Intelligent planning and production scheduling method and system based on large model, product and medium

Abstract

The invention discloses an intelligent planning production scheduling method and system based on a large model, a product and a medium, and relates to the field of production scheduling, and the method comprises the steps: inputting a text description record into a preset production large model for semantic analysis, extracting decision elements of a production expert in different scenes and a weight relation thereof, and obtaining a production scheduling knowledge base; inputting the scene description text into a preset large production model, and extracting core decision logic of experts in the associated scene; the method comprises the following steps: performing predictive analysis on real-time data of a production field based on a preset production large model to obtain a production state evolution trend in a preset time period, identifying productivity conflict points from the production state evolution trend, generating a plurality of alternative adjustment schemes based on the productivity conflict points, process constraint conditions and expert core decision logic, and inputting the plurality of alternative adjustment schemes into a preset target evaluation function for calculation, and selecting the alternative adjustment scheme with the highest comprehensive score as a planned production scheduling scheme. By implementing the method, the enterprise production plan adjustment efficiency can be improved.

Description

TECHNICAL FIELD

[0001] The application relates to the field of production scheduling, and in particular to an intelligent production scheduling method and system based on a large model, a product and a medium. BACKGROUND

[0002] With the development of manufacturing industry towards intelligence and flexibility, the demand for dynamic adjustment of production plans is increasing. In the production mode of multi-variety and small batch, various uncertain factors such as equipment failure, material shortage and urgent orders frequently occur, which makes it difficult to execute the original production plan on schedule, and the plan needs to be adjusted in time to ensure the continuity of production and the timely delivery of orders.

[0003] In the prior art, a rule-based plan adjustment method is usually used, a series of fixed adjustment rules are preset, such as order priority rules and equipment selection rules. When an abnormality occurs in the production process, the system automatically adjusts the production plan according to these preset rules, for example, moving the affected orders to the back, transferring the processing tasks to the standby equipment, etc.

[0004] However, due to the complex and changeable actual situation in the production site, the preset fixed rules are often too simple and mechanical, and cannot fully consider various constraint conditions and influencing factors in actual production. With the expansion of production scale and the increase of product types, the combination of various abnormal situations becomes more complex, and relying on preset rules leads to low efficiency of production plan adjustment of enterprises. SUMMARY

[0005] The application provides an intelligent production scheduling method and system based on a large model, a product and a medium, which is used to improve the efficiency of production plan adjustment of enterprises.

[0006] In a first aspect, the application provides an intelligent plan scheduling method based on a large model, applied to an intelligent plan scheduling system. The method comprises: obtaining a textual description record of a production expert processing production plan adjustment, the textual description record containing a triggering scenario, a consideration factor, an adjustment idea and an adjustment result; inputting the textual description record into a preset production large model for semantic analysis, extracting decision elements and their weight relationships of the production expert in different scenarios to obtain a production scheduling knowledge base; collecting production site real-time data and standardizing the production site real-time data into a scene description text, the production site real-time data including equipment operating status, material inventory level and in-process product processing progress; inputting the scene description text into the preset production large model, analyzing the correlation degree of the current scene and the historical decision-making scene based on the production scheduling knowledge base, and extracting the core decision-making logic of the expert in the associated scene; based on the preset production large model, performing prediction analysis on the production site real-time data to obtain a production state evolution trend in a preset time period, and identifying a capacity conflict point from the production state evolution trend, the capacity conflict point including a conflict time interval and a production order involved; generating a plurality of alternative adjustment schemes based on the capacity conflict point, process constraint conditions and the core decision-making logic of the expert, inputting the plurality of alternative adjustment schemes into a preset target evaluation function for calculation, and selecting the alternative adjustment scheme with the highest comprehensive score as the plan scheduling scheme.

[0007] In the above embodiment, the large model performs semantic analysis on the production expert's historical decisions to construct a production scheduling knowledge base containing decision elements and weight relationships. Based on real-time data analysis, a capacity conflict point is identified, and the core decision-making logic highly associated with the current scene is extracted from the knowledge base. Through the evaluation function, the intelligent evaluation and screening of multiple alternative schemes are realized, which realizes the accurate dynamic adjustment of production plan and improves the response speed and scheduling accuracy of plan scheduling.

[0008] In some embodiments of the first aspect, in some embodiments, the production site real-time data is analyzed by the preset production large model to obtain a production state evolution trend in a preset time period, and a capacity conflict point is identified from the production state evolution trend, the capacity conflict point including a conflict time interval and a step involving a production order, specifically including: determining a standard capacity of each device in the preset time period based on the device operating state to obtain a device capacity benchmark value; calculating a device capacity demand of the order to be processed in the preset time period according to the product processing progress and the process route to obtain a device load prediction value; dividing the preset time period into a plurality of time units according to the minimum production unit time, and calculating a device load rate in each time unit, and marking the time unit as a load overrun unit when the device load prediction value in the first time unit is greater than the device capacity benchmark value; calculating a material demand replenishment balance in each time unit, and marking the time unit as a material shortage unit when the material demand in the second time unit is greater than the material inventory level; determining the union of the load overrun unit and the material shortage unit as the conflict time interval, and determining the order to be processed in the conflict time interval as the production order involved.

[0009] In the above embodiment, the preset time period is divided into the minimum production unit, the device capacity benchmark value and the load prediction value are calculated respectively, and the load overrun unit is identified through the load rate calculation. At the same time, the material shortage unit is determined by combining the material demand replenishment balance analysis, and the union of the two types of abnormal units is taken as the conflict time interval. This detailed conflict identification method accurately locates the capacity and material bottlenecks, and provides a basis for formulating targeted adjustment strategies subsequently.

[0010] In some embodiments of the first aspect, in some embodiments, a plurality of alternative adjustment schemes are generated based on the capacity conflict point, the process constraint condition and the expert core decision logic, the plurality of alternative adjustment schemes are input into a preset target evaluation function for calculation, and the scheme with the highest comprehensive score is selected as the plan scheduling scheme, specifically including: generating an order adjustment strategy set according to the production order involved in the conflict time interval and the process constraint condition, the order adjustment strategy set including an order splitting strategy, an order transfer strategy and an order postponement strategy for the production order involved; screening the order adjustment strategy set according to the expert core decision logic, selecting the adjustment strategy conforming to the historical decision mode, and combining the adjustment strategies conforming to the historical decision mode to generate a plurality of alternative adjustment schemes; inputting the plurality of alternative adjustment schemes into the preset target evaluation function to obtain the score results of the plurality of alternative adjustment schemes, and selecting the alternative adjustment scheme with the highest comprehensive score as the plan scheduling scheme based on the score results.

[0011] In the above embodiment, according to the order generation order adjustment strategy set in the conflict time interval, including splitting, transfer and postponement and other strategies. Based on the expert core decision logic, the adjustment strategy that meets the historical decision mode is selected and combined to generate the alternative scheme, and the comprehensive score is calculated by the preset target evaluation function. This multi-dimensional scheme generation and evaluation mechanism ensures the feasibility and optimality of the adjustment scheme.

[0012] In combination with some embodiments of the first aspect, in some embodiments, after the step of generating a plurality of alternative adjustment schemes based on the capacity conflict point, the process constraint condition and the expert core decision logic, inputting the plurality of alternative adjustment schemes into the preset target evaluation function for calculation, and selecting the scheme with the highest comprehensive score as the plan scheduling scheme, the method further comprises: performing production verification on the plurality of alternative adjustment schemes, analyzing the execution effect of each scheme under the conditions of equipment fluctuation and material fluctuation, obtaining a scheme verification result; selecting an optimal adjustment scheme according to the scheme verification result; recording the relevant information of the capacity conflict point, the expert core decision logic, the plurality of alternative adjustment schemes and the optimal adjustment scheme; and storing the relevant information into a production scheduling knowledge base.

[0013] In the above embodiment, the production verification is performed on the alternative adjustment scheme, and the execution effect under the conditions of equipment fluctuation and material fluctuation is analyzed, so as to select the optimal adjustment scheme. The verification result is recorded together with the information of the capacity conflict point and the decision logic, and is stored into the production scheduling knowledge base, forming a closed-loop feedback mechanism. Based on the continuous optimization of actual verification, the scheduling knowledge base is continuously enriched and improved, and the robustness and adaptability of the plan scheduling scheme are improved.

[0014] In combination with some embodiments of the first aspect, in some embodiments, before the step of generating a plurality of alternative adjustment schemes based on the capacity conflict point, the process constraint condition and the expert core decision logic, inputting the plurality of alternative adjustment schemes into the preset target evaluation function for calculation, and selecting the scheme with the highest comprehensive score as the plan scheduling scheme, the method further comprises: extracting the expert core decision logic from the production scheduling knowledge base; constructing a target evaluation function based on the expert core decision logic, the target evaluation function including a production efficiency index, a delivery date satisfaction index and a changeover loss index; and setting the target evaluation function as the preset target evaluation function.

[0015] In the above embodiment, the expert core decision logic is extracted from the production scheduling knowledge base, and the target evaluation function including the production efficiency, the delivery date satisfaction and the changeover loss is constructed. The evaluation function quantifies the expert experience into specific index weights, realizing the multi-dimensional comprehensive evaluation of the alternative scheme. The evaluation system constructed based on the expert decision logic enhances the scientificity and accuracy of the scheme evaluation.

