared with existing technologies, this invention can improve the completeness of production data and the accuracy of cross-enterprise data interaction. Therefore, it can solve the problems of limited data acquisition dimensions in traditional systems and the distortion of collaborative information transmission caused by semantic differences in cross-enterprise data.

[0018] 3. This invention establishes a semantic association model between production progress and quality data through a two-way collaboration module to achieve linkage between anomaly tracing and progress adjustment. At the same time, the scenario intent recognition and proactive collaboration module proactively identifies potential collaboration needs and accurately pushes related information based on user behavior profiles and real-time scenario features. Compared with existing technologies, this invention can improve the timeliness of quality anomaly handling and the targeting of collaboration services. Therefore, it can solve the problems of low production collaboration efficiency caused by the separation of quality and progress management and passive response to collaboration needs in traditional collaboration systems. Attached Figure Description

[0019] Figure 1 This is a schematic diagram of the overall system architecture of the present invention; Figure 2 This is a schematic diagram of the multi-source production data acquisition module of the present invention; Figure 3 This is a schematic diagram of the bidirectional collaborative module of the present invention. Detailed Implementation

[0020] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below. Obviously, the described embodiments belong to some, but not all, embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0021] The terminology used in the embodiments of this invention is for the purpose of describing particular embodiments only and is not intended to limit the invention. The singular forms “said” and “the” as used in the embodiments of this invention and the appended claims are also intended to include the plural forms, and “multiple” generally includes at least two unless the context clearly indicates otherwise.

[0022] Depending on the context, the word "if" or "if" as used here can be interpreted as "when," "when," "in response to determination," or "in response to detection." Similarly, depending on the context, the phrase "if determination" or "if detection (of the stated condition or event)" can be interpreted as "when determination," "in response to determination," "when detection (of the stated condition or event)," or "in response to detection (of the stated condition or event)."

[0023] Furthermore, the timing of the steps in the following method embodiments is merely an example and not a strict limitation.

[0024] In practice, the server-side equipment deployed in the intelligent production schedule collaborative operation management system based on a large model of the present invention may consist of one or more devices. The system can be implemented as business instances, virtual machines, or hardware devices. For example, it can be implemented as a business instance deployed on one or more devices in a cloud node, i.e., software deployed on the cloud node, used to provide intelligent management services for collaborative production schedule operations to various enterprises; it can also be implemented as a virtual machine deployed on one or more devices in a cloud node, with application software installed for cross-enterprise production collaborative management; or it can be implemented as a server-side system composed of numerous identical or different types of hardware devices, with one or more hardware devices specifically designated to provide collaborative management services to upstream and downstream enterprises in the industry chain.

[0025] In terms of implementation, the system of this invention is compatible with the enterprise end. That is, when the system is implemented as an application installed on a cloud service platform, the enterprise end acts as a client that establishes a communication connection with the application; when the system is implemented as a website, the enterprise end accesses it through a webpage; when the system is implemented as a cloud service platform, the enterprise end can access it through a mini-program in an instant messaging application.

[0026] The system of this invention can be set up in a cloud server. In terms of implementation, it can be used as one or more service devices, or as an application installed in the cloud, or developed as a dedicated industrial collaborative management website.

[0027] In this embodiment of the invention, each module of the system can be implemented independently and can call other modules. Here, "calling" refers to a module connecting to multiple modules of another type and providing corresponding services to those connected modules. In the system provided by this embodiment, the applicability of the system architecture can be adjusted by adding modules and directly calling them without modifying the program code, achieving cluster-based horizontal expansion and enabling quick and flexible expansion of system application scenarios. In practical applications, the above modules can be set in the same device or different devices, or in virtual devices, such as service instances in a cloud server.

[0028] The following detailed descriptions of the various components and specific workflows of the system of the present invention, with reference to specific embodiments, are provided below.

[0029] Example 1, please refer to Figure 1 and Figure 2 As shown: A production schedule collaborative operation intelligent management system based on a large model. The system includes a multi-source production data acquisition module, a cross-enterprise semantic adaptation module, a two-way collaboration module, a scene intent recognition and proactive collaboration module, and a large model decision engine module. Each module is linked based on data flow and decision flow. The multi-source production data acquisition module is used to collect multi-source production data, preprocess the collected data, and output standardized production data.

[0030] In this embodiment of the invention, when the multi-source production data acquisition module performs the acquisition and preprocessing of multi-source production data, it is specifically used for: Collect order information, bill of materials, and process progress data from the enterprise's internal ERP, MES, and QMS systems; Collect material flow data, order delivery data, and logistics transportation data from upstream and downstream enterprises in the industrial chain; The collected text data such as process documents, sensor time-series data, and structured data such as orders and inventory are cleaned, deduplicated, and formatted. Data of different formats are transformed into a unified data format that can be processed by large models, forming and outputting standardized production data.

[0031] Specifically, for data collection from internal enterprise systems, the multi-source production data collection module establishes a stable data transmission channel through the APIs reserved by each system, and adopts a hybrid collection mode that combines timed retrieval with real-time push.

[0032] For data with low update frequency, such as order information and bill of materials, set up a timed fetch once per hour; For data with high timeliness requirements, such as process progress data and real-time production status, a real-time push mechanism based on message queues is adopted to ensure that the data delay does not exceed 3 seconds.

[0033] The SSL protocol is used during data transmission to prevent data from being tampered with or leaked.

[0034] Specifically, for data collection from upstream and downstream enterprises in the industrial chain, a cross-enterprise encrypted data exchange platform is built using the multi-source production data collection module. The platform uses the HTTPS protocol for data transmission and establishes an enterprise identity authentication mechanism, allowing only authorized partner enterprises to access the platform.

