Liquid crystal display module production resource scheduling method and system

Abstract

The application provides a liquid crystal display module production resource scheduling method and system, relates to the field of industrial data management and production operation optimization, and comprises the following steps: acquiring multi-source operation data of a hot-pressing production line to generate a global operation portrait of the production line; matching the global operation portrait with a plurality of preset management risk assessment rules; if any rule is matched, generating a scheduling decision suggestion for manual intervention; otherwise, generating an automatic scheduling scheme by a pre-trained resource scheduling strategy model; in response to the scheduling decision suggestion for manual intervention, generating a first candidate scheduling suggestion scheme, performing virtual running, and outputting an evaluation conclusion of recommendation, postponement or non-recommendation; if the deviation exceeds a preset tolerance range, triggering a management review and strategy optimization process; the system comprises five modules of data fusion and risk assessment, decision interaction and scheme generation, simulation evaluation, scheme execution and monitoring, and model iteration; and the application improves OTD and OEE indexes of the production line through data driving, simulation verification and closed-loop optimization.

Description

Technical Field

[0001] This invention relates to the field of industrial data management and production operation optimization technology, and in particular to a method and system for scheduling production resources for liquid crystal display modules. Background Technology

[0002] The hot-press bonding process for LCD modules is a core step in the manufacturing process. The efficiency of its production scheduling and management directly determines the company's on-time delivery rate (OTD), overall equipment efficiency (OEE), and overall operating costs.

[0003] The management of related hot press production lines largely relies on the manual experience of scheduling administrators. These administrators need to make decisions based on limited and fragmented information from the production line (such as equipment alarms and order delivery dates). Administrators often cannot form a unified understanding of the overall operational status of the production line, and decisions are often based on lagging information, resulting in slow decision-making responses. After the scheduling plan is formulated, the decision-making process lacks forward-looking simulation and quantitative evaluation, and the scheduling plan is often verified through trial production, which results in high trial and error costs and is prone to quality accidents or delivery delays. Furthermore, the decision-making mechanism is mostly a passive response (such as equipment failure or quality anomalies) rather than proactive intervention, making it difficult to make optimized decisions. Summary of the Invention

[0004] The technical problem to be solved by this invention is to provide a method and system for scheduling production resources of liquid crystal display modules. Through technology integration and closed-loop optimization, it solves the pain points of relying on manual labor, slow response, and poor adaptability in the scheduling of liquid crystal display module production, and realizes the intelligent, lean and sustainable improvement of the scheduling process.

[0005] To solve the above-mentioned technical problems, the technical solution of the present invention is as follows:

[0006] A method for scheduling production resources for liquid crystal display modules, comprising the following steps executed by a production resource scheduling system:

[0007] Step S1: Obtain multi-source operation data of the hot pressing production line, including production order information and equipment status data. Generate a global operation profile of the production line based on the multi-source operation data. Match the global operation profile of the production line with multiple preset management risk assessment rules. If any rule is matched, generate a scheduling decision suggestion for manual intervention. Otherwise, generate an automated scheduling scheme from a pre-trained resource scheduling strategy model.

[0008] Step S2: In response to the manually intervened scheduling decision suggestion, automatically generate a decision support report and push it to the management terminal; receive the scheduling adjustment intention confirmed by the scheduling administrator based on the decision support report, and convert the scheduling adjustment intention into a first candidate scheduling suggestion scheme;

[0009] Step S3: Input the first candidate scheduling suggestion or the automated scheduling scheme into the simulation evaluation module for virtual operation, predict its order delivery on-time rate index and equipment efficiency index, and output the evaluation conclusion of recommendation, postponement or non-recommendation based on the prediction results.

[0010] Step S4: If the evaluation conclusion is a recommendation, the corresponding first candidate scheduling suggestion or the automated scheduling scheme is pushed to the production management terminal for execution, and the actual operational performance data after the scheme is executed is obtained; if the deviation between the indicators in the actual operational performance data of the scheme and the predicted indicators exceeds the preset tolerance range, the management review and strategy optimization process is triggered.

[0011] Step S5: Store the actual operational performance data and the corresponding production line global operational profile and scheduling scheme as training samples for iterative updates of the resource scheduling strategy model.

[0012] Furthermore, in step S1:

[0013] The production order information and equipment operating status data are cleaned and fused.

[0014] The production line global operation profile is calculated through a pre-set scoring model. The quantitative dimensions of the production line global operation profile include at least: equipment maintenance risk score calculated based on equipment alarm frequency and performance degradation trend, delivery risk score calculated based on remaining order delivery time and work-in-process quantity, and process stability score calculated based on process parameter volatility.

[0015] Furthermore, the management risk assessment rules include:

[0016] If the equipment maintenance risk score is lower than the preset safety threshold, equipment maintenance decision recommendations will be generated.

[0017] If the delivery risk score exceeds a preset tolerance threshold, a decision suggestion for adjusting order priority will be generated.

[0018] If the process stability score exceeds the preset warning threshold, a process review decision recommendation will be generated.

[0019] If the planned occupancy rate of any critical resource exceeds the preset saturation threshold for an extended period, a decision suggestion to generate resource reallocation or production cycle adjustment will be triggered.

[0020] If no rules are triggered, the current global operation profile of the production line is input into the resource scheduling strategy model, and the automated scheduling scheme is output. The automated scheduling scheme includes the order execution priority sequence, equipment resource allocation ratio, and process parameter adjustment range.