[0016] In some embodiments of the first aspect, after the step of generating a plurality of alternative adjustment schemes based on the capacity conflict point, the process constraint, and the expert core decision logic, inputting the plurality of alternative adjustment schemes into a preset target evaluation function for calculation, and selecting the scheme with the highest comprehensive score as the planned production scheduling scheme, the method further comprises: collecting device running status, personnel operation records, and material consumption data of each process of the production line; calculating the standard capacity and the actual capacity of each process of the production line to obtain the capacity utilization rate of the production line; identifying the bottleneck process in the production line according to the capacity utilization rate, and determining the key resources corresponding to the bottleneck process; formulating a device configuration adjustment scheme, a personnel configuration adjustment scheme, and a material configuration adjustment scheme based on the usage status of the key resources; and combining the device configuration adjustment scheme, the personnel configuration adjustment scheme, and the material configuration adjustment scheme to form a resource configuration scheme. In the above embodiments, the capacity utilization rate is calculated based on the device running status, the personnel operation records, and the material consumption data, the bottleneck process in the production line and the key resources are identified. The configuration adjustment schemes of the device, the personnel, and the material are formulated for the key resources and combined to form the resource configuration scheme. The multi-dimensional resource collaborative configuration improves the overall capacity utilization level of the production line and eliminates the local bottleneck constraints.

[0017] In some embodiments of the first aspect, after the step of generating a plurality of alternative adjustment schemes based on the capacity conflict point, the process constraint, and the expert core decision logic, inputting the plurality of alternative adjustment schemes into a preset target evaluation function for calculation, and selecting the scheme with the highest comprehensive score as the planned production scheduling scheme, the method further comprises: setting a supplier evaluation index, the supplier evaluation index comprising a production capacity index, a delivery capacity index, and a quality level index; obtaining historical supply records of the supplier, evaluating the supplier according to the supplier evaluation index, and generating a supplier evaluation result; selecting a target supplier according to the material demand of the production plan and the supplier evaluation result; sending a purchase demand to the target supplier, and obtaining supply time, supply quantity, and supply quality information of the target supplier; setting a material supply early warning rule, including a material inventory early warning value and a supply delay early warning value; when the material inventory is lower than the material inventory early warning value or the supply time exceeds the supply delay early warning value, starting emergency procurement or adjusting the production sequence according to the urgency of the production plan.

[0018] In the above embodiments, a supplier evaluation system including production capacity, delivery capacity, and quality level is established, and a target supplier is selected based on historical supply records. A material inventory and supply delay early warning rule is set, and emergency procurement or production sequence adjustment is dynamically started according to the urgency of the production plan. The collaborative operation of the supplier evaluation and the material early warning mechanism realizes the deep integration of the material supply chain and the production plan, and guarantees the continuous and stable execution of the production plan.

[0019] In a second aspect, the embodiments of the present application provide an intelligent planned production scheduling system, comprising: one or more processors and a memory; the memory is coupled to the one or more processors, and the memory is configured to store computer program code, the computer program code comprising computer instructions, the one or more processors invoking the computer instructions to cause the intelligent planned production scheduling system to perform the method described in the first aspect and any possible implementation manner of the first aspect.

[0020] In a third aspect, the embodiments of the present application provide a computer program product comprising instructions, which, when executed on an intelligent planned production scheduling system, cause the intelligent planned production scheduling system to perform the method described in the first aspect and any possible implementation manner of the first aspect.

[0021] In a fourth aspect, the embodiments of the present application provide a computer-readable storage medium comprising instructions, which, when executed on an intelligent planned production scheduling system, cause the intelligent planned production scheduling system to perform the method described in the first aspect and any possible implementation manner of the first aspect.

[0022] It can be understood that the intelligent planned production scheduling system provided by the second aspect, the computer program product provided by the third aspect, and the computer storage medium provided by the fourth aspect are all used to execute the method provided by the embodiments of the present application. Therefore, the beneficial effects that can be achieved are referred to the beneficial effects in the corresponding method, which will not be described here.

[0023] The one or more technical solutions provided in the embodiments of the present application have at least the following technical effects or advantages: 1. In the present application, a large model is used to perform semantic analysis on historical decision-making of production experts, and a production scheduling knowledge base containing decision-making elements and weight relationships is constructed. Based on real-time data analysis, conflict points of production capacity are identified, and core decision-making logic highly related to the current scene is extracted from the knowledge base. Through an evaluation function, multiple alternative schemes are intelligently evaluated and screened, realizing accurate dynamic adjustment of production planning and improving the response speed and scheduling accuracy of planned production.

[0024] 2. In the present application, a preset time period is divided into the smallest production unit, the device capacity benchmark value and the load prediction value are calculated respectively, and the load overrun unit is identified through load rate calculation. At the same time, combined with material demand replenishment balance analysis, the material shortage unit is determined, and the union of the two types of abnormal units is taken as the conflict time interval. This fine conflict identification method accurately locates the production capacity and material bottlenecks, providing a basis for formulating targeted adjustment strategies subsequently.

[0025] 3、The application generates an order adjustment strategy set including splitting, transferring and postponing, etc. based on the conflict time interval involving order generation. The adjustment strategies that meet the historical decision mode are selected based on the expert core decision logic and combined to generate alternative solutions. The preset target evaluation function is used for comprehensive scoring. This multi-dimensional solution generation and evaluation mechanism guarantees the feasibility and optimality of the adjustment scheme. BRIEF DESCRIPTION OF DRAWINGS

[0026] Figure 1 is a flowchart of an intelligent plan scheduling method based on a large model in an embodiment of the application; Figure 2 is another flowchart of an intelligent plan scheduling method based on a large model in an embodiment of the application; Figure 3 is an entity device structure schematic diagram of an intelligent plan scheduling system in an embodiment of the application. DETAILED DESCRIPTION

[0027] The terms used in the following embodiments of the application are only for the purpose of describing specific embodiments and are not intended to be limiting on the application. As used in the specification, the singular forms "a," "an," and "the" are intended to include the plural forms as well, unless the context clearly indicates otherwise. It will be further understood that the terms "and / or" used in the application means any or all possible combinations of one or more of the listed items.

[0028] Hereinafter, the terms "first" and "second" are only for the purpose of description, and cannot be understood as implying or suggesting relative importance or implicitly indicating the number of indicated technical features. Therefore, the features defined with "first" and "second" can explicitly or implicitly include one or more of the features, and in the description of the embodiments of the application, the meaning of "multiple" is two or more, unless otherwise specified.

[0029] For ease of understanding, the application scenarios of the embodiments of the application are introduced as follows.

[0030] In a large automotive parts manufacturing enterprise, production lines are simultaneously processing multiple types of transmission housings. Due to high customization, small batch size, and tight delivery schedule, production plans often need to be dynamically adjusted. For example, a key machining center was found to have a spindle failure that needs to be repaired for 4 hours, while the supplier notified a delay in the arrival of a batch of aluminum alloy blanks, plus a new urgent order was temporarily inserted. These unexpected situations cause the original processing sequence and equipment allocation plan for 20 orders to be unable to be executed on schedule. The workshop supervisor needs to quickly adjust the production plan considering factors such as equipment repair time, material arrival time, process route restrictions, and changeover loss, etc., to minimize the impact on overall delivery. This complex plan adjustment decision often relies on experienced production supervisors, but manual decision-making is inefficient and prone to oversight.

[0031] A certain home appliance manufacturing enterprise uses a fixed rule-based scheduling system to handle production plan adjustments. The system pre-sets some basic scheduling rules, such as "when a device fails, the affected order sequence is moved back", "when there is a shortage of materials, orders with sufficient inventory are prioritized", "when a new urgent order is inserted, the original planned order is delayed", etc. When the injection molding process equipment upstream of a refrigerator assembly line fails, the system automatically postpones the original planned production orders for the next 4 hours, and searches for available backup equipment. However, this simple rule cannot handle more complex situations. For example, when multiple abnormalities such as mold replacement delay, injection molding part quality anomaly, and urgent order insertion occur simultaneously, the system cannot weigh various constraint conditions and influencing factors to make the optimal adjustment decision. Moreover, the pre-set rules are too rigid and cannot be flexibly adjusted according to actual production conditions, resulting in poor plan execution results.

[0032] In a precision machinery manufacturing workshop, an intelligent scheduling system based on large models is applied. The system first accumulates a large number of production expert processing plan adjustment cases, including analysis ideas and specific adjustment schemes for various abnormal scenarios. One morning, the system monitors that three abnormal situations occur simultaneously on an automated production line: a precision machining center needs to replace the tool assembly, which is expected to take 2 hours, a batch of key parts is found to have a 3% rejection rate and needs to be reworked, and a high-priority order requires early delivery. The system immediately analyzes the current situation, extracts the decision elements and processing ideas of experts in similar scenarios from the knowledge base. By analyzing the constraints of device capacity, material supply, process route, etc., the system automatically generates multiple feasible adjustment schemes. These schemes include strategies such as transferring some orders to backup equipment, adjusting the size of the processing batch, and optimizing the changeover sequence. The system quantitatively evaluates the production efficiency, delivery impact, and changeover loss of each scheme, and finally selects an optimal scheme that balances various indicators. The entire decision-making process takes only a few minutes, ensuring the feasibility of the scheme and significantly improving the efficiency of plan adjustment. This intelligent scheduling method effectively combines expert experience and data analysis, enabling more flexible response to complex production abnormalities.