[0035] The collected material flow data includes key information such as material number, flow quantity, flow time, and handler; order delivery data includes order number, delivery quantity, delivery time, and acceptance results; logistics and transportation data includes waybill number, transportation route, transit status, and estimated delivery time.

[0036] Specifically, the data preprocessing stage employs differentiated processing strategies for different data types during the cleaning process. For text data such as process documents, the multi-source production data acquisition module uses regular expressions to remove invalid characters, redundant spaces, and duplicate paragraphs, while natural language processing technology corrects typos and grammatical errors. For sensor time-series data, the 3σ principle is used to identify outliers, meaning that data values ​​exceeding the mean ± 3 times the standard deviation are considered outliers. Missing values ​​are filled using the median imputation method to ensure the continuity of the time-series data. For structured data such as orders and inventory, the multi-source production data acquisition module verifies data integrity through a field mapping table, deletes records with missing key fields such as order number, material code, or quantity, and corrects data format errors.

[0037] Specifically, the deduplication operation is achieved by building a data fingerprint database. The multi-source production data acquisition module performs SHA-256 hash calculation on each data to generate a unique data fingerprint. The fingerprint of the newly acquired data is compared with the existing fingerprints in the fingerprint database. If there is a duplicate, it is discarded directly to ensure data uniqueness.

[0038] Specifically, during the format standardization process, the multi-source production data acquisition module uniformly converts all text data into UTF-8 encoding format, sensor time series data into JSON format, and structured data into CSV format. At the same time, it standardizes the naming of data fields to ensure that large models can be efficiently parsed and processed.

[0039] Furthermore, during the data acquisition process, the multi-source production data acquisition module establishes a data quality verification mechanism to verify the completeness and accuracy of each piece of data collected. Completeness verification is achieved by checking for missing key fields, while accuracy verification is achieved through comparison with historical data and cross-validation of data across systems. For example, the multi-source production data acquisition module compares the order quantity in the ERP system with the production plan quantity in the MES system. If the deviation exceeds 5%, a data anomaly alarm is triggered, notifying relevant personnel to conduct an investigation.

[0040] Furthermore, to address high-concurrency data acquisition scenarios, the multi-source production data acquisition module adopts a distributed data acquisition architecture, distributing acquisition tasks to multiple acquisition nodes. Each node is responsible for acquiring data from a specific enterprise or of a specific type. Tasks are allocated among the acquisition nodes using a load balancing algorithm to ensure the stability and efficiency of the acquisition system. Simultaneously, the multi-source production data acquisition module establishes a data caching mechanism, caching frequently accessed basic data locally to reduce access pressure on the source system.

[0041] In summary, the multi-source production data acquisition module adopts a hybrid acquisition mode that combines timed retrieval with real-time push, which can balance the access pressure of the source system while ensuring the timeliness of data, and provide timely and comprehensive raw data support for subsequent data processing.

[0042] In summary, the multi-source production data acquisition module builds a cross-enterprise encrypted data exchange platform and establishes an identity authentication mechanism, which can effectively ensure the security and reliability of cross-enterprise data acquisition, solve the trust problem in the cross-enterprise data sharing process, ensure that data is not illegally obtained or tampered with during transmission and sharing, and provide a secure data interaction environment for supply chain collaboration.

[0043] In summary, the multi-source production data acquisition module employs differentiated cleaning and deduplication strategies for different data types, accurately addressing various data quality issues and improving data accuracy and consistency. The format normalization operation standardizes data from different sources and in different formats, eliminating data format barriers and enabling large models to quickly parse and process data, thus improving data processing efficiency.

[0044] In summary, the multi-source production data acquisition module establishes a data quality verification mechanism and a distributed acquisition architecture, which can cope with complex production data acquisition scenarios, ensure the quality of acquired data and the stability of the acquisition system, and efficiently complete data acquisition tasks even under high concurrency, providing a basic guarantee for the stable operation of the system.

[0045] Example 2, please refer to Figure 1 As shown: A production schedule collaborative operation intelligent management system based on a large model. The cross-enterprise semantic adaptation module is used to receive standardized production data, construct an industrial semantic graph based on the large model, perform semantic parsing and alignment on cross-enterprise heterogeneous system data, and output semantically aligned production data.

[0046] In this embodiment of the invention, the cross-enterprise semantic adaptation module, when performing semantic parsing and alignment operations, is specifically used for: Based on large-scale model training, we construct an industrial semantic graph that includes data terms, semantic connotations, and relationships. Extract textual descriptions and data tags of production-related data from different industrial systems and map them to corresponding nodes in the industrial semantic graph; The semantic association weights are generated by calculating the association paths and similarities of differentiated representations in the semantic graph using a large model. Based on semantic association weights, the reliability of data sources, and the contextual features of data representation, the semantic alignment confidence between cross-enterprise data representations is calculated. When the semantic alignment confidence reaches a preset threshold, the semantic alignment is confirmed to be effective and the semantically aligned production data is output.

[0047] Specifically, in the construction of the industrial semantic graph, the cross-enterprise semantic adaptation module selects a pre-trained large model based on the Transformer architecture as the base model and uses an industrial corpus covering multiple sub-sectors such as machinery manufacturing, electronic processing, automobile production, and aerospace for fine-tuning and training. The corpus contains text data such as technical manuals, process specifications, system operation documents, and industry standards, totaling more than 5 million entries. During training, the cross-enterprise semantic adaptation module optimizes the model's semantic understanding capabilities through masked language modeling tasks and sentence ranking tasks, enabling the model to accurately grasp the semantic connotations and relationships of industrial terms. The final industrial semantic graph contains more than 8,000 core term nodes and more than 12,000 semantic relationships. The relationship types between nodes are clearly defined by labels such as synonyms, subordinate relationships, associations, or causal relationships, forming a complete industrial semantic knowledge network.