[0021] Furthermore, in step S2:

[0022] In response to the scheduling decision suggestions made by human intervention, the production resource scheduling system automatically generates a decision support report containing the triggering cause, related data snapshots, impact scope analysis and recommended solutions, and pushes it to the management terminal;

[0023] Based on the decision support report, the dispatch administrator provides the following optional operations in the management terminal interface:

[0024] The automated scheduling scheme recommended by the resource scheduling strategy model shall be adopted;

[0025] Retrieve and select the parameter set that is most similar to the current overall operation profile of the production line from the historical success case database;

[0026] Manually enter the new scheduling adjustment intention;

[0027] The content of the scheduling adjustment intention includes: adjusting the priority of the order queue, reserving capacity quotas for specific orders, and approving the trial use of new process parameter combinations within a limited quantity;

[0028] The production resource scheduling system converts the final confirmed scheduling adjustment intention into a structured first candidate scheduling suggestion and attaches a management constraint label, which includes: applicable product batch range, suggestion validity period, and performance deviation review trigger threshold.

[0029] Furthermore, the evaluation process of the simulation evaluation module includes:

[0030] Level 1 rapid simulation: Based on the historical average cycle time of the production line and the capacity model of bottleneck equipment, estimate the theoretical production cycle, peak resource consumption, and whether there are resource conflicts of the proposed scheme;

[0031] The second level of fine simulation: Based on historical quality data and process parameters, a multiple regression model is used to predict key quality indicators after the implementation of the scheme. The key quality indicators include the predicted bonding failure rate and the predicted proportion of batches with misalignment.

[0032] Based on the combined prediction results of the two-level simulations, a final evaluation is made on the first candidate scheduling proposal or the automated scheduling scheme: if all indicators are better than the preset recommended thresholds, it is evaluated as recommended; if the indicators are better than the safe operation baseline but some indicators do not reach the target thresholds, it is evaluated as postponed and added to the queue for evaluation; if any indicator violates the management red line threshold, it is evaluated as not recommended.

[0033] Furthermore, if the plan is assessed as postponed, the following queuing and reassessment strategy will be implemented:

[0034] Continuously monitor the actual production schedule margin, planned changeover window time, and high-priority order insertion status in the manufacturing execution system;

[0035] Based on the preset optimization objectives, the optimal time for re-evaluation of the proposed solution is dynamically calculated;

[0036] When a management window that meets the preset conditions is detected, a re-evaluation reminder is automatically pushed to the administrator, or the simulation evaluation process is automatically re-executed according to the preset policy; if the re-evaluation is recommended, it is pushed to the management terminal again.

[0037] Furthermore, the iterative update of the resource scheduling strategy model includes:

[0038] Set the admission rules for training samples: the samples must cover a complete order cycle; the deviation between the predicted indicators and the actual performance must be within the preset allowable error range;

[0039] When the number of newly added compliant training samples reaches the preset scale or the system's built-in performance monitor detects that the performance of the resource scheduling strategy model has degraded to the performance drift threshold, the incremental learning process is automatically triggered.

[0040] The incremental learning process employs a transfer learning strategy, which uses new samples to fine-tune the model parameters while retaining the knowledge of the original model.

[0041] The updated model was validated using a reserved validation set. The core validation metrics were: improved correlation coefficient between predicted on-time delivery rate and actual on-time delivery rate, and reduced mean square error between predicted resource utilization rate and actual resource utilization rate.

[0042] If the verification passes, the new version of the model is deployed; if the verification fails, the model is rolled back to the previous version and the model version audit log is recorded.

[0043] Furthermore, the processes that trigger management review and strategy optimization include:

[0044] Generate verification suggestions for sequentially performing production cycle time compliance verification, quality target achievement verification, and equipment health boundary verification;

[0045] Once the verification suggestion is executed, if any verification fails, the current implementation of the scheme will be terminated, and a scheduling scheme that is adopted and rolled back to the previous evaluation recommendation will be generated.

[0046] After the rollback suggestion is implemented, a management review report is automatically generated. The management review report includes quantitative analysis of performance deviation, a snapshot of the associated production line global operation profile, a comparative analysis of the implementation plan and simulation prediction, and management strategy optimization items.

[0047] The management review report is pushed to the management terminal and stored in the historical case library.

[0048] Furthermore, a liquid crystal display module production resource scheduling system includes:

[0049] The data fusion and risk assessment module is configured to collect multi-source operational data of the hot pressing production line, generate a global operational profile of the production line based on the multi-source operational data, and match the global operational profile of the production line with multiple preset management risk assessment rules. If any rule is matched, a manual intervention scheduling decision suggestion is generated; otherwise, an automated resource scheduling suggestion scheme is generated by a pre-trained resource scheduling strategy model.

[0050] The decision interaction and solution generation module is configured to respond to the scheduling decision suggestions provided by the human intervention, automatically generate a decision support report and push it to the management terminal, receive and convert the scheduling adjustment intention confirmed by the scheduling administrator, and generate a first candidate scheduling suggestion solution.

[0051] The simulation evaluation module is configured to virtually run the first candidate scheduling suggestion scheme or the automated scheduling scheme, and output a judgment of recommendation, postponement or non-recommendation as the evaluation conclusion.

[0052] The scheme execution and monitoring module is configured to push the corresponding first candidate scheduling suggestion scheme to the operation interface of the manufacturing execution system for execution if the evaluation conclusion is recommended, and obtain the actual operational performance data of the scheme after execution; if the deviation between the actual indicators and the predicted indicators of the scheme exceeds the preset tolerance range, the management review and strategy optimization process is triggered.

[0053] The model iteration module is configured as a training sample for management decision optimization, triggering and executing incremental training and version verification of the resource scheduling strategy model.