[0033] For ease of understanding, the method provided by the present embodiment will be described in the flow below in combination with the above scenario. Please refer to Figure 1 , which is a flowchart of the intelligent plan scheduling method based on large models in the embodiments of the present application.

[0034] S101, acquire the textual description record of the production expert processing the production plan adjustment, the textual description record containing the triggering scenario, the consideration factor, the adjustment idea and the adjustment result.

[0035] Among them, the production expert means a technical personnel with rich practical experience and professional knowledge in the field of production plan adjustment. The textual description record refers to the detailed text material formed by the expert when handling the production plan adjustment problem. The triggering scenario is used to represent the specific situation that triggers the plan adjustment, such as equipment failure, material shortage, etc. The consideration factor is the various conditions that the expert needs to weigh when formulating the adjustment scheme, including device capacity, material supply, personnel configuration, etc. The adjustment idea represents the specific logical process of the expert analyzing and solving the problem. The adjustment result is the specific action scheme taken and its implementation effect.

[0036] This step is continuously performed during system initialization and daily operation to accumulate expert experience data. Specifically, the system first collects detailed records of production experts handling various abnormal situations through various channels such as interview records, operation logs, and decision reports. Each record needs to fully record the specific scene description of the abnormal occurrence, the key factors considered by the expert, the problem-solving process, and the final adjustment measures and their effect evaluation. These records are organized using structured templates to ensure the completeness and consistency of the information.

[0037] In some embodiments, the collection and recording of expert experience data can be achieved in various ways: optionally, the system can conduct in-depth communication with production experts for typical abnormal scenes through semi-structured interviews, guide experts to describe the specific process of handling problems in detail, including on-site situation evaluation, scheme development consideration, and execution effect analysis, and convert the interview content into standardized written records; optionally, the system can embed an expert decision recording module in the production management software, and guide the expert to record the decision process simultaneously when the expert performs plan adjustment operation through the preset form, including problem description, influence factor analysis, alternative scheme evaluation, and final scheme explanation. It can be understood that other ways can also be used to collect expert experience data, such as video recording, operation log analysis, etc., which are not limited here.

[0038] S102, input the written description record into the preset production large model for semantic analysis, extract the decision elements and their weight relationships of the production expert in different scenes, and obtain a production scheduling knowledge base.

[0039] Among them, the preset production large model represents an intelligent language model system trained on production field data. Semantic analysis refers to the process of deep understanding and structured processing of text content. Decision elements are used to represent key variables and conditions that affect expert judgment. Weight relationships refer to the importance and mutual influence of each decision element. The production scheduling knowledge base represents a structured data system that stores expert decision knowledge.

[0040] This step is performed after the collection of written description records to convert unstructured expert experience into computable knowledge representation. Specifically, the system inputs the collected written records into the pre-trained production large model, the model first performs word segmentation and semantic understanding on the text, and identifies the key information entities in the text. Then, through deep learning algorithm analysis, the expert considers various factors in different scenes, and extracts the logical relationships and influence degrees between these factors. Finally, the analysis results are stored in the production scheduling knowledge base according to the pre-defined knowledge graph structure, forming an expert knowledge model that can be used for subsequent decision reference.

[0041] In some embodiments, the intelligent conversion of expert experience can be achieved in various ways: optionally, the system can use a deep learning model based on attention mechanism to perform multi-level semantic analysis on the text, including entity recognition, relationship extraction and logical reasoning, to build a complete decision knowledge graph; optionally, the system can combine rule templates and machine learning methods, first extract standardized information from the text based on pre-defined semantic rules, then identify decision patterns in different scenarios through clustering analysis, and finally determine the weight relationship between elements through correlation analysis. It can be understood that other ways can also be used to achieve the intelligent conversion of expert experience, such as reinforcement learning, knowledge distillation, etc., which are not limited here.

[0042] S103, collecting production site real-time data and standardizing the production site real-time data into scene description text, the production site real-time data including equipment running state, material inventory level and in-process product processing progress.

[0043] Among them, the production site real-time data represents the real-time running data collected from various sensors and information systems on the production line. The equipment running state refers to the working state, running parameters and fault information of the production equipment. The material inventory level is used to represent the real-time inventory quantity and location information of various raw materials and semi-finished products. The in-process product processing progress represents the completion of the product batch being processed. The scene description text refers to the conversion of various data into standardized and structured text descriptions. The standardization process is used to represent the process of converting data of different sources and formats into a unified format.

[0044] This step is continuously executed during the operation of the production system, for real-time acquisition of the state information of the production site. Specifically, the system collects equipment running parameters, material inventory changes and production progress information in real time through various data collection devices deployed on the production line, including equipment state monitoring systems, warehouse management systems and production execution systems, etc. After cleaning and preprocessing, the collected data is converted into standardized text descriptions through pre-defined data conversion templates, including equipment state description, material inventory status description and production progress description, etc., to form a complete scene description text.

[0045] In some embodiments, the collection and standardization of production data can be achieved in various ways: optionally, the system can deploy an industrial Internet of Things platform, first collect equipment operation data and environmental data through a distributed sensor network, then perform data preprocessing and preliminary analysis through an edge computing node, and finally upload the processed data to a cloud platform for unified management and conversion; optionally, the system can be based on a production management information system, first extract relevant data from each business system through an interface program at regular intervals, then use data cleaning rules to standardize the data, and finally convert the data into structured text descriptions through a natural language generation model. It can be understood that other methods such as mobile terminal collection, manual input, etc. can also be used to achieve the collection and standardization of production data, which are not limited here.

[0046] S104, input the scene description text into the preset production large model, analyze the correlation degree of the current scene and the historical decision-making scene based on the production scheduling knowledge base, and extract the core decision-making logic of experts in the associated scene.

[0047] Among them, the scene description text represents the standardized processing of the production site state description. The correlation degree refers to the similarity of the current scene and the historical decision-making scene in various aspects. The historical decision-making scene is used to represent the similar scenes handled by experts stored in the knowledge base. The core decision-making logic represents the main decision-making principles and methods adopted by experts when handling specific types of problems. Similarity calculation refers to quantitatively evaluating the similarity between scenes through feature matching and other methods.

[0048] This step is executed after obtaining the current scene description, and is used to extract the decision-making ideas that can be learned from historical experience. Specifically, the system inputs the standardized scene description text into the production large model, and extracts and encodes the scene features through a deep learning algorithm. Then, based on the historical decision-making scenes stored in the knowledge base, the similarity between the current scene and each historical scene is calculated. For historical scenes with high similarity, the system further analyzes the decision-making methods and considerations adopted by experts in these scenes, and extracts the core decision-making logic with universal guiding significance.

[0049] In some embodiments, scene matching and decision logic extraction can be implemented in various ways: optionally, the system can use a vector similarity-based matching method, first converting scene descriptions into high-dimensional feature vectors using a pre-trained model, then calculating the cosine similarity between vectors, and finally filtering out relevant scenes based on a similarity threshold and extracting common decision rules through a decision tree algorithm; optionally, the system can use a graph neural network-based method, first converting scene descriptions into knowledge graph representations, then calculating the structural similarity between scenes through graph matching algorithms, and finally extracting common decision patterns using subgraph mining techniques. It can be understood that other ways can also be used to implement scene matching and decision logic extraction, such as fuzzy reasoning, case reasoning, etc., which are not limited here.

[0050] S105, based on the preset production large model, the production site real-time data is analyzed to obtain the production state evolution trend in the preset time period, and a capacity conflict point is identified from the production state evolution trend, the capacity conflict point including a conflict time interval and a production order involved.

[0051] Among them, the prediction analysis represents the deduction and estimation of the production state change in the future period. The production state evolution trend refers to the change trend of key indicators such as equipment operation, material consumption and order progress. The preset time period is used to represent the time span of the prediction analysis, usually several production shifts or working days. The capacity conflict point represents the time point of the mismatch between supply and demand of production capacity in the prediction period. The conflict time interval refers to the duration range of the capacity conflict. The production order involved refers to the specific order set affected by the capacity conflict. The capacity supply and demand relationship is used to represent the matching degree between equipment capacity and order demand.

[0052] This step is continuously executed after obtaining real-time data, which is used to discover potential capacity problems in advance. Specifically, the system uses the preset production large model to perform time series prediction based on the current collected equipment state, material inventory and production progress data, to obtain the change trend of each production indicator in the future period. Through analysis of the prediction results, the system identifies the time period of capacity supply and demand imbalance, including equipment load exceeding the limit and material supply shortage, etc. For each identified conflict point, the system determines its impact range and the specific orders involved, providing a basis for subsequent plan adjustment.