[0048] Specifically, during the text description and data tag extraction process, the cross-enterprise semantic adaptation module uses the jieba word segmentation algorithm to segment the text description, and combines it with a conditional random field model to complete part-of-speech tagging, selecting core semantic keywords such as nouns, verbs, or adjectives. For example, after segmenting "engine cylinder block processing completion rate," the core keywords "engine cylinder block," "processing process," and "completion rate" are extracted. For data tags, the cross-enterprise semantic adaptation module extracts their attribute information, including data type and the business scenario to which they belong.

[0049] Specifically, the cross-enterprise semantic adaptation module calculates the cosine similarity between core semantic keywords and term nodes in the semantic graph, initially determining an initial mapping relationship with a similarity greater than 0.7. The initial mapping relationship is then corrected based on the attribute information of the data tags. For example, if the core keyword "material arrival time" has a similarity of 0.82 with the "material delivery date" node in the graph, but the data tag indicates that this data belongs to the "logistics and transportation" scenario, while the "material delivery date" node belongs to the "order delivery" scenario, then the mapping relationship is corrected to the "material delivery time" node under the "logistics and transportation" scenario, ensuring the accuracy of the mapping relationship.

[0050] Specifically, in the semantic association weight calculation process, the large model traverses the node paths corresponding to differentiated representations in the semantic graph, statistically analyzes features such as path length, number of associated nodes, and semantic association strength between nodes, and uses a weighted path algorithm to calculate the similarity between representations. The semantic association weight is represented by a value between 0 and 1, with a larger value indicating a stronger semantic association between representations.

[0051] Specifically, the reliability assessment of data sources is comprehensively evaluated through indicators such as enterprise credit rating, historical data interaction quality, and data verification pass rate. The cross-enterprise semantic adaptation module uses the analytic hierarchy process to determine the weight of each indicator and calculates a reliability score between 0 and 1.

[0052] Specifically, contextual feature analysis focuses on the sentence structure, semantic context, and business scenario description of the data representation. The cross-enterprise semantic adaptation module extracts scenario-related feature vectors and calculates the scenario matching degree through a neural network model.

[0053] Specifically, the semantic alignment confidence score is calculated using a weighted summation formula, as follows: Confidence score = Semantic association weight × 0.5 + Data source reliability × 0.3 + Context scene matching degree × 0.2.

[0054] The preset threshold is determined by statistically analyzing the accuracy of historical semantic alignment data. When the confidence level is greater than 0.85, the semantic alignment is deemed valid, and semantically aligned production data is output.

[0055] Furthermore, to enhance the dynamic adaptability of the semantic graph, the cross-enterprise semantic adaptation module establishes a semantic graph update mechanism. This mechanism regularly collects new industry terms, technical specifications, and new expressions arising during cross-enterprise data interactions. Through a combination of manual review and machine learning, the nodes and relationships in the semantic graph are updated. The update cycle can be dynamically adjusted based on the speed of industry technology updates and the frequency of enterprise data interactions, but is generally set to once a month.

[0056] Furthermore, an anomaly handling mechanism is established in the cross-enterprise semantic adaptation module during the semantic alignment process. When the semantic alignment confidence is lower than the preset threshold, the data is marked as semantically misaligned data, triggering a manual review process. Professional technicians then manually align the data, and the manual alignment results are fed back to the large model as training samples to continuously optimize the semantic alignment capabilities of the large model.

[0057] In summary, the cross-enterprise semantic adaptation module constructs an industrial semantic graph based on a large model fine-tuned from a large-scale industrial corpus. It can accurately capture the semantic connotations and relationships of industrial terms, providing a solid knowledge foundation for cross-enterprise data semantic alignment and solving the problem of insufficient semantic understanding accuracy in traditional methods.

[0058] In summary, the cross-enterprise semantic adaptation module improves the accuracy of mapping data to semantic graph nodes by combining core keyword matching with data tag attribute correction. This avoids mapping deviations caused by a single matching method and ensures that heterogeneous data across enterprises can be correctly associated with the corresponding nodes in the semantic graph.

[0059] In summary, the cross-enterprise semantic adaptation module calculates semantic alignment confidence by comprehensively considering semantic association weights, data source reliability, and contextual scenario matching. This allows for a comprehensive assessment of multiple factors in semantic alignment, improving the reliability of the results. The preset threshold setting provides a clear criterion for judging the effectiveness of semantic alignment, ensuring the quality of the output data.

[0060] In summary, the establishment of the semantic graph update mechanism and anomaly handling mechanism in the cross-enterprise semantic adaptation module enables the system to adapt to the dynamic changes in technological development and data interaction in the industrial field, continuously optimize semantic alignment capabilities, and effectively process unaligned data to avoid affecting subsequent collaborative management processes due to data semantic inconsistencies.

[0061] Example 3, please refer to Figure 1 and Figure 3 As shown: A production schedule collaborative operation intelligent management system based on a large model. The two-way collaboration module is used to receive semantically aligned production data, establish a semantic association model of production schedule and quality data, realize the linkage between quality anomaly tracing and schedule adjustment, and output schedule and quality linkage data.