[0054] The above-described solution of the present invention has at least the following beneficial effects:

[0055] The above-mentioned solution of the present invention automatically collects multi-source data such as equipment status and order information in real time, and quickly generates a global operational profile of the production line based on a pre-set scoring model. The system can complete status assessment and decision path selection within minutes, shortening the time required for traditional manual analysis, enabling scheduling decisions to respond promptly to changes in the production environment and reducing the risk of production interruption.

[0056] By using management risk assessment rules for automatic matching, the bias of subjective judgment is avoided; through the rapid and detailed simulation of the simulation assessment module, the production cycle, resource conflicts and quality indicators of the plan can be predicted in advance, ensuring the feasibility of the plan and reducing resource waste or delivery delays caused by unreasonable plans.

[0057] By dynamically adjusting equipment allocation ratios, order priorities, and process parameters, the system can balance production line load and avoid equipment idleness or overload.

[0058] Based on process stability scoring and quality prediction through detailed simulation, the system can monitor process parameter fluctuations in real time and trigger timely reviews or adjustments.

[0059] Through the model iteration module, the system can automatically extract training samples from historical execution data and update the resource scheduling model using an incremental learning strategy, enabling the system to adapt to long-term challenges such as aging production line equipment and changes in order structure, and continuously improve scheduling performance.

[0060] Through real-time risk scoring and early warning mechanisms, risks such as equipment maintenance and delivery delays can be identified in advance, and decision support reports containing data snapshots and impact analysis can be generated. The management review process can quickly locate the root cause when deviations exceed limits and automatically revert to a stable solution, reducing production downtime.

[0061] The system provides an intuitive interface that allows dispatch administrators to flexibly select automation solutions, retrieve historical cases, or make manual adjustments. It leverages the efficiency of automation while retaining the flexibility of human experience, lowering the operational threshold and improving management efficiency. Attached Figure Description

[0062] Figure 1 This is a flowchart of the liquid crystal display module production resource scheduling method provided by the present invention. Detailed Implementation

[0063] Exemplary embodiments of the present disclosure will now be described in more detail with reference to the accompanying drawings. While exemplary embodiments of the present disclosure are shown in the drawings, it should be understood that the present disclosure may be implemented in various forms and should not be limited to the embodiments set forth herein. Rather, these embodiments are provided so that this disclosure will be thorough and complete, and will fully convey the scope of the disclosure to those skilled in the art.

[0064] like Figure 1 As shown, an embodiment of the present invention proposes a method for scheduling production resources for liquid crystal display modules, wherein the production resource scheduling system executes the following steps:

[0065] Step S1: Obtain multi-source operation data of the hot pressing production line. The multi-source operation data includes production order information and equipment status data. Generate a global operation profile of the production line based on the multi-source operation data. Match the global operation profile of the production line with multiple preset management risk assessment rules. If any rule is matched, generate a scheduling decision suggestion for manual intervention. Otherwise, generate an automated scheduling scheme from a pre-trained resource scheduling strategy model.

[0066] Step S2: In response to the scheduling decision suggestions from manual intervention, automatically generate a decision support report and push it to the management terminal; receive the scheduling adjustment intention confirmed by the scheduling administrator based on the decision support report, and convert the scheduling adjustment intention into a first candidate scheduling suggestion scheme;

[0067] Step S3: Input the first candidate scheduling suggestion or automated scheduling scheme into the simulation evaluation module for virtual operation, predict its order delivery on-time rate index and equipment efficiency index, and output the evaluation conclusion of recommendation, postponement or non-recommendation based on the prediction results;

[0068] Step S4: If the evaluation conclusion is recommended, the corresponding first candidate scheduling suggestion or automated scheduling scheme will be pushed to the production management terminal for execution, and the actual operational performance data after the scheme is executed will be obtained; if the deviation between the indicators in the actual operational performance data of the scheme and the predicted indicators exceeds the preset tolerance range, the management review and strategy optimization process will be triggered.

[0069] Step S5: Store the actual operational performance data and the corresponding production line global operational profile and scheduling plan as training samples for iterative updates of the resource scheduling strategy model.

[0070] In this embodiment of the invention, the specific application scenario is a liquid crystal display module production line containing three hot pressing machines (denoted as hot presses HP-01, HP-02, and HP-03). The production resource scheduling system (hereinafter referred to as the system) is deployed on a server in the workshop data center. This server interacts with the PLCs of the three hot pressing machines and the temperature and pressure sensors installed on the machines in real time via the OPCUA industrial communication protocol. Simultaneously, the system interfaces with the upper-level Manufacturing Execution System (MES) and Enterprise Resource Planning System (ERP) through RESTful API interfaces to synchronize production orders and resource status information.

[0071] Step S1: Obtain multi-source operational data: The system periodically (e.g., every minute) obtains the latest data from the following data sources:

[0072] Production order information: obtained from the ERP system, including order number A1 (e.g., quantity 5000 pieces, product model 15.6-inch FHD, delivery deadline 24 hours later) and order number A2 (e.g., quantity 3000 pieces, product model 14.0-inch HD, delivery deadline 48 hours later).

[0073] Equipment status data: Read in real time from the equipment controller and sensors via the OPCUA protocol. For example, HP-01 is currently in operation with a current hot platen temperature of 165°C; HP-02 is in standby mode, but the system log shows that it triggered 3 temperature over-limit alarms in the past hour; HP-03 is in operation with a hot platen temperature of 162°C (in this embodiment, the set value is 165°C, and the hot platen temperature shows a slight decay trend).

[0074] Generate a global operational profile of the production line: The system cleans (e.g., filters sensor transient noise) and fuses the raw data obtained above, and inputs it into a preset scoring model. The model calculates the global operational profile of the current production line according to predefined calculation rules and data analysis methods. The profile contains three quantitative dimensions, and the scores of each dimension are based on a percentage system. The safety, early warning, and other thresholds are preset based on the historical operating data of the production line (e.g., the statistical results of the past year).