[0053] In some embodiments, production state prediction and conflict identification can be achieved in various ways: optionally, the system can use a deep learning time series prediction model, first perform feature engineering on historical data to build a training data set, then train a recurrent neural network model to learn the time series variation law, and finally use the trained model to predict the future state and set a threshold for conflict detection; optionally, the system can be based on a simulation model method, first establish a digital twin model containing equipment, materials and orders, etc. elements, then based on the current state for rapid simulation deduction, and finally through the constraint condition verification to find the potential conflict point. It can be understood that other ways can also be used to achieve production state prediction and conflict identification, such as statistical prediction, expert system, etc. method, not limited here.

[0054] In some embodiments, the step specifically includes: Determine the standard capacity of each device in the preset time period based on the device operating state, and obtain the device capacity reference value.

[0055] Wherein, the device operating state represents the working parameters and performance indicators of the device, including running speed, stability, failure rate, etc. data. The standard capacity represents the theoretical production capacity of the device under normal operating conditions. The preset time period represents the time range for capacity calculation. The device capacity reference value represents the effective capacity level considering actual operation factors. The running efficiency factor represents the comprehensive correction coefficient of various factors affecting the actual capacity of the device.

[0056] The capacity reference value is calculated by analyzing the device operating data. First, obtain the standard working parameters of the device, including standard processing speed, standard working time, etc. basic data. Then collect the actual operating data of the device, calculate the device running efficiency factor: device running efficiency = actual running time / planned working time × (1-failure downtime rate) × (1-planned downtime rate) × quality pass rate. Multiply the standard capacity by the running efficiency factor to obtain the device capacity reference value. For example, a device with a standard capacity of 100 pieces / hour and a running efficiency of 85%, its capacity reference value is 85 pieces / hour. For multiple devices, calculate the capacity reference value of each device to form a device capacity reference data set.

[0057] According to the product processing progress and process route, calculate the device capacity demand of the to-be-processed order in the preset time period, and obtain the device load prediction value.

[0058] Wherein, the product processing progress represents the completion of the order being processed. The process route represents the sequence and processing requirements of product processing. The to-be-processed order represents the production task that needs to be completed in the preset time period. The device load prediction value represents the demand amount of order processing on device capacity. The processing time quota represents the standard processing time required to complete a single product.

[0059] Load calculation is based on product status and orders to be processed. First, the process progress data of the product is obtained, and the remaining processing procedures and processing quantities of each order are determined. Then, according to the process route, the orders to be processed are expanded according to the process, and the processing time of each process is calculated: Process processing time = order quantity x single piece processing time quota. The processing time of the orders that need to use the same equipment in the same time period is accumulated to obtain the load prediction value of the equipment. For example, a certain equipment needs to process A order 100 pieces (each piece 2 hours) and B order 50 pieces (each piece 3 hours) in a preset time period, and the load prediction value of the equipment is 350 hours.

[0060] The preset time period is divided into multiple time units according to the minimum production unit time, and the equipment load rate is calculated in each time unit. When the equipment load prediction value in the first time unit is greater than the equipment capacity benchmark value, the time unit is marked as a load overrun unit.

[0061] Among them, the minimum production unit time represents the minimum time granularity for load analysis, such as division by hour or by shift. Time unit represents the specific time period divided according to the minimum production unit time. Equipment load rate represents the ratio of equipment load prediction value to capacity benchmark value. Load overrun unit represents the time unit whose equipment load rate exceeds 100%.

[0062] Fine load analysis is performed on the preset time period. First, the preset time period is divided into multiple consecutive time units according to the minimum production unit time (such as 1 hour). Then, in each time unit, the equipment load rate = load prediction value in the time unit / equipment capacity benchmark value x 100%. When the load rate of a certain time unit is greater than 100%, the time unit is marked as a load overrun unit. For example, the load prediction value of a certain equipment in the 8:00-9:00 time unit is 95 pieces / hour, and the capacity benchmark value is 85 pieces / hour, and the load rate is 111.8%, so the time unit is marked as a load overrun unit.

[0063] Material demand replenishment balance is calculated in each time unit. When the material demand in the second time unit is greater than the material inventory level, the time unit is marked as a material shortage unit.

[0064] Among them, the material demand replenishment balance represents the comparison between the material demand and the available inventory. Material demand represents the amount of material needed by the production order in the time unit. Material inventory level represents the available material quantity considering the in-transit replenishment. Material shortage unit represents the time unit in which the material demand exceeds the inventory level.

[0065] Material balance analysis is performed in the time unit dimension. First, the material demand in each time unit is calculated: based on the process BOM and production plan, the material consumption of all orders in this time unit is counted. Then the material inventory level of the time unit is determined: initial inventory + expected arrival quantity - safety stock. When the material demand of a time unit is greater than the inventory level, mark this time unit as a material shortage unit. For example, the material demand in a certain time unit is 1000 pieces, the current inventory is 800 pieces, the expected arrival is 100 pieces, and the safety stock is 200 pieces. The available inventory of this time unit is 700 pieces, which is marked as a material shortage unit.

[0066] The union of load overrun units and material shortage units is determined as the conflict time interval, and the orders to be processed in the conflict time interval are determined as the production-related orders.

[0067] wherein the conflict time interval represents a continuous time period where there is a capacity conflict or material shortage. The production-related orders represent the set of orders that need to be processed in the conflict time interval. The union operation represents a mathematical operation that unifies the identification of both load overrun and material shortage problems.

[0068] The identified problem time units are analyzed. First, the load overrun units and material shortage units are merged to obtain the set of all problem time units. A continuity analysis is performed on these time units, and adjacent problem time units are merged into conflict time intervals. Then, all orders to be processed in these conflict time intervals are queried, and these orders are determined as production-related orders. For example, a certain device has load overrun from 8:00 to 10:00 and material shortage from 9:00 to 11:00, so the final conflict time interval is determined to be 8:00-11:00, and all orders to be processed during this period are marked as production-related orders.

[0069] S106, based on the capacity conflict points, process constraint conditions and expert core decision logic, generate a plurality of alternative adjustment schemes, input the plurality of alternative adjustment schemes into a preset target evaluation function for calculation, and select the alternative adjustment scheme with the highest comprehensive score as the plan scheduling scheme.

[0070] wherein the process constraint condition represents the technical requirements and limitation rules that must be followed in the production process. The alternative adjustment scheme refers to a set of feasible solutions designed for the identified capacity conflicts. The preset target evaluation function is used to represent a mathematical model for measuring the pros and cons of the scheme, which includes multiple evaluation indexes. The comprehensive score represents the weighted score of the scheme in each index. The plan scheduling scheme is the final determined production plan adjustment scheme. The evaluation index system is used to represent multiple dimensions of evaluating the scheme, including production efficiency, on-time delivery rate, etc.

[0071] This step is executed after the capacity conflict points are identified, for generating and optimizing solutions. Specifically, the system first generates multiple feasible adjustment schemes according to the production order characteristics and process constraints within the conflict time interval, combined with expert core decision logic. Each scheme contains specific order adjustment strategies, such as order splitting, process transfer, batch merging, etc. Then the system inputs these alternative schemes into the pre-set target evaluation function, quantitatively evaluates them from multiple dimensions such as production efficiency, delivery date satisfaction, and changeover loss, and calculates the comprehensive score of each scheme. Finally, the scheme with the highest score is selected as the final plan adjustment scheme.

[0072] In some embodiments, scheme generation and evaluation can be implemented in various ways: alternatively, the system can use a rule-based scheme generation method, first construct an adjustment strategy library according to expert experience, then select applicable adjustment strategies based on conflict characteristics, finally generate multiple alternative schemes through strategy combination, and use a weighted scoring model for scheme evaluation; alternatively, the system can use heuristic optimization algorithms, first establish an optimization model considering multiple constraint conditions, then search the solution space to generate alternative schemes through genetic algorithms or other methods, and finally screen schemes based on the Pareto optimality principle. It can be understood that other ways can also be used to implement scheme generation and evaluation, such as reinforcement learning, mixed integer programming, etc., which are not limited here.

[0073] In some embodiments, this step specifically includes: Extracting expert core decision logic from the production scheduling knowledge base.

[0074] Wherein, the production scheduling knowledge base represents a structured data system that stores historical scheduling experience and expert processing methods. Expert core decision logic represents the key decision rules and processing principles adopted by experts in handling production scheduling problems. Decision rules include processing conditions, execution actions, and expected results. Processing principles represent the basic guidelines that need to be followed in the decision-making process. Conflict characteristics represent the key attributes of the current production conflict, including conflict type, degree, and impact range. Similarity calculation represents the numerical value that measures the similarity between the current conflict and historical cases.

[0075] Based on the characteristics of the current production conflict, relevant expert decision logic is retrieved and extracted from the knowledge base. First, a conflict feature vector is constructed, including conflict type (load overrun / material shortage), conflict degree (percentage of exceeding the benchmark value), and order feature (order quantity, urgency) dimension information. Then, using vector similarity calculation method, similar cases with similarity higher than 80% are found in the knowledge base. Decision rule extraction is performed on the selected cases: the standard process of experts dealing with problems is identified through case analysis; the specific measures taken by experts under different conflict degrees are summarized; the key factors considered by experts in decision-making are extracted, such as equipment efficiency, delivery time requirement, and changeover cost. The extracted decision rules are structured and described to form a standard processing template, including trigger conditions, execution actions, constraint conditions, and expected effects, which serves as a guide for subsequent scheme generation.