[0062] In this embodiment of the invention, when the bidirectional collaboration module performs the establishment of a semantic association model and realizes the linkage between quality anomaly tracing and progress adjustment, it is specifically used for: By establishing semantic associations between production schedule nodes and corresponding quality inspection results, material qualification certificates and equipment operating parameters through a large model, a schedule and quality semantic association model is formed. When a quality anomaly is detected, the production progress node corresponding to the anomaly is located based on the progress and quality semantic association model, and the material batch information, operation execution records and equipment operating parameters associated with that node are traced along the semantic association link. Based on the traceability results, a schedule adjustment plan is generated, which includes suggestions for process adjustments and resource allocation plans, and outputs schedule and quality linkage data.

[0063] Specifically, in the process of building the semantic association model for progress and quality, the bidirectional collaboration module uses semantically aligned production data as training samples. The samples contain more than 100,000 association records of production progress nodes and quality-related data, covering typical production scenarios in multiple industries such as machinery manufacturing, electronics production, and automobile assembly.

[0064] The bidirectional collaboration module employs a graph neural network model to construct an association model, using production progress nodes as core nodes and quality inspection results, material qualification certificates, and equipment operating parameters as associated nodes. The association strength between nodes is represented by semantic association weights. During model training, the bidirectional collaboration module uses a cross-entropy loss function to optimize model parameters and employs 5-fold cross-validation to ensure the model's generalization ability. Ultimately, the model's association prediction accuracy reaches over 92%.

[0065] Specifically, the quality anomaly detection employs a dual detection mechanism combining real-time monitoring and periodic verification. During real-time monitoring, the two-way collaborative module collects quality detection data and compares it with preset quality standard thresholds. When the data exceeds the threshold range, it is immediately identified as a quality anomaly. During periodic verification, the two-way collaborative module uses statistical process control methods to analyze the fluctuation trend of quality data by calculating statistical quantities such as the mean, standard deviation, and range, and to identify potential quality anomaly risks. For example, when five consecutive quality data points show an upward or downward trend, a quality anomaly warning is triggered.

[0066] Specifically, during the anomaly tracing process, the two-way collaborative module quickly locates the production progress node corresponding to the anomaly based on the association model. For example, when the dimensional accuracy of a batch of products is found to be unqualified, the association model is used to locate the corresponding processing step node of the product, and traces it layer by layer along the semantic association link to obtain key data such as the material batch number, supplier information, operator number, operation time, equipment number, operating temperature, operating speed, and tool wear status associated with the node, forming a complete traceability chain.

[0067] Specifically, during the process of generating the schedule adjustment plan, the two-way collaboration module first analyzes the severity of quality anomalies based on the traceability results, classifying the anomalies into three levels: general anomalies, important anomalies, and severe anomalies. General anomalies refer to those that do not affect the core performance of the product and can be repaired through simple rectification, such as minor scratches on the product surface; important anomalies refer to those that affect some performance of the product and require adjustment of process parameters to repair, such as product dimensional deviations that are within the allowable range but close to the upper limit; severe anomalies refer to those that affect the core performance of the product, cannot be repaired through rectification, and require re-production, such as damage to the core components of the product.

[0068] Specifically, the two-way collaboration module assesses the schedule criticality of the process where the anomaly occurs, categorizing processes into critical, important, and general processes. Critical processes refer to core processes that affect the overall production schedule, such as the automotive engine assembly process, where delays will lead to overall delivery delays. Important processes refer to processes that affect local production flows, such as the parts cleaning process. General processes refer to processes that have a relatively small impact on the overall schedule, such as the product packaging process.

[0069] Specifically, the two-way collaboration module, combined with the flexibility of resource substitution, analyzes whether there are readily available alternative materials, backup equipment, or idle manpower. For example, when a production piece of equipment malfunctions and causes a quality defect, it assesses whether there is backup equipment that can be put into use immediately; when a batch of materials is substandard, it assesses whether there are qualified materials from other suppliers that can be quickly allocated.

[0070] Specifically, the two-way collaboration module uses the analytic hierarchy process (AHP) to quantify the impact of quality anomalies on production schedule and constructs an impact assessment index system, including three primary indicators: anomaly severity, process criticality, and resource substitution flexibility. Each primary indicator has 3 to 4 secondary indicators. The weight of each indicator is determined by pairwise comparison, and the final quantitative value of the impact is calculated.

[0071] Based on the quantified impact level, corresponding adjustment plans are generated: when the impact level is below 0.3, simple process parameter adjustment suggestions are generated; when the impact level is between 0.3 and 0.7, process adjustment suggestions and local resource allocation plans are generated; when the impact level is above 0.7, comprehensive process adjustment, material replacement, equipment scheduling, and manpower allocation plans are generated.

[0072] Furthermore, to ensure the feasibility of the schedule adjustment plan, the two-way collaboration module establishes a plan verification mechanism. The generated adjustment plan is input into the production simulation system to simulate its execution effect and evaluate its impact on production schedule, cost, and resource utilization. If the simulation results show that the plan cannot achieve the expected goals, it is returned for re-optimization; if the simulation passes, the final schedule adjustment plan is output.

[0073] Furthermore, the two-way collaboration module establishes an adjustment plan execution tracking mechanism to monitor the plan execution status in real time, collect data feedback during the execution process, and adjust the plan in a timely manner if deviations in plan execution or new quality anomalies are found, so as to ensure a dynamic balance between production progress and quality.

[0074] In summary, the bidirectional collaboration module constructs a semantic association model between progress and quality based on a graph neural network model. This model can effectively capture the complex relationship between production progress and quality data, providing accurate model support for quality anomaly tracing and progress adjustment linkage, and solving the problem of the disconnect between progress and quality control in traditional management.

[0075] In summary, the two-way collaborative module adopts a quality anomaly detection mechanism that combines real-time monitoring with periodic verification. This mechanism can quickly identify explicit quality anomalies and potential quality risks, ensuring that quality anomalies are detected and addressed early, and reducing their impact on production schedules.