[0075] Quantitative dimensions include:

[0076] Equipment maintenance risk score: calculated based on alarm frequency and performance degradation trend. For example, HP-02 has a high alarm frequency, so its equipment maintenance risk score is 35 points. In this embodiment, the full score is 100 points, and the lower the score, the higher the risk.

[0077] Delivery risk score: Calculated based on the remaining time for order delivery and the quantity of work-in-process. For example, considering the urgent delivery of order A1 and the risk of HP-02, the current delivery risk score is 75 points. In this embodiment, the threshold is 60 points. A score exceeding 60 points indicates the presence of risk, and the higher the score, the higher the risk.

[0078] Process stability score: calculated based on the temperature fluctuation rate of HP-03. For example, its process stability score is 85 points. In this embodiment, the preset stability warning threshold is 80 points, and the process stability score is at a normal level.

[0079] Rule matching and decision path selection: The system matches the above-mentioned global operational profile of the production line with the preset management risk assessment rules.

[0080] Rule matching result: If the equipment maintenance risk score is lower than the safety threshold (set to 40 points in this embodiment), a manual intervention scheduling decision suggestion will be generated. Since HP-02's score (e.g., 35 points) is lower than the threshold, the system determines that the match is successful.

[0081] Decision Path: Therefore, the system generates a manual intervention scheduling decision suggestion: Warning: The HP-02 hot press has a high maintenance risk; it is recommended to immediately check and adjust the production plan. The system does not proceed to the process of generating an automated scheduling scheme from the resource scheduling strategy model.

[0082] Step S2: Generate and Push Decision Support Report: In response to the scheduling decision recommendations generated in Step S1, the system automatically generates a decision support report. This report includes: the triggering cause (HP-02 equipment risk), related data snapshots (HP-02 alarm logs, current status), impact scope analysis (potential impact on order A1 delivery), and recommended solutions (e.g., transferring tasks on HP-02 to other equipment and scheduling maintenance for HP-02). The report is pushed to the scheduling administrator's management terminal.

[0083] Confirming the scheduling adjustment intention and generating candidate solutions: After reviewing the report, the scheduling administrator confirms the scheduling adjustment intention on the terminal interface: to adjust part of the tasks (2000 units) of order A1, originally planned to be produced on HP-02, to be produced on the currently idle HP-01. The system receives this intention and converts it into a structured first candidate scheduling suggestion, which clearly specifies the specific allocation quantity and production sequence of each order on the three machines after the adjustment.

[0084] Step S3: Virtual Operation: The system inputs the first candidate scheduling suggestion generated in Step S2 into the simulation evaluation module. This module performs virtual operation based on the constructed production line digital twin model, which includes components such as equipment capacity model, process flow model, and quality prediction model.

[0085] Key performance indicators (KPIs) are predicted: The digital twin model simulates the execution of this scheduling scheme and predicts its results. For example, the prediction shows that the on-time delivery rate of this scheme is expected to reach 98% (compared to 85% in the original scheme), and the equipment efficiency indicator (Overall Equipment Effectiveness, OEE) is expected to be 78% (compared to 75% in the original scheme).

[0086] Output evaluation conclusions: The system compares the predicted indicators with preset thresholds: for example, the on-time delivery rate (98%) is much higher than the recommended threshold (95%), and the equipment efficiency (78%) is higher than the recommended threshold (75%). Therefore, the simulation evaluation module outputs the recommended evaluation conclusions.

[0087] Step S4: Solution Push and Execution: Since the evaluation conclusion is a recommendation, the system pushes the first candidate scheduling suggestion to the production management terminal. The scheduling administrator adopts the solution and issues it to the production line for execution through the MES system.

[0088] Obtain actual performance and monitoring deviations: the actual execution status of the system monitoring plan. For example, after 8 hours, obtain actual operational performance data: the actual order delivery on-time rate is 97.5%, and the actual equipment efficiency is 77%.

[0089] Deviation Judgment and Process Triggering: The system calculates the deviation: for example, the on-time delivery rate deviation is |97.5%-98%|=0.5%, and the equipment efficiency deviation is |77%-78%|=1%. Neither deviation exceeds the preset tolerance range (e.g., 5%), therefore the management review and strategy optimization process is not triggered.

[0090] If any deviation exceeds the tolerance range, the system will automatically trigger the management review and strategy optimization process to correct the scheduling strategy.

[0091] Step S5: The system associates all the data from the entire scheduling process, including actual operational performance data (delivery rate 97.5%, OEE 77%), the corresponding global operational profile of the production line (equipment maintenance risk score, delivery risk score, process stability score), and the executed scheduling scheme (first candidate scheduling suggestion scheme), into a complete training sample and stores it in the sample database. This sample serves as a training sample for subsequent iterative updates to the resource scheduling strategy model.

[0092] It should be noted that the pre-set scoring model of this invention can be implemented in various ways, as long as it can convert multi-source operational data into a quantifiable risk score. For example, a rule-based weighted scoring method can be used to assign weights to different indicators (such as alarm frequency, temperature deviation, etc.) and perform linear combinations; alternatively, a machine learning model trained on historical data (such as logistic regression, gradient boosting decision trees, etc.) can be used for prediction. Those skilled in the art can choose an appropriate model construction method based on the characteristics of the actual production line and the amount of data accumulated.