[0076] Based on the expert core decision logic, a target evaluation function is constructed, which includes production efficiency index, delivery time satisfaction index, and changeover loss index.

[0077] Wherein, the target evaluation function represents a mathematical model for evaluating the pros and cons of the scheme. The production efficiency index represents the utilization efficiency level of equipment and resources. The delivery time satisfaction index represents the matching degree of order completion time and required delivery time. The changeover loss index represents the time and material loss caused by product switching. The index weight represents the importance of each evaluation index.

[0078] Based on the extracted expert decision logic, a standardized target evaluation system is established. Production efficiency index part: equipment utilization rate = actual production time / planned working time × 100%, personnel efficiency = standard working hours / actual working hours × 100%, the weight of this part is 40%. Delivery time satisfaction index part: delivery time achievement rate = on-time order completion number / total order number × 100%, delay impact degree = Σ(delay days × order priority), the weight of this part is 40%. Changeover loss index part: time loss = changeover frequency × standard changeover time, material loss = Σ(loss quantity × material unit price), the weight of this part is 20%. The total evaluation function is constructed: scheme score = production efficiency score × 0.4 + delivery time satisfaction score × 0.4 + changeover loss score × 0.2. Wherein, each sub-item score adopts percentage system, and different dimension indexes are converted into unified standard through linear normalization processing.

[0079] The target evaluation function is set as a preset target evaluation function.

[0080] Wherein, the preset target evaluation function represents a standard evaluation model for subsequent scheme evaluation. The function parameters represent the coefficients and thresholds in the evaluation function. Standardization processing represents the conversion of different indexes into comparable unified scale. The weight coefficient represents the importance of each evaluation index.

[0081] The constructed target evaluation function is standardized and parameters are determined. The calculation standard of each index is set: the production efficiency index takes full load operation as the benchmark (100 points), and deducts points when the actual efficiency is less than 80%; the delivery satisfaction index takes complete on-time delivery as the benchmark (100 points), and deducts 5 points for each day of delay; the changeover loss index takes the planned changeover times as the benchmark (100 points), and deducts 10 points for each additional changeover. The score interval is determined: 90 points and above is an excellent solution, 80-90 points is a good solution, 70-80 points is an acceptable solution, and 70 points and below is a solution to be optimized. The specific form and parameter value of the evaluation function are stored in the system as a unified standard for subsequent evaluation of each adjustment solution.

[0082] An order adjustment strategy set is generated according to the production order and process constraint conditions within the conflict time interval, the order adjustment strategy set including order splitting strategies, order transfer strategies and order delay strategies for the production order.

[0083] Among them, the order adjustment strategy represents a specific processing method to solve production conflicts. The order splitting strategy represents a processing solution that splits the order into multiple small batches. The order transfer strategy represents a processing solution that transfers the order to other production resources. The order delay strategy represents a processing solution that adjusts the order completion time. The process constraint condition represents the technical rules that the order adjustment must follow.

[0084] For production orders within the conflict time interval, adjustment strategies are developed under process constraints. Order splitting strategy development: analyze the matching relationship between order size and equipment capacity, the equipment unit time capacity is P, and the order total is Q, then the minimum batch number =┌Q / P┐ (round up); calculate the economic batch size, considering changeover cost and inventory cost; determine the processing sequence of each batch. Order transfer strategy development: identify alternative equipment with processing capacity, verify that the process parameter matching degree of the equipment needs to reach 95% or more; evaluate the logistics cost of the transfer path; plan the specific transfer time node. Order delay strategy development: calculate the maximum delay time of the order based on customer requirements and contract terms; evaluate the impact of the delay on subsequent orders; develop delay compensation measures. Each strategy is verified for process constraints to ensure that device adaptability, process sequence and quality requirements and other constraint conditions are met.

[0085] The order adjustment strategy set is filtered according to the expert core decision logic, and adjustment strategies that meet the historical decision pattern are selected, and the adjustment strategies that meet the historical decision pattern are combined to generate multiple alternative adjustment solutions.

[0086] Among them, the historical decision-making mode represents the typical solution and processing idea adopted by experts when dealing with similar problems. The adjustment strategy that meets the historical decision-making mode represents an order adjustment method that is highly matched with the processing method of experts. The strategy combination represents the process of combining multiple adjustment strategies according to a specific logic to form a complete scheme. The alternative adjustment scheme represents a candidate solution set formed after screening and combination.

[0087] The order adjustment strategies are screened and combined to form a complete alternative scheme. First, the strategy screening criteria are set based on the core decision-making logic of experts: analyze the preferred processing method of experts in historical cases, such as preferentially adopting order splitting or equipment transfer; extract the key rules of experts in handling problems, such as adopting the splitting strategy when the order size is greater than 2 times the daily production capacity of the equipment. Then, the strategy set is screened: calculate the matching degree of each strategy with the historical decision-making mode, and the matching degree calculation method is the weighted similarity of strategy features and historical case features, and the strategies with a matching degree of more than 85% are retained. Finally, the screened strategies are combined: a single strategy can be used to form a scheme, such as a "pure splitting scheme" that splits the order into 3 batches for continuous production; or multiple strategies can be combined to form a scheme, such as a "splitting + transfer scheme" that splits the order into 2 batches and transfers one of the batches to an alternative equipment. Each alternative scheme specifies the execution steps, resource requirements, and expected results.

[0088] The multiple alternative adjustment schemes are input into the preset target evaluation function respectively to obtain scoring results of the multiple alternative adjustment schemes, and the alternative adjustment scheme with the highest comprehensive score is selected as the plan scheduling scheme based on the scoring results.

[0089] Among them, the scoring result represents the score of each alternative scheme under the target evaluation function. The comprehensive score represents the weighted total score of the scheme in each evaluation dimension. The plan scheduling scheme represents the final determined production plan adjustment scheme. The scoring index represents the specific index item for evaluating the pros and cons of the scheme.

[0090] The generated alternative adjustment schemes are evaluated and compared one by one. First, each alternative scheme is input into the preset target evaluation function for scoring: calculate the production efficiency score of the scheme, including equipment utilization and personnel efficiency; calculate the on-time delivery satisfaction score of the scheme, including on-time delivery rate and delay impact; calculate the changeover loss score of the scheme, including time loss and material loss. The scoring is performed using the preset scoring standards and weights, such as production efficiency index weight 40%, on-time delivery satisfaction index weight 40%, and changeover loss index weight 20%. Then, the scoring results are sorted: arrange the comprehensive scores of the schemes from high to low; analyze the main reasons for the score difference; compare the details of the schemes with similar scores. Finally, the scheme with the highest comprehensive score is selected as the final plan scheduling scheme, and the specific content, scoring result, and selection basis of the scheme are recorded.

[0091] The method provided by the embodiment is further described in a more specific flow. Please refer to Figure 2 , another flowchart of the intelligent scheduling method based on a large model in the embodiment of the present application.

[0092] S201, production verification is performed on the multiple alternative adjustment schemes, execution effects of each scheme under device fluctuation and material fluctuation conditions are analyzed, and a scheme verification result is obtained.

[0093] Among them, production verification refers to evaluating the feasibility and stability of the scheme through simulation or small-scale test. Device fluctuation represents random changes in device performance and availability, including device failure, maintenance requirements, etc. Material fluctuation represents uncertain changes in material supply and quality, including delayed arrival, quality deviation, etc. Execution effect refers to the performance of key indicators of the scheme under various fluctuation conditions. The scheme verification result represents the stability evaluation result of each alternative scheme.

[0094] This step verifies the alternative schemes comprehensively by constructing a simulation environment containing fluctuation factors. First, for each alternative scheme, set multiple different device fluctuation scenarios, including random device failure, planned maintenance, performance fluctuation, etc. At the same time, set different material fluctuation conditions, such as supplier delay, material quality anomaly, etc. Under these fluctuation conditions, execute the plan scheme respectively, and record the changes of production progress, device utilization, material consumption, etc. key indicators. Through statistical analysis of the fluctuation range and deviation degree of each indicator, the robustness of the scheme is evaluated. For key processes, more detailed bottleneck analysis and sensitivity test are performed to identify potential risk points in the execution process of the scheme.

[0095] S202, selecting the optimal adjustment scheme according to the scheme verification result.

[0096] Among them, the optimal adjustment scheme represents the scheme with the best overall performance under various fluctuation conditions. The scheme evaluation index includes production efficiency stability, delivery time guarantee ability, resource utilization efficiency, etc. quantitative indicators.

[0097] Based on the index data obtained in the verification phase, multi-dimensional scheme evaluation and comparison are performed. First, calculate the mean and standard deviation of each scheme under different fluctuation scenarios to evaluate the average performance and stability of the scheme. For the production efficiency index, calculate the fluctuation range and minimum guarantee level of device utilization. For the delivery time index, analyze the delay risk and maximum delay degree of order completion time. For the resource utilization index, evaluate the efficiency of device and material use and the change of inventory level. Through setting index weight to make comprehensive score, select the scheme with the highest total score and each index meeting the basic requirements as the optimal scheme.