[0076] In summary, the bidirectional collaboration module, based on the complete traceability chain of semantic association links, can accurately locate the root cause of quality anomalies, providing a reliable basis for formulating targeted schedule adjustment plans and avoiding resource waste and schedule delays caused by adjustments without a basis.

[0077] In summary, the two-way collaboration module, based on the analytic hierarchy process (AHP) for quantifying the degree of impact and generating hierarchical adjustment schemes, can formulate scientific and reasonable adjustment strategies according to the actual impact of quality anomalies, ensuring the relevance and feasibility of the adjustment schemes. At the same time, through scheme verification and execution tracking mechanisms, it ensures that the schemes can be effectively implemented, achieving dynamic collaboration between production progress and quality.

[0078] Example 4, please refer to Figure 1 As shown: A production schedule collaborative operation intelligent management system based on a large model. The scenario intent recognition and proactive collaboration module is used to receive semantically aligned production data, analyze the interaction behavior and real-time operation scenarios of collaborative roles, identify potential collaborative needs and push related semantically aligned production data and decision support information, and output collaborative need data.

[0079] In this embodiment of the invention, when the scene intent recognition and proactive collaboration module identifies potential collaboration needs and pushes related information, it is specifically used for: Collect historical query records, operation execution records, and real-time operation behaviors of collaborative roles to build collaborative role behavior profiles and scenario feature libraries; Based on the reasoning capabilities of large models, combined with the behavioral profiles of collaborative roles and real-time operational scenarios, we can identify the potential collaborative needs of collaborative roles. Based on the identified potential collaboration needs, relevant information is extracted from semantically aligned production data to generate decision support information, which is then pushed out through system pop-ups, mobile notifications, and industrial tablet prompts to output collaboration need data.

[0080] Specifically, during the data collection process of collaborative role behavior, the scene intent recognition and proactive collaboration module collects the collaborative role's historical query records, operation execution records, and real-time operation behaviors for the past 6 months. The historical query records include query keywords, query time, and query frequency. The operation execution records include operation modules, operation steps, operation duration, and operation results. The real-time operation behaviors include the current operation interface, input content, dwell time, and click behavior. During the data collection process, relevant regulations on data privacy protection are strictly followed, and sensitive information of the collaborative role is encrypted.

[0081] Specifically, the construction of collaborative role behavioral profiles includes three core dimensions: basic attributes, behavioral preferences, and demand characteristics. Basic attributes include the type of collaborative role, department, or scope of responsibility; behavioral preferences include frequently used query dimensions, high-frequency operation types, and preferred information display methods; demand characteristics include the types of historical collaborative demands, the urgency of demands, and the methods of fulfilling demands. For example, the behavioral profile of a production scheduler might show frequent queries about process progress, a preference for graphical data displays, and urgent demands primarily related to resource allocation.

[0082] Specifically, the scenario feature library is built based on the entire production collaboration process. The scenario intent recognition and proactive collaboration module is divided into five core scenarios: order receiving scenario, material preparation scenario, production execution scenario, quality inspection scenario, and logistics delivery scenario. Each scenario includes feature information such as scenario triggering conditions, typical operational behaviors, core data requirements, and common collaboration requirements. For example, the triggering condition for the order receiving scenario is that the collaborative role opens the order management module and views new order information. Typical operational behaviors include querying material inventory, confirming production capacity, and allocating production tasks. Core data requirements include material inventory status, equipment idle status, process capacity data, and supplier delivery capabilities. Common collaboration requirements include material procurement coordination, production resource scheduling, and delivery time negotiation.

[0083] Specifically, in the process of identifying potential collaborative needs, the scenario intent recognition and proactive collaboration modules use a large model finely tuned for industrial scenarios for reasoning and analysis. Real-time data on the operational behaviors of collaborative roles is collected to generate feature vectors for the current operational scenario of each role. These feature vectors contain feature values ​​across multiple dimensions, including operational modules, operational steps, dwell time, and input content.

[0084] By fusing feature vectors with behavioral profiles of collaborative roles and inputting them into a large model, the model's semantic reasoning capabilities predict the collaborative support that the role may need next. For example, when a production scheduler frequently queries material delivery information for a particular order, the model determines that their potential need is to confirm whether the materials have arrived on time and whether the production plan needs to be adjusted. When a quality inspector reviews the inspection data of a batch of products, the model determines that their potential need is to obtain relevant information such as the certificate of conformity for that batch of materials and the operating parameters of the production equipment. Validated using historical demand matching data, the accuracy rate for identifying these needs reaches over 88%.

[0085] Specifically, during the process of extracting related information, the scenario intent recognition and proactive collaboration module filters the most relevant information from semantically aligned production data based on the identified potential collaboration needs. For example, for material arrival confirmation needs, it extracts information such as the inventory quantity, transit status, estimated delivery time, and supplier contact information of related materials; for production plan adjustment needs, it extracts information such as the current progress of related processes, equipment load, manpower allocation, and alternative resource availability.

[0086] Specifically, the decision support information generation adopts natural language generation technology, and the scene intent recognition and proactive collaboration module transforms structured data into concise and clear text descriptions, while providing 2 to 3 optimization suggestions in combination with production management experience and algorithm models.

[0087] Specifically, the selection of push channels prioritizes preferences based on the behavioral profiles of collaborative roles. For urgent needs, push notifications are prioritized via system pop-ups and mobile notifications to ensure timely reception by collaborative roles; for routine needs, push notifications are sent via industrial tablets. Simultaneously, the scene intent recognition and proactive collaboration modules record information such as push time, collaborative role viewing time, viewing duration, and operation feedback, providing data support for subsequent demand identification and optimization.