[0093] In step S1:

[0094] Perform data cleaning and fusion processing on production order information and equipment operating status data;

[0095] The overall operational profile of the production line is calculated using a pre-set scoring model. The quantitative dimensions of the overall operational profile of the production line include at least: equipment maintenance risk score calculated based on equipment alarm frequency and performance degradation trend, delivery risk score calculated based on remaining order delivery time and work-in-process quantity, and process stability score calculated based on process parameter volatility.

[0096] After acquiring raw data from multiple sources including ERP, MES, equipment controllers, and sensors, the system first performs data cleaning and fusion processing to ensure data quality and provide a consistent and reliable data foundation for subsequent profile generation.

[0097] The system performs the following cleaning operations:

[0098] Handling missing values: When consecutive missing values ​​are detected in the data stream (such as no data at two consecutive time points of a temperature sensor), linear interpolation is used to fill in the missing values ​​based on the valid data points before and after the missing values.

[0099] Noise and outlier filtering: Data that deviates significantly from the process range (e.g., a transient peak of 300 MPa returned by a pressure sensor, while the normal range is 250 ± 10 MPa) is identified and filtered based on a preset threshold.

[0100] Format standardization: unify the timestamps of different systems to the ISO8601 format to ensure the correctness of time sequence association.

[0101] The system integrates data in the following ways:

[0102] Key identifier association: Using the device number (e.g., HP01) and order number (e.g., A1) as key identifiers, ERP order information, MES execution status, device sensor data, etc. are associated.

[0103] Build a unified data view: form structured data records, such as [timestamp: 2024XXXX10:00:00, device: HP01, order: A1, status: running, temperature: 165℃, pressure: 255MPa], to provide input for profile calculation.

[0104] The cleaned and fused data is input into a pre-set scoring model. This model encapsulates computational logic based on knowledge of the hot pressing field of liquid crystal display modules. In this embodiment, a graded scoring rule is used as an example to demonstrate the generation process of each quantitative dimension in the overall operation profile of the production line:

[0105] Equipment maintenance risk score:

[0106] Base score determination: The base score is set according to the current status of the equipment. The base score for running equipment with no active alarms is set to 70 points; the base score for standby equipment or equipment with active alarms is set to 60 points; and the base score for equipment stopped or in emergency stop is set to 30 points.

[0107] In this embodiment, HP-02 is in standby mode, so the base score is 60 points.

[0108] Alarm frequency deduction: Points will be deducted based on the recent alarm frequency. The rule is: 5 points will be deducted for each relevant alarm that occurs in the past hour.

[0109] In this embodiment, HP-02 triggered a temperature over-limit alarm once in the past hour, resulting in a deduction of 5 points.

[0110] Performance degradation penalty points: Penalty points are applied when a clear performance degradation trend is detected (such as a critical process parameter continuously deviating from the set value). The rules are as follows: 20 points are deducted when there is a significant degradation trend (such as a process parameter continuously deviating from the set value by more than 3% in the past hour); 10 points are deducted when there is a slight trend (such as a process parameter continuously deviating from the set value by 1%-3% in the past hour); and no points are deducted when there is no trend (such as a process parameter continuously deviating from the set value by less than or equal to 1% in the past hour).

[0111] In this embodiment, HP-02's hot platen temperature deviated from the set value by 6% in the past hour, resulting in a penalty of 20 points.

[0112] In this embodiment, the full score is 100 points. 70 points and above is low risk, 40 to 70 points is medium risk, and below 40 points is high risk. The lower the score, the higher the risk. Therefore, the HP-02 equipment maintenance risk score is 35 points, which is a high risk state.

[0113] Delivery risk score:

[0114] Bonus points for time urgency: Points are awarded based on the remaining time for order delivery. The rules are: less than 48 hours remaining, 40 points; less than 24 hours remaining, 40 points.

[0115] In this embodiment, the remaining delivery time for order A1 is 24 hours, plus 40 minutes.

[0116] Work-in-process (WIP) load bonus: Bonus points are awarded based on the quantity of WIP and the status of the production line. The rules are as follows: if the quantity of WIP exceeds 80% of the normal production line capacity, 20 points are awarded; additionally, if any critical equipment is unavailable, an extra 15 points are awarded.

[0117] In this embodiment, the number of work-in-process is large, and HP-02 has a high risk (which can be considered unusable), so add 40+35=75 points.

[0118] Process stability rating:

[0119] Base score determination: The base score is set based on the current status of the equipment. The base score for the running status is 85 points.

[0120] Volatility deduction: Deductions are made based on the volatility of key process parameters. The rules are as follows: no deductions are made for volatility within ±1%; 5 points are deducted for volatility between ±1% and ±2%; and 10 points are deducted for volatility exceeding ±2%.

[0121] In this embodiment, the temperature fluctuation rate of HP-03 is within the normal range (e.g., within ±1%), so no points are deducted.

[0122] The rules for managing risk assessment include:

[0123] If the equipment maintenance risk score is lower than the preset safety threshold, equipment maintenance decision recommendations will be generated.

[0124] If the delivery risk score exceeds the preset tolerance threshold, a decision suggestion for adjusting the order priority will be triggered.

[0125] If the process stability score exceeds the preset warning threshold, a process review decision recommendation will be generated.

[0126] If the planned occupancy rate of any critical resource exceeds the preset saturation threshold for an extended period, a decision suggestion to generate resource reallocation or production cycle adjustment will be triggered.

[0127] If no rules are triggered, the current global operational profile of the production line will be input into the resource scheduling strategy model, and an automated scheduling plan will be output. The automated scheduling plan includes the order execution priority sequence, equipment resource allocation ratio, and process parameter adjustment range.