[0098] S203, record the relevant information of the capacity conflict point, expert core decision logic, multiple alternative adjustment schemes and the optimal adjustment scheme.

[0099] Among them, the relevant information represents various data and results generated in the scheme generation and verification process. The capacity conflict point information includes specific data such as conflict time, impact range, and related orders. The expert core decision logic includes decision rules, considerations and solutions. The alternative scheme information includes scheme content, adjustment strategy and expected effect. The optimal scheme information includes specific implementation content and verification results.

[0100] The complete decision process information is recorded and stored in the production scheduling knowledge base to establish a standardized information storage structure. First, record the specific situation of the capacity conflict that triggers the adjustment, including the time period of the conflict, the capacity gap of the equipment, the degree of material shortage, etc. Numerical data. Then record the expert core decision logic to form a structured decision rule description. The specific content of each alternative scheme is recorded in detail, including order adjustment strategy, resource allocation scheme, etc. Finally, record the selection basis and verification data of the optimal scheme, including the specific values of various indicators and the results of fluctuation analysis. These records are organized according to the predefined template format to ensure the completeness and traceability of the data.

[0101] S204, store the relevant information into the production scheduling knowledge base.

[0102] Among them, the production scheduling knowledge base represents a structured database for storing production plan adjustment related knowledge and experience. The relevant information includes capacity conflict characteristics, decision logic rules, scheme content and verification results, etc. The storage structure refers to the data organization form in the knowledge base, including the entity relationship model and attribute definition. The knowledge index represents the retrieval mechanism for quickly retrieving and calling knowledge.

[0103] According to the predefined knowledge storage structure, the information generated in the capacity conflict handling process is classified and organized and stored in the knowledge base. First, establish the feature vector of the capacity conflict scene, including device type, conflict degree, impact range, etc. Dimensional information. Then convert the expert decision logic into a rule set, each rule contains trigger conditions, considerations and execution actions. Structurally describe the alternative scheme, record the adjustment strategy, resource allocation and expected effect of the scheme. Establish a data table of scheme verification results, including test data and statistical results of various performance indicators. Through the relationship model, these information is stored in association to establish a knowledge chain of scene-decision-scheme, and create a multi-dimensional index to support subsequent knowledge retrieval and application.

[0104] S205, collect the equipment running state, personnel operation record and material consumption data of each process of the production line.

[0105] The device running state includes device boot time, running speed, fault record, and other running parameters. The personnel operation record represents the work log of the operator, including operation items, operation time, and abnormal handling information. The material consumption data includes raw material usage, scrap rate, and in-process quantity, and other material flow information. The process represents each processing link in the production process.

[0106] Through the data acquisition equipment deployed at each process of the production line, the production process data is recorded in real time. At the device level, the running state information of the acquisition equipment is collected, including device running time, processing speed, quality parameters, fault alarm, and other data, which are directly read through the device interface or obtained through sensor measurement. At the personnel level, the work process of the operator is recorded, including process operation content, operation time, quality inspection results, and other information, which are recorded through the workstation terminal or mobile equipment. At the material level, the usage of various materials is tracked and recorded, including feeding amount, output amount, and defective product quantity, and other data, which are achieved through barcode scanning or RFID technology. After preliminary processing and formatting, these data are uniformly transmitted to the central database for storage.

[0107] S206, the standard capacity and actual capacity of each process of the production line are calculated to obtain the capacity utilization rate of the production line.

[0108] The standard capacity represents the maximum production capacity of the process under ideal conditions. The actual capacity refers to the effective output capacity of the process in actual production. The capacity utilization rate represents the ratio of the actual capacity to the standard capacity. The calculation period represents the time unit of capacity statistics.

[0109] Based on the collected production data, each capacity index is calculated. First, the standard capacity of each process is calculated according to the device technical parameters and process standards, and the calculation formula is: standard capacity = device theoretical speed × standard working time × standard output rate. Then, the actual production data is counted to calculate the actual capacity, and the calculation method is: actual capacity = actual output quantity ÷ actual production time. The actual output quantity deducts the defective product quantity, and the actual production time deducts the device downtime and process switching time. Finally, the capacity utilization rate is calculated: capacity utilization rate = actual capacity ÷ standard capacity × 100%. For a multi-process production line, the capacity level of the entire production line is determined by identifying the capacity of the bottleneck process. The calculation results can be summarized according to different periods such as hours, shifts, and days to form a capacity analysis report.

[0110] S207, according to the capacity utilization rate, the bottleneck process in the production line is identified, and the corresponding key resource of the bottleneck process is determined.

[0111] The bottleneck process represents the process that restricts the production capacity of the entire production line, usually with the highest capacity utilization or the lowest actual capacity. The key resource refers to the production element that plays a decisive role in the bottleneck process, including core equipment, professionals, and key materials. The capacity restricting factor refers to the specific reason that causes the process to become a bottleneck, such as equipment performance, personnel skills, or material supply, etc.

[0112] According to the capacity utilization data of each process, the bottleneck analysis is performed. First, compare the capacity utilization of each process to identify the link with a capacity utilization exceeding 95% or significantly higher than other processes as the bottleneck process. For the identified bottleneck process, analyze the specific reasons for its capacity limitation: analyze whether there is a problem of insufficient equipment performance or frequent failure through equipment operation data; analyze whether there is a problem of personnel skill shortage or insufficient manpower through personnel operation records; analyze whether there is a problem of insufficient material supply or quality fluctuation through material consumption data. Based on the analysis results, determine the key resources that cause the bottleneck, such as specific model processing equipment, personnel with specific skills, specific specification raw materials, etc.

[0113] S208, based on the usage status of the key resources, device configuration adjustment scheme, personnel configuration adjustment scheme and material configuration adjustment scheme are formulated.

[0114] The device configuration adjustment scheme includes specific measures such as equipment upgrading and modification, equipment layout optimization, and standby equipment allocation. The personnel configuration adjustment scheme includes specific measures such as personnel skill training, shift adjustment, and manpower supplement. The material configuration adjustment scheme includes specific measures such as inventory strategy optimization, supplier management, and material distribution optimization. The resource usage status represents the utilization efficiency and management level of various resources.

[0115] For the usage status of the key resources, corresponding adjustment schemes are formulated respectively. At the device level, analyze the equipment operation data, formulate equipment upgrading and modification plan, including replacing key parts, increasing automation devices, optimizing process parameters, etc.; evaluate the equipment layout, formulate equipment reorganization scheme, improve the efficiency of equipment combination; plan standby equipment management strategy to ensure quick response in case of equipment failure. At the personnel level, evaluate the skill level of employees, formulate targeted training plan; analyze the work load, optimize shift arrangement and personnel configuration; establish talent reserve mechanism to ensure the stability of key position personnel. At the material level, optimize material inventory strategy, establish safety inventory management mechanism; strengthen supplier management, formulate material quality improvement scheme; optimize material distribution process to reduce material turnover time.

[0116] S209, the device configuration adjustment scheme, the personnel configuration adjustment scheme and the material configuration adjustment scheme are combined to form the resource configuration scheme.

[0117] The resource configuration scheme represents a complete implementation scheme formed by integrating various resource adjustment measures. The scheme combination principles include resource coordination, implementation feasibility, and cost effectiveness. Resource coordination represents the coordination and support relationship between various resource adjustment measures. Implementation node represents the execution timing of each adjustment measure.

[0118] The adjustment schemes of equipment, personnel, and materials are systematically integrated. First, a correlation matrix of resource adjustment measures is established to identify the dependency between measures, such as equipment upgrade requiring supporting personnel training, and material specification adjustment requiring equipment parameter optimization. Then, according to the implementation difficulty and urgency of the adjustment measures, the measures are sorted and combined into implementation packages. A specific implementation plan is developed to clarify the execution order, time node, and responsible personnel of each implementation package. Monitoring indicators for the implementation process are established, including equipment efficiency improvement indicators, personnel effectiveness indicators, and material turnover indicators. A complete resource configuration scheme document is formed, including specific implementation content, schedule, resource requirements, and expected goals.

[0119] S210, set a supplier evaluation index, which includes production capacity index, delivery capacity index, and quality level index.

[0120] The production capacity index represents the production scale, equipment level, and technical capacity of the supplier, including quantitative indicators such as annual capacity, equipment automation degree, and process advancement. The delivery capacity index represents the logistics distribution and delivery guarantee capacity of the supplier, including data indicators such as on-time delivery rate, average delivery cycle, and emergency response time. The quality level index represents the product quality assurance capacity of the supplier, including product qualification rate, quality stability, and quality system certification evaluation elements. The evaluation dimension represents the multi-angle investigation direction of the comprehensive strength of the supplier. The index weight represents the importance of each evaluation index in the overall assessment.