[0088] Furthermore, the scene intent recognition and proactive collaboration module establishes a demand recognition optimization mechanism. It periodically analyzes feedback data from collaborative roles regarding pushed information, including whether they viewed the message, adopted suggestions, or supplemented their queries, calculating the accuracy of demand recognition and the usefulness of the pushed information. Based on the analysis results, the inference parameters of the large model are adjusted, and the behavioral profiles of collaborative roles and the scene feature library are optimized to continuously improve the accuracy of demand recognition and the relevance of pushed information.

[0089] Furthermore, the scene intent recognition and proactive collaboration module supports customized push rules for collaborative roles. Collaborative roles can set push channels, push frequency, and information display methods according to their own work habits, thereby improving the user experience. For example, collaborative roles can be set to receive non-urgent push information only during working hours and display frequently viewed data in the form of a dashboard.

[0090] In summary, the scene intent recognition and proactive collaboration module constructs multi-dimensional behavioral profiles of collaborative roles and scene feature libraries, which can comprehensively depict the behavioral habits of collaborative roles and the characteristic needs of different scenarios, providing rich evidence for the identification of potential collaborative needs and solving the problem of insufficient understanding of collaborative role needs in traditional collaborative management.

[0091] In summary, the scenario intent recognition and proactive collaboration module identifies potential collaborative needs based on the semantic reasoning capabilities of the fine-tuned large model. It can accurately capture implicit needs that are not explicitly expressed by collaborative roles, realizing the transformation of collaborative need recognition from passive response to proactive prediction, and improving the foresight and proactiveness of cross-enterprise production collaborative management.

[0092] In summary, the scene intent recognition and proactive collaboration module accurately extracts relevant information and generates personalized decision support information, providing direct and effective decision-making references for collaborative roles. This helps them quickly resolve collaborative issues and improve work efficiency. The multi-channel push mechanism and custom rule settings enhance the timeliness of information delivery and the user experience for collaborative roles.

[0093] In summary, the establishment of the scene intent recognition and proactive collaboration module demand recognition optimization mechanism can continuously optimize system performance based on feedback from collaborative roles, ensuring that demand recognition and information push always meet the actual needs of collaborative roles, thereby improving the long-term applicability of the system and the satisfaction of collaborative roles.

[0094] Example 5, please refer to Figure 1 As shown: A production schedule collaborative operation intelligent management system based on a large model. The large model decision engine module provides semantic reasoning support for the cross-enterprise semantic adaptation module, the two-way collaboration module, and the scenario intent recognition and proactive collaboration module. At the same time, it integrates standardized production data, semantically aligned production data, progress and quality linkage data, and collaboration requirement data to construct a multi-objective decision model. Based on the system operation feedback, the large model is optimized to output a production scheduling plan.

[0095] In this embodiment of the invention, the large model decision engine module, when providing semantic reasoning support, is specifically used for: By calling pre-trained large-scale models in the industrial field, inference support is provided for the semantic parsing process of the cross-enterprise semantic adaptation module, and the accuracy of semantic mapping is verified. Provide reasoning verification for the progress and quality correlation logic of the two-way collaborative module to ensure the correctness of the anomaly tracing path; It provides semantic reasoning support for the requirement identification process of the scene intent recognition and proactive collaboration module.

[0096] In this embodiment of the invention, the large model decision engine module, when constructing a multi-objective decision model and optimizing the large model, is specifically used for: It aggregates semantically aligned production data from the cross-enterprise semantic adaptation module, progress and quality linkage data from the bidirectional collaboration module, and collaboration requirement data from the scene intent recognition and proactive collaboration module. A multi-objective decision-making model is constructed by taking production schedule coordination, resource allocation optimization, and quality risk control as decision-making objectives. Generate and output a production scheduling plan based on a multi-objective decision-making model; Based on system operation feedback data, including the execution effect of production scheduling schemes, feedback from collaborative roles on collaborative push information, and the accuracy of semantic alignment results, the large model is dynamically optimized.

[0097] Specifically, the large industrial model invoked during the semantic reasoning support process is a large model based on the Transformer architecture that has been fine-tuned from 5 million industrial corpora. This model has been specifically optimized for production collaboration scenarios and has strong domain semantic understanding and reasoning capabilities.

[0098] For the cross-enterprise semantic adaptation module, the model verifies the rationality of the mapping relationship by comparing the data representation before and after semantic mapping with the business scenario. When the probability of the mapping result being unreasonable exceeds 10%, a remapping process is triggered. For the bidirectional collaboration module, the model verifies the completeness and logic of the anomaly tracing path through reverse reasoning, checks for any missing related nodes, and ensures that the tracing results are comprehensive and accurate. For the scenario intent recognition and proactive collaboration module, the model optimizes the accuracy of requirement recognition and reduces misjudgments and omissions by analyzing the causal relationship between the behavior of collaborative roles and potential requirements.

[0099] Specifically, in the process of constructing the multi-objective decision-making model, the large-scale model decision engine module first standardizes various input data, converting data of different magnitudes and dimensions into unified evaluation indicators. Evaluation indicators for production schedule coordination objectives include order delivery on-time rate, process seamlessness, and production cycle compliance rate; evaluation indicators for resource allocation optimization objectives include resource utilization rate, scheduling cost, and energy consumption level; and evaluation indicators for quality risk control objectives include quality anomaly occurrence rate, anomaly rectification completion rate, and product qualification rate.