[0128] In this embodiment of the invention, during production line operation, at time point T1, the overall production line operation profile is displayed as follows:

[0129] Equipment maintenance risk score (HP-02): 35 (safety threshold 40 points, the lower the score, the higher the risk);

[0130] Delivery risk score: 75 points (tolerance threshold: 60 points, the higher the score, the higher the risk).

[0131] Process stability score (HP-03): 85 points (stability threshold: 85 points, the higher the score, the more stable the process and the lower the risk).

[0132] Planned utilization rate of key resources (e.g., HP-01 hot press): 70% (saturation threshold 85%)

[0133] The system iteratively matches the overall production line operation profile at time T1 with all preset management risk assessment rules:

[0134] Rule 1 Application (Equipment Maintenance Risks):

[0135] Rule condition: If the equipment maintenance risk score is lower than the preset safety threshold (40 points).

[0136] Match criterion: HP-02's score (35 points) is below the safety threshold (40 points). The condition is met, and the match is successful.

[0137] Trigger action: The system immediately triggers and generates a device maintenance decision suggestion, which may include: Alarm: The HP-02 hot press has a high maintenance risk, and it is recommended to arrange maintenance immediately.

[0138] Rule 2 Application (Delivery Risk):

[0139] Rule condition: If the delivery risk score exceeds the preset tolerance threshold (60 points).

[0140] Matching criterion: The delivery risk score (75 points) is higher than the tolerance threshold (60 points). The condition is met, and the match is successful.

[0141] Triggering action: The system triggers the generation of an order priority adjustment decision suggestion, the content of which may be: Prompt: Order A1 has a high delivery risk, it is recommended to evaluate and increase its production priority.

[0142] Application of Rule 3 (Process Stability):

[0143] Rule condition: If the process stability score is lower than the preset excellent threshold (80 points).

[0144] Matching criteria: HP-03's process stability score (75 points) is below the excellent threshold (80 points). The condition is met, and the match is successful.

[0145] Triggering action: The system triggers and generates a process optimization decision suggestion, which may include: Prompt: The process stability of the HP-03 hot press is not good. It is recommended to review the parameters or suspend high-precision orders.

[0146] Rule 4 Application (Resource Utilization):

[0147] Rule condition: If the planned occupancy rate of any critical resource continuously exceeds the preset saturation threshold (85%).

[0148] Matching criterion: The planned utilization rate (70%) of critical resource HP-01 is lower than the saturation threshold (85%). The condition is not met, and the match fails. The system will not trigger decision recommendations based on this rule.

[0149] At time T1, rules one, two, and three are all triggered. Therefore, the system generates a scheduling decision suggestion with multiple recommendations for human intervention and proceeds to step S2 (human decision interaction) instead of executing the automated path that was executed when no rules were triggered.

[0150] For example, when the timeline advances to point T1.5: due to the insertion of an emergency order, the system predicts that HP-01 will have a planned occupancy rate of 90% for the next 4 shifts.

[0151] Rule 4 Application: The system detected that the planned occupancy rate of HP-01 has continuously exceeded the saturation threshold of 85%. The condition is met, and the match is successful.

[0152] Triggering action: The system triggers and generates a resource reallocation decision suggestion, which may include: Warning: The future capacity of the HP-01 hot press will continue to be oversaturated. It is recommended to divert some tasks to HP-03 or adjust the production cycle.

[0153] For example, as time progresses to point T2, the production line status changes:

[0154] Equipment maintenance risk score (HP-02): 55 points (risk eliminated if above 40 points);

[0155] Delivery risk score: 55 points (risk eliminated if below 60 points);

[0156] Process stability score (HP-03): 82 points (above 80 points, the process is stable);

[0157] Critical resource (HP-01) planned utilization rate: 80% (below 85%, not continuously oversaturated);

[0158] The system performs rule matching again:

[0159] Rule 1: A score of 55 is higher than 40, so this rule is not triggered.

[0160] Rule 2: If the score is 55 or lower than 60, it will not be triggered.

[0161] Rule 3: 82 points is higher than 80 points, so this rule is not triggered.

[0162] Rule 4: If the percentage is below 85%, it will not be triggered.

[0163] At time T2, no rules were triggered. The system generated a suggestion for manual intervention, but the content was only a process optimization prompt.

[0164] For example, when time progresses to point T3, the production line is in an ideal stable state:

[0165] All equipment maintenance risk scores are higher than 60 points (far exceeding the safety threshold).

[0166] The delivery risk score is below 50 (far below the tolerance threshold);

[0167] All process stability scores were between 78 and 82 (fluctuating normally around the threshold, not consistently above the excellent threshold).

[0168] All key resource utilization rates were in the 70%-80% range (not exceeding the saturation threshold continuously).

[0169] After the system performed rule matching, it was confirmed that none of the management risk assessment rules (rules one to four) were triggered.

[0170] Since no rules were triggered, the system executed the automated path: inputting the global operational profile of the production line at time T3 into the pre-trained resource scheduling strategy model.

[0171] The resource scheduling strategy model quickly outputs an automated scheduling plan based on the current production line status. This plan is a structured set of instructions, such as including:

[0172] The order execution priority sequence optimized based on delivery urgency and process compatibility is: [Order A2, Order A1].

[0173] Equipment resource allocation ratio: Allocate 40% of the available production time of HP-01 to order A2 and 60% to order A1; allocate all available production time of HP-03 to order A2. Process parameter adjustment range: It is recommended to fine-tune the hot pressing temperature of HP-01 within the range of [163℃, 167℃] in order to increase production capacity by 5% while ensuring quality.