[0121] Based on the procurement management requirements, a complete supplier evaluation system is constructed. In terms of production capacity, set the annual capacity utilization rate index (weight 30%), not less than 80%; set the equipment automation rate index (weight 10%), not less than 60%; set the process level score index (weight 10%), full score 100 points, not less than 80 points. In terms of delivery capacity, set the on-time delivery rate index (weight 15%), not less than 95%; set the average delivery cycle index (weight 10%), required to meet the procurement plan cycle; set the emergency order response time index (weight 5%), not more than 48 hours. In terms of quality level, set the product qualification rate index (weight 10%), not less than 99%; set the quality system certification requirement (weight 5%), require ISO9001 certification; set the quality traceability system score (weight 5%), full score 100 points, not less than 90 points.

[0122] S211, obtain the historical supply records of the supplier, evaluate the supplier according to the supplier evaluation index, and generate a supplier evaluation result.

[0123] Wherein, the historical supply record represents the past supply data of the supplier, including order completion, product quality data and service response record. The evaluation period represents the time range of statistical analysis of the performance of the supplier. The evaluation method represents the calculation rule of converting the actual performance of the supplier into a score. The evaluation result represents the score and comprehensive rating of the supplier on each index.

[0124] Based on the set evaluation index system, the historical performance of the supplier is quantitatively evaluated. First, collect the supply records in the past year, including the delivery time, delivery quantity, quality inspection data and other detailed information of each batch of orders. Then process the historical data according to the calculation rules of the evaluation indexes: production capacity score = actual annual supply quantity / promised capacity × 100 × 0.3 + equipment automation rate × 0.1 + process score × 0.1; delivery capacity score = on-time delivery order quantity / total order quantity × 100 × 0.15 + delivery cycle compliance rate × 0.1 + emergency response compliance rate × 0.05; quality level score = product qualified quantity / total supply quantity × 100 × 0.1 + quality system score × 0.05 + traceability system score × 0.05. Finally, the scores of each item are summarized, the comprehensive score of the supplier is calculated, and the supplier grade is determined according to the score, such as 90 points or more for A level, 80-90 points for B level, and 70-80 points for C level.

[0125] S212, select the target supplier according to the material demand of the production plan and the evaluation result of the supplier.

[0126] Wherein, the material demand represents the material specification, quantity and delivery period required by the production plan. The target supplier represents the finally selected material supplier. The supplier classification represents the grade division of the supplier according to the evaluation result. The selection rule represents the decision basis and process of supplier selection.

[0127] According to the material demand of the production plan and the evaluation result of the supplier, the supplier selection process is executed. First, analyze the material demand characteristics, including the technical specification requirements, procurement batch, delivery period and other key factors of the material. Then filter out the candidate suppliers who meet the basic conditions from the supplier library, and give priority to the suppliers with A level evaluation grade. Compare the candidate suppliers comprehensively: evaluate whether their production capacity meets the procurement batch requirement; verify whether their delivery cycle meets the production plan node; confirm whether their product quality meets the technical specification requirement. Among the suppliers who meet the basic conditions, the supplier with the highest comprehensive score is selected as the target supplier. For key materials, develop a primary and backup supplier configuration scheme to ensure the stability of the supply chain.

[0128] S213, send the procurement demand to the target supplier, and obtain the delivery time, delivery quantity and delivery quality information of the target supplier.

[0129] The procurement demand represents a material procurement application to the supplier, including material specifications, procurement quantity, delivery period and other specific requirements. The delivery time represents the material delivery time node promised by the supplier. The delivery quantity represents the material quantity that the supplier can provide. The delivery quality information represents the product quality standards and warranty conditions promised by the supplier. The information confirmation represents the explicit consensus reached with the supplier on the procurement conditions.

[0130] The standardized procurement demand file is sent to the target supplier, which specifically includes the following information: material specifications, technical parameters and quality requirements; specific quantity and minimum order quantity of procurement; required delivery date and delivery location; quality inspection standards and acceptance conditions; payment method and settlement period. After receiving the procurement demand, the supplier needs to reply within 24 hours with a detailed delivery plan: confirming the specific delivery time, including the production cycle and transportation time; confirming the actual available quantity, and explaining the batch delivery plan if there is capacity limitation; providing quality assurance commitment, including product qualification and warranty period. The reply information of the supplier is confirmed and recorded, and the delivery information is entered into the supplier management system as the basis for subsequent monitoring and evaluation.

[0131] S214, set material supply warning rules, including material inventory warning value and delivery delay warning value.

[0132] The material inventory warning value represents the minimum inventory level that triggers inventory warning, which is usually determined based on material supply cycle and safety inventory strategy. The delivery delay warning value represents the maximum delay time that triggers delivery period warning, which is determined based on production plan node and material importance. The warning rule represents the condition judgment logic that triggers the warning. The safety inventory represents the buffer inventory maintained to cope with supply fluctuations.

[0133] A complete warning management mechanism is established according to the characteristics of the material and the production demand. For the setting of the material inventory warning value: calculate the daily consumption of the material, determine the procurement lead time of the material, consider the minimum delivery quantity of the supplier, and set the warning value = daily consumption × (procurement lead time + safety factor). For example, the daily consumption of a certain material is 100 pieces, the procurement lead time is 7 days, and the safety factor is 1.5, then the inventory warning value is set to 1050 pieces. For the setting of the delivery delay warning value: based on the standard delivery period promised by the supplier, combined with the influence degree of the material on production, determine the maximum acceptable delay time. For example, the delay warning value of a key material is set to 24 hours, and that of a general material is set to 72 hours. The warning rules are configured in the material management system to realize automatic monitoring and warning.

[0134] S215, when the material inventory is lower than the material inventory early warning value or the delivery time exceeds the delivery delay early warning value, starting emergency procurement or adjusting the production sequence according to the emergency degree of the production plan.

[0135] Wherein, the emergency degree represents the time urgency of the production plan to the material demand. The emergency procurement represents the procurement behavior of obtaining the required material in an urgent manner. The production sequence adjustment represents changing the original production plan sequence to adapt to the material supply situation. The alternative solution represents the alternative solution adopted when the original material cannot be obtained in time.

[0136] Establish a standard response process after the early warning trigger. When the system detects that the material inventory is lower than the early warning value: first confirm the demand time node of the material in the production plan; when the demand time is urgent (such as within 48 hours), immediately start the emergency procurement program, including contacting alternative suppliers, negotiating urgent production, arranging fast logistics and other measures; when the demand time is relatively loose, evaluate the feasibility of adjusting the production sequence, move the production order of the material to the back, and arrange other producible orders first. When the delivery delay exceeds the early warning value: for the delay time within 72 hours, evaluate the possibility of using inventory turnover or borrowing other order materials; for the delay time exceeding 72 hours, start the alternative supplier procurement program, and seek material alternative solutions if necessary. Update the production plan and adjust the production sequence of the affected order.

[0137] The intelligent production scheduling system in the embodiment of the present application will be described from the perspective of hardware processing. Please refer to Figure 3 , which is a schematic diagram of an entity device structure of the intelligent production scheduling system in the embodiment of the present application.

[0138] It should be noted that, Figure 3 The structure of the intelligent production scheduling system shown is only an example, and should not bring any limitation to the function and use range of the embodiment of the present application.

[0139] As Figure 3As shown, the intelligent scheduling system includes a central processing unit (CPU) 301 which can perform various appropriate actions and processes in accordance with a program stored in a read-only memory (ROM) 302 or a program loaded from a storage section 308 into a random access memory (RAM) 303, such as performing the methods described in the above embodiments. In the RAM 303, various programs and data required for the operation of the system are also stored. The CPU 301, the ROM 302, and the RAM 303 are connected to each other through a bus 304. An input / output (I / O) interface 305 is also connected to the bus 304.

[0140] Connected to the I / O interface 305 are an input section 306 including an audio input device, a push button switch, and the like; an output section 307 including a liquid crystal display (LCD), an audio output device, an indicator, and the like; a storage section 308 including a hard disk and the like; and a communication section 309 including a network interface card such as a LAN (Local Area Network) card, a modem, and the like. The communication section 309 performs communication processing via a network such as the Internet. A drive 310 is also connected to the I / O interface 305 as necessary. A removable media 311 such as a magnetic disk, an optical disk, a magneto-optical disk, a semiconductor memory, and the like is attached to the drive 310 as necessary, so that a computer program read therefrom is installed into the storage section 308 as necessary.

[0141] In particular, the processes described above with reference to the flowcharts can be implemented as a computer software program in accordance with embodiments of the present application. For example, embodiments of the present application include a computer program product comprising a computer program carried on a computer readable medium, the computer program containing a computer program for performing the methods shown in the flowcharts. In such embodiments, the computer program can be downloaded and installed from a network via the communication section 309, and / or installed from the removable media 311. When the computer program is executed by the central processing unit (CPU) 301, various functions defined in the present application are performed.

[0142] Note that specific examples of computer-readable storage media can include but are not limited to an electrical connection having one or more wires, a portable computer diskette, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or Flash memory), an optical fiber, a portable compact disc read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the foregoing. In the present disclosure, computer-readable storage media can be any tangible medium that can contain, or store a program for use by or in connection with an instruction execution system, apparatus, or device.