[0100] Specifically, the large-scale model decision engine module uses a weighted summation method to construct a multi-objective decision model and determines the weights of each objective through the analytic hierarchy process (AHP). An evaluation team composed of experts in production management, quality management, and resource scheduling was invited to conduct pairwise comparisons of the importance of the three decision objectives, construct a judgment matrix, calculate the weight vector, and perform consistency checks. The final determined weights are: production schedule coordination weight 0.4, resource allocation optimization weight 0.3, and quality risk control weight 0.3.

[0101] Specifically, the model training employs the gradient descent algorithm to optimize parameters. The training samples include over 50,000 historical production scheduling data points, covering typical cases from different industries and production scales. During training, the large model decision engine module divides the historical data into a training set (70%), a validation set (15%), and a test set (15%). The model's hyperparameters are adjusted using the validation set, and the model's performance is evaluated using the test set. After training, the model's scheduling scheme achieves a satisfaction rate of over 90%.

[0102] Specifically, during the production scheduling plan generation process, the model performs multi-scheme simulation calculations on production process arrangement, resource allocation, and schedule adjustment based on real-time production data, collaborative needs, and decision target weights, selecting the scheme with the optimal overall benefits as the final output. The scheduling plan includes process adjustment plans, resource allocation lists, schedule node planning, and key quality control points.

[0103] Specifically, the large model optimization process adopts an online learning mode. The large model decision engine module is set to collect feedback data every 24 hours. Each cycle collects system operation feedback data, including the execution effect of the production scheduling plan, the feedback of collaborative roles to collaborative push information, and the accuracy of semantic alignment results.

[0104] Specifically, the large model decision engine module calculates the performance indicators of each stage of the model based on feedback data. When the performance indicator of a certain stage falls below a preset threshold, local fine-tuning of the model is triggered. For example, when the semantic alignment accuracy is below 85%, the model part corresponding to the cross-enterprise semantic adaptation module is fine-tuned using newly added semantic alignment samples; when the demand recognition accuracy is below 80%, the model part corresponding to the scene intent recognition is optimized by supplementing collaborative role behavior data; when the satisfaction of the scheduling scheme is below 85%, the weight coefficients of the multi-objective decision model are adjusted or the algorithm parameters are optimized.

[0105] Furthermore, to ensure the stability and reliability of the decision-making model, the large-scale model decision engine module establishes a model performance monitoring mechanism to monitor performance indicators such as inference speed, accuracy, and robustness in real time. When significant fluctuations occur in model performance, a model rollback mechanism is automatically triggered to restore the model to a previous stable version, and the causes of the fluctuations are analyzed for targeted optimization.

[0106] Furthermore, the large model decision engine module supports manual intervention in the decision-making process. When encountering complex production scenarios or special collaborative needs, production managers can manually adjust the parameters of the decision model or directly modify the scheduling plan. The system will use the results of manual adjustments as training samples to feed back to the model, continuously improving the model's ability to handle complex scenarios.

[0107] In summary, the large model decision engine module calls upon a large model specifically optimized for industrial scenarios to provide semantic reasoning support. This enables precise reasoning support for the core processes of each module, ensuring the accuracy of semantic alignment, anomaly tracing, and demand identification, thereby improving the overall system performance.

[0108] In summary, the large-scale model decision engine module constructs a multi-objective decision model that takes into account production schedule, resource allocation, and quality control. Through scientific weight allocation and training with a large amount of historical data, it can generate a production scheduling scheme with optimal comprehensive benefits, overcoming the limitations of traditional single-objective decision-making and meeting the diverse needs of complex production collaboration scenarios.

[0109] In summary, the large model decision engine module adopts an online learning mode and a feedback-driven model optimization mechanism, which enables the model to continuously adapt to the dynamic changes in the production scenario, continuously optimize performance based on actual operating results, and ensure that the model maintains good adaptability and decision accuracy in the long term.

[0110] In summary, the model performance monitoring mechanism and manual intervention function of the large model decision engine module can ensure the stability and reliability of the decision model, while providing diversified data support for model optimization and improving the system's ability to cope with complex scenarios and special needs.

[0111] It will be apparent to those skilled in the art that the present invention is not limited to the details of the exemplary embodiments described above, and that the present invention can be implemented in other specific forms without departing from the spirit or essential characteristics of the present invention.

[0112] The embodiments of this application can acquire and process relevant data based on artificial intelligence technology. Artificial intelligence is the theory, method, technology, and application system that uses digital computers or machines controlled by digital computers to simulate, extend, and expand human intelligence, perceive the environment, acquire knowledge, and use that knowledge to obtain optimal results.

[0113] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention.

Claims

1. An intelligent management system for collaborative production schedule operations based on a large model, characterized in that, The system includes a multi-source production data acquisition module, a cross-enterprise semantic adaptation module, a two-way collaboration module, a scene intent recognition and proactive collaboration module, and a large model decision engine module. Each module is linked based on data flow and decision flow. The multi-source production data acquisition module is used to collect multi-source production data, preprocess the collected data, and output standardized production data. The cross-enterprise semantic adaptation module is used to receive the standardized production data, construct an industrial semantic graph based on a large model, perform semantic parsing and alignment on cross-enterprise heterogeneous system data, and output semantically aligned production data. The bidirectional collaboration module is used to receive the semantically aligned production data, establish a semantic association model of production progress and quality data, realize the linkage between quality anomaly tracing and progress adjustment, and output progress and quality linkage data. The scenario intent recognition and proactive collaboration module is used to receive the semantically aligned production data, analyze the interaction behavior and real-time operation scenario of the collaborative roles, identify potential collaborative needs and push the associated semantically aligned production data and decision support information, and output collaborative need data. The large model decision engine module is used to provide semantic reasoning support for the cross-enterprise semantic adaptation module, bidirectional collaboration module, and scene intent recognition and proactive collaboration module. At the same time, it integrates the standardized production data, semantically aligned production data, progress and quality linkage data, and collaboration requirement data to construct a multi-objective decision model. Based on system operation feedback, it optimizes the large model and outputs a production scheduling plan.