[0174] In step S2:

[0175] In response to manual scheduling decision suggestions, the production resource scheduling system automatically generates a decision support report that includes the triggering cause, related data snapshots, impact scope analysis and recommended solutions, and pushes it to the management terminal;

[0176] Based on the decision support report, the dispatch administrator provides the following optional operations in the management terminal interface:

[0177] Adopt the automated scheduling scheme recommended by the resource scheduling strategy model;

[0178] Retrieve and select the parameter set that is most similar to the current overall operation profile of the production line from the historical success case database;

[0179] Manually enter the new scheduling adjustment intention;

[0180] The proposed scheduling adjustments include: adjusting order queue priorities, reserving capacity quotas for specific orders, and approving the trial use of new process parameter combinations within a limited quantity.

[0181] The production resource scheduling system converts the final confirmed scheduling adjustment intentions into structured first-candidate scheduling suggestions and attaches management constraint labels, which include: applicable product batch range, suggestion validity period, and performance deviation review trigger threshold.

[0182] In this embodiment of the invention, in response to a manual scheduling decision suggestion triggered at time T1, the system automatically generates a comprehensive decision support report and immediately pushes it to the management terminal of the scheduling administrator. For example, the report presents the following key information in a structured manner:

[0183] Triggering reason:

[0184] Equipment maintenance risk: The equipment maintenance risk score (35 points) of the HP-02 hot press is lower than the safety threshold (40 points).

[0185] Delivery risk: The delivery risk score (75 points) for order A1 exceeds the tolerance threshold (60 points).

[0186] Process stability needs to be reviewed: Although the process stability score (82 points) of the HP-03 hot press is higher than the excellent threshold (80 points), its temperature fluctuation rate (±1.5%) is higher than the normal level. It is recommended to review the performance to optimize it.

[0187] Related data snapshot:

[0188] Detailed alarm logs for HP-02 (temperature over-limit alarm records and corresponding timestamps for the past hour).

[0189] Detailed information for order A1 (order number, product model, remaining quantity, delivery deadline).

[0190] The HP-03 temperature fluctuation trend chart shows that the temperature fluctuated between 162°C and 168°C over the past 30 minutes.

[0191] Analysis of the scope of impact:

[0192] Impact on delivery: If HP-02 malfunctions and stops, delivery of order A1 will be delayed by at least 12 hours, which may result in a default.

[0193] Impact on quality: Process variations in HP-03 may increase the defect rate of the current batch of products by 1%-2%.

[0194] Impact on overall capacity: A potential shutdown of HP-02 will reduce the capacity of the entire production line by approximately 30%.

[0195] Recommended solution:

[0196] For HP-02: It is recommended to immediately transfer all tasks on HP-02 to other devices and schedule maintenance.

[0197] For order A1: It is recommended to raise its production priority to the highest level and prioritize its allocation to the stable HP-01.

[0198] For HP-03: It is recommended to review and calibrate its temperature control system.

[0199] After reviewing the decision support report, the dispatch administrator can choose from the following options in the graphical user interface of the management terminal:

[0200] Operation 1: Adopt the automated scheduling scheme recommended by the resource scheduling strategy model;

[0201] Administrators can click to view what automated solutions the system would normally generate if the rules were not triggered. For example, the solution might suggest prioritizing order A2. However, in this high-risk scenario, the administrator believes this solution does not adequately consider the current emergency situation and therefore did not adopt this option.

[0202] Step 2: Retrieve and select a similar parameter set from the historical successful case database;

[0203] The administrator selected this function. Based on the current production line's overall operational profile, the system searched the case library for the historical case with the highest similarity. The system found a case with 88% similarity, which successfully mitigated the risk by diverting high-risk equipment tasks to idle equipment and prioritizing urgent orders.

[0204] After reviewing the detailed parameters of the case (such as equipment allocation ratio and order sequence), the administrator selects to apply the core parameters of this case. The system automatically fills in some scheduling intentions, such as transferring the HP-02 task and increasing the priority of order A1.

[0205] Step 3: Manually enter or confirm the new scheduling adjustment intention;

[0206] Based on the case parameters and their own experience, the administrator manually inputs, adjusts, and confirms the specific scheduling adjustment intentions through operation three, as follows:

[0207] Adjust order queue priority: Set the priority of order A1 to the highest and the priority of order A2 to high.

[0208] Reserve capacity quota for specific orders: Reserve 60% of the capacity quota for HP-01 for order A1 to ensure that it is not affected by other orders.

[0209] Approval for trial use of a new combination of process parameters within a limited quantity: To improve efficiency, order A2 is approved to trial the process parameters of 165°C (originally 163°C) on HP-03, but only for the first batch of 500 products, and quality must be closely monitored.

[0210] After the administrator confirms the final adjustment intention, the system converts it into a structured first-candidate scheduling suggestion. This suggestion is clearly defined, for example:

[0211] Order execution priority sequence: Order A1 (highest priority), Order A2 (high priority).

[0212] Equipment resource allocation plan:

[0213] HP-01: 60% of the production capacity is used for order A1, and 40% is used for order A2.

[0214] HP-02: Shutdown for maintenance, capacity allocation is 0%.

[0215] HP-03: 100% of the production capacity is used for order A2 (the first 500 pieces are for trial use of the new 165℃ process).

[0216] Process parameter adjustment: The hot pressing temperature of HP-03 is set to 165℃ (applicable to the first 500 pieces of order A2).

[0217] At the same time, the system adds management constraint tags to the solution to clarify its scope and conditions of application:

[0218] Applicable product batch range: This solution applies to all batches (5000 pieces) of order A1 and the current production batch (3000 pieces) of order A2, with the new process parameters specifically applicable to the first 500 pieces of order A2.

[0219] Recommended validity period: Valid for 24 hours from time T1 (due to the urgency of equipment maintenance and order delivery).