[0143] The flow diagrams and the block diagrams in the drawings are illustrations of architectures, functional processes, and operational processes, according to various embodiments of the present disclosure. Each block in the flow diagrams and the block diagrams can represent a module, a procedure, or a part of code that comprises one or more executable instructions for implementing the specific logical functions specified for the block. It should also be noted that in some alternative implementations, the functions noted in the blocks can occur out of the order noted in the figures.

[0144] Specifically, the intelligent plan scheduling system of the embodiment includes a processor and a memory, and the memory stores a computer program. When the computer program is executed by the processor, the intelligent plan scheduling method based on a large model provided in the above embodiment is implemented.

[0145] As another aspect, the present disclosure also provides a computer-readable storage medium. The storage medium can be included in the intelligent plan scheduling system described in the above embodiments, or can exist separately and not be assembled into the intelligent plan scheduling system. The storage medium carries one or more computer programs. When the one or more computer programs are executed by a processor of the intelligent plan scheduling system, the intelligent plan scheduling system implements the intelligent plan scheduling method based on a large model provided in the above embodiments.

[0146] The above embodiments are only used to illustrate the technical solutions of the present disclosure, rather than limit them. Although the present disclosure has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that they can still modify the technical solutions recorded in the foregoing embodiments, or make equivalent replacements to some technical features. The modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present disclosure.

[0147] In the above embodiments, the term "when" can be interpreted to mean "if" or "after" or "in response to determining" or "in response to detecting" depending on the context. Similarly, the phrase "on determining" or "if detecting (a stated condition or event)" can be interpreted to mean "if determining" or "in response to determining" or "on detecting (a stated condition or event)" or "in response to detecting (a stated condition or event)" depending on the context.

[0148] Those skilled in the art can understand that all or part of the processes in the above-mentioned embodiments can be instructed by a computer program to relevant hardware, and the program can be stored in a computer readable storage medium. When the program is executed, the processes of the above-mentioned embodiments can be included. The storage medium includes ROM, RAM, magnetic disk or optical disk, and various storage media that can store program codes.

Claims

1. An intelligent production planning method based on a large model, characterized in that: Applied to an intelligent planning and scheduling system, the method includes: Obtaining a textual description record of a production expert's handling of a production plan adjustment, wherein the textual description record includes a triggering scenario, considerations, adjustment ideas, and adjustment results; Input the text description record into a preset production model for semantic analysis, extract the decision factors and weight relationships of the production experts in different scenarios, and obtain a production scheduling knowledge base; Collecting real-time data from the production site and standardizing the data into scene description text, wherein the real-time data from the production site includes equipment operating status, material inventory level, and work-in-progress processing progress; Input the scenario description text into the preset production model, analyze the correlation between the current scenario and the historical decision scenarios based on the production scheduling knowledge base, and extract the core decision logic of the experts in the correlated scenarios; Based on the preset production model, the real-time data of the production site is predicted and analyzed to obtain the production status evolution trend within a preset time period, and the production capacity conflict point is identified from the production status evolution trend, wherein the production capacity conflict point includes the conflict time interval and the production orders involved; Based on the capacity conflict points, process constraints and the expert core decision logic, multiple alternative adjustment plans are generated, the multiple alternative adjustment plans are input into a preset target evaluation function for calculation, and the alternative adjustment plan with the highest comprehensive score is selected as the planned production scheduling plan.

2. The method according to claim 1, characterized in that The method includes performing a predictive analysis on the real-time data of the production site based on the preset production macro model to obtain a production status evolution trend within a preset time period, and identifying a capacity conflict point from the production status evolution trend. The capacity conflict point includes a conflicting time interval and steps involving production orders, specifically including: Determine the standard production capacity of each device within the preset time period based on the device operating status, and obtain a device production capacity benchmark value; Calculate the equipment capacity requirements of the order to be processed within the preset time period based on the work-in-progress processing progress and process route to obtain a predicted equipment load value; Divide the preset time period into multiple time units according to the minimum production unit time, calculate the equipment load rate in each time unit, and mark the time unit as a load-exceeded unit when the equipment load forecast value in the first time unit is greater than the equipment production capacity benchmark value; Calculating a material demand replenishment balance in each time unit, and marking the time unit as a material shortage unit when the material demand in a second time unit is greater than the material inventory level; The union of the load-exceeding units and the material-shortage units is determined as the conflict time interval, and the to-be-processed orders within the conflict time interval are determined as the involved production orders.

3. The method according to claim 1, characterized in that The step of generating multiple alternative adjustment plans based on the capacity conflict point, process constraints, and the expert core decision logic, inputting the multiple alternative adjustment plans into a preset target evaluation function for calculation, and selecting the plan with the highest comprehensive score as the planned production scheduling plan specifically includes: generating the order adjustment strategy set according to the production order involved in the conflicting time interval and the process constraint condition, wherein the order adjustment strategy set includes an order splitting strategy, an order transfer strategy, and an order postponement strategy for the production order involved; screening the set of order adjustment strategies according to the expert core decision logic, selecting the adjustment strategies that conform to the historical decision patterns, and combining the adjustment strategies that conform to the historical decision patterns to generate the multiple alternative adjustment plans; The multiple alternative adjustment plans are respectively input into a preset target evaluation function to obtain scoring results of the multiple alternative adjustment plans, and based on the scoring results, the alternative adjustment plan with the highest comprehensive score is selected as the planned production scheduling plan.

4. The method according to claim 1, wherein After the steps of generating a plurality of alternative adjustment plans based on the capacity conflict point, the process constraints, and the expert core decision logic, inputting the plurality of alternative adjustment plans into a preset target evaluation function for calculation, and selecting the plan with the highest comprehensive score as the planned production scheduling plan, the method further includes: Performing production verification on the multiple alternative adjustment plans, analyzing the execution effect of each plan under the conditions of the equipment fluctuation and the material fluctuation, and obtaining plan verification results; Select the optimal adjustment plan based on the verification results of the plan; Recording relevant information of the capacity conflict point, the expert core decision logic, the multiple alternative adjustment plans, and the optimal adjustment plan; The relevant information is stored in the production scheduling knowledge base.

5. The method according to claim 1, wherein Before the step of generating a plurality of alternative adjustment plans based on the capacity conflict point, the process constraints, and the expert core decision logic, inputting the plurality of alternative adjustment plans into a preset target evaluation function for calculation, and selecting the plan with the highest comprehensive score as the planned production scheduling plan, the method further includes: Extracting the expert core decision logic from the production scheduling knowledge base; Constructing a target evaluation function based on the expert core decision logic, wherein the target evaluation function includes a production efficiency index, a delivery satisfaction index, and a changeover loss index; The target evaluation function is set as the preset target evaluation function.

6. The method according to claim 1, characterized in that After the steps of generating a plurality of alternative adjustment plans based on the capacity conflict point, the process constraints, and the expert core decision logic, inputting the plurality of alternative adjustment plans into a preset target evaluation function for calculation, and selecting the plan with the highest comprehensive score as the planned production scheduling plan, the method further includes: Collecting equipment operating status, personnel operation records and material consumption data for each process of the production line; Calculate the standard capacity and actual capacity of each process of the production line to obtain the capacity utilization rate of the production line; Identifying the bottleneck process in the production line according to the capacity utilization rate, and determining the key resources corresponding to the bottleneck process; Based on the usage status of the key resources, formulate equipment configuration adjustment plans, personnel configuration adjustment plans and material configuration adjustment plans; The equipment configuration adjustment plan, the personnel configuration adjustment plan and the material configuration adjustment plan are combined to form a resource configuration plan.

7. The method according to claim 1, characterized in that After the steps of generating a plurality of alternative adjustment plans based on the capacity conflict point, the process constraints, and the expert core decision logic, inputting the plurality of alternative adjustment plans into a preset target evaluation function for calculation, and selecting the plan with the highest comprehensive score as the planned production scheduling plan, the method further includes: Setting supplier evaluation indicators, including production capacity indicators, delivery capacity indicators and quality level indicators; Obtaining the supplier's historical supply records, evaluating the supplier based on the supplier evaluation indicators, and generating a supplier evaluation result; Selecting a target supplier based on the material requirements of the production plan and the supplier evaluation results; Sending a purchase request to the target supplier to obtain the target supplier's delivery time, supply quantity and supply quality information; Set material supply warning rules, including material inventory warning values ​​and supply delay warning values; When the material inventory is lower than the material inventory warning value or the delivery time exceeds the delivery delay warning value, emergency procurement is initiated or the production sequence is adjusted according to the urgency of the production plan.

8. An intelligent production planning system, characterized in that: The intelligent planning and scheduling system includes: one or more processors and a memory; the memory is coupled to the one or more processors, the memory is used to store computer program code, the computer program code includes computer instructions, and the one or more processors call the computer instructions to enable the intelligent planning and scheduling system to execute the method as described in any one of claims 1-7.

9. A computer-readable storage medium comprising instructions, characterized in that: When the instruction is executed on the intelligent planning and scheduling system, the intelligent planning and scheduling system is caused to execute the method according to any one of claims 1 to 7.

10. A computer program product, characterized in that When the computer program product is run on an intelligent planning and scheduling system, the intelligent planning and scheduling system is enabled to execute the method according to any one of claims 1 to 7.

Images