2. The intelligent management system for collaborative production schedule operations based on a large model according to claim 1, characterized in that, The multi-source production data acquisition module is specifically used for: Collect order information, bill of materials, and process progress data from the enterprise's internal ERP, MES, and QMS systems; Collect material flow data, order delivery data, and logistics transportation data from upstream and downstream enterprises in the industrial chain; The collected text data such as process documents, sensor time-series data, and structured data such as orders and inventory are cleaned, deduplicated, and formatted. Data of different formats are transformed into a unified data format that can be processed by large models, forming and outputting standardized production data.

3. The intelligent management system for collaborative production schedule operations based on a large model according to claim 1, characterized in that, The cross-enterprise semantic adaptation module is specifically used for: Based on large-scale model training, we construct an industrial semantic graph that includes data terms, semantic connotations, and relationships. Extract textual descriptions and data tags of production-related data from different industrial systems and map them to the corresponding nodes of the industrial semantic graph; The large model is used to calculate the association paths and similarities of differentiated representations in the semantic graph, and semantic association weights are generated. Based on the semantic association weights, the reliability of the data source, and the contextual features of the data representation, the semantic alignment confidence between cross-enterprise data representations is calculated. When the semantic alignment confidence level reaches a preset threshold, the semantic alignment is confirmed to be effective and the semantically aligned production data is output.

4. The intelligent management system for collaborative production schedule operations based on a large model according to claim 1, characterized in that, The bidirectional collaboration module is specifically used for: By establishing semantic associations between production schedule nodes and corresponding quality inspection results, material qualification certificates and equipment operating parameters through a large model, a schedule and quality semantic association model is formed. When a quality anomaly is detected, the production progress node corresponding to the anomaly is located based on the progress and quality semantic association model, and the material batch information, operation execution records and equipment operating parameters associated with the node are traced along the semantic association link. Based on the traceability results, a schedule adjustment plan is generated, which includes suggestions for process adjustments and resource allocation plans, and outputs schedule and quality linkage data.

5. The intelligent management system for collaborative production schedule operations based on a large model according to claim 1, characterized in that, The scene intent recognition and proactive collaboration module is specifically used for: Collect historical query records, operation execution records, and real-time operation behaviors of collaborative roles to build user behavior profiles and scenario feature libraries; Based on the reasoning capabilities of large models, combined with user behavior profiles and real-time operation scenarios, potential collaborative needs of collaborative roles are identified. Based on the identified potential collaboration needs, relevant information is extracted from semantically aligned production data to generate decision support information, which is then pushed out through system pop-ups, mobile notifications, and industrial tablet prompts to output collaboration need data.

6. The intelligent management system for collaborative production schedule operations based on a large model according to claim 1, characterized in that, The large model decision engine module is specifically used to provide semantic reasoning support for: By calling pre-trained large-scale models in the industrial field, inference support is provided for the semantic parsing process of the cross-enterprise semantic adaptation module, and the accuracy of semantic mapping is verified. Provide reasoning verification for the progress and quality correlation logic of the two-way collaborative module to ensure the correctness of the anomaly tracing path; It provides semantic reasoning support for the requirement identification process of the scene intent recognition and proactive collaboration module.

7. The intelligent management system for collaborative production schedule operations based on a large model according to claim 1, characterized in that, The large model decision engine module is specifically used in constructing multi-objective decision models for: It aggregates semantically aligned production data from the cross-enterprise semantic adaptation module, progress and quality linkage data from the bidirectional collaboration module, and collaboration requirement data from the scene intent recognition and proactive collaboration module. A multi-objective decision-making model is constructed by taking production schedule coordination, resource allocation optimization, and quality risk control as decision-making objectives. A production scheduling scheme is generated and output based on the multi-objective decision model.

8. The intelligent management system for collaborative production schedule operations based on a large model according to claim 3, characterized in that, The cross-enterprise semantic adaptation module is specifically used to extract textual descriptions and data tags of production-related data from different industrial systems and map them to corresponding nodes in the industrial semantic graph. The text is segmented and tagged with parts of speech to extract core semantic keywords; The core semantic keywords are matched with the term nodes in the semantic graph to determine the initial mapping relationship; Based on the attribute information of the data tags, including data type and business scenario, the initial mapping relationship is corrected to obtain the final mapping node.

9. The intelligent management system for collaborative production schedule operations based on a large model according to claim 4, characterized in that, When the bidirectional collaboration module generates a schedule adjustment plan that includes process adjustment suggestions and resource allocation plans based on the traceability results, it is specifically used for: Based on the traceability results, and considering the severity of the quality anomaly, the schedule criticality of the process, and the flexibility of resource substitution, the impact of the quality anomaly on the production schedule is quantified. Based on the magnitude of the impact, corresponding level of process adjustment suggestions and resource allocation plans are generated.

10. The intelligent management system for collaborative production schedule operations based on a large model according to claim 5, characterized in that, The scene intent recognition and proactive collaboration module is specifically used for: extracting related information and generating decision support information in the following ways: Determine push channels based on user behavior profiles; Based on the scene feature library, semantically aligned production data that is highly relevant to the current scene is matched, including the inventory status of related materials and the equipment readiness status of the next process. The extracted correlation information and the generated decision support information are packaged together, pushed out through the determined push channels, and the push time and user viewing status are recorded.

Images