[0220] Performance deviation review trigger threshold: When the deviation of order delivery on time rate exceeds 3% or the deviation of equipment overall efficiency (OEE) exceeds 2%, the management review process will be automatically triggered.

[0221] Finally, a first candidate scheduling proposal is generated, complete with information and constraint labels.

[0222] The evaluation process for the simulation evaluation module includes:

[0223] Level 1 rapid simulation: Based on the historical average cycle time of the production line and the capacity model of bottleneck equipment, estimate the theoretical production cycle, peak resource consumption, and whether there are resource conflicts of the plan;

[0224] The second level of fine simulation: Based on historical quality data and process parameters, a multiple regression model is used to predict key quality indicators after the implementation of the plan. Key quality indicators include the predicted bonding failure rate and the predicted proportion of batches with misalignment.

[0225] Based on the combined prediction results of the two-level simulations, a final evaluation is made on the first candidate scheduling proposal or the automated scheduling scheme: if all indicators are better than the preset recommended thresholds, it is evaluated as recommended; if the indicators are better than the safe operation baseline but some indicators do not reach the target thresholds, it is evaluated as postponed and added to the queue for evaluation; if any indicator violates the management red line threshold, it is evaluated as not recommended.

[0226] In this embodiment of the invention, the first candidate scheduling suggestion scheme: HP-02 shutdown for maintenance, order A1 priority is increased and 60% capacity is reserved on HP-01, and order A2 is produced on HP-03 and a new process is tested, is used as input for virtual operation and feasibility assessment.

[0227] The simulation evaluation module received the first candidate scheduling suggestion as input. This suggestion is scheduled to be executed within the next 24 hours.

[0228] Evaluation process:

[0229] Level 1 Rapid Simulation: This level of simulation focuses on quickly assessing the macro-level production efficiency and resource feasibility of a solution. Its core is a capacity model built based on historical production line data.

[0230] Simulation basis: For example, the historical average cycle time of the production line: Based on data from the past month, the average production cycle time of this production line is 2.5 minutes per piece (i.e., 24 pieces per hour).

[0231] Bottleneck Equipment Capacity Model: The HP-01 hot press is identified as the bottleneck equipment in the current production line. For example, the theoretical maximum capacity of the HP-01 hot press is 28 pieces per hour. After considering efficiency losses (such as equipment aging), the actual usable capacity is about 20 pieces per hour.

[0232] Simulation execution and prediction results:

[0233] Theoretical production cycle estimation: The simulation module estimates the shortest theoretical production cycle (e.g., 28 hours) required to complete all tasks based on the order allocation in the plan (5000 units remaining for order A1 and 3000 units remaining for order A2) and the capacity of bottleneck equipment HP-01 (20 pieces per hour), combined with the impact of HP-02 shutdown.

[0234] Peak resource utilization analysis: For example, simulations show that during the 10th to 15th hour of the program's execution, the capacity utilization rate of HP-01 will reach a peak of 95%, while the peak utilization rate of HP-03 will be approximately 85%.

[0235] Resource conflict check: Simulate and analyze the smoothness of material flow and the adequacy of personnel operating space to identify potential resource conflicts.

[0236] The second level of refined simulation focuses on evaluating the impact of the scheme on product quality. Its core is to establish a quantitative relationship between process parameters and quality indicators by using a multiple regression model trained based on historical production data.

[0237] Simulation Basics:

[0238] Historical quality database: Contains 50,000 production records from the past year. Each record includes process parameters such as hot pressing temperature, pressure, and time, as well as corresponding product quality inspection results (such as poor bonding, misalignment, etc.).

[0239] Multiple regression model: Based on historical data of the hot pressing process of LCD display modules, a quantitative mapping relationship was established between specific process parameters such as hot pressing temperature, pressure, and time and specific quality defects such as bonding failure rate and misalignment. This model is used to predict the possible results of product quality indicators under given process parameters.

[0240] Simulation execution and prediction results:

[0241] Predicted bonding defect rate: A prediction was made for the instruction in the first candidate scheduling proposal to try out the new 165°C process on HP-03. For example, the prediction results show that, compared with the original 163°C process, the predicted bonding defect rate for order A2 produced at 165°C will increase slightly from the baseline of 1.5% to 1.8%.

[0242] Predicted proportion of batches with misalignment: For example, it is predicted that the proportion of batches with misalignment will decrease from 3% to 2.2% because the plan will concentrate order A1 on the stable HP-01. After order A2 adopts the new 165°C process on HP-03, its predicted proportion of batches with misalignment will be approximately 2.5%.

[0243] The system compares the prediction results of the two-level simulation with the preset multi-level thresholds for comprehensive evaluation.

[0244] Preset multi-level thresholds (set based on statistical analysis of historical production line data):

[0245] Recommended thresholds (high standard, expected target): production cycle less than or equal to 30 hours; peak equipment occupancy less than or equal to 90%; bonding defect rate less than or equal to 2.0%; misalignment rate less than or equal to 2.5%.

[0246] Safety operation baseline (minimum requirements, must be met): production cycle less than or equal to 35 hours; peak equipment occupancy less than or equal to 95%; bonding defect rate less than or equal to 2.5%; misalignment rate less than or equal to 3.5%.

[0247] Management red line thresholds (absolutely prohibited): Production cycle exceeds 40 hours; peak equipment occupancy exceeds 98%; bonding defect rate exceeds 3.0%; misalignment rate exceeds 4.0%.

[0248] Evaluation and comparison:

[0249] Production cycle: 28 hours is less than 30 hours (recommended threshold), which is better than the recommended standard.

[0250] Peak resource usage: Below the safety operation baseline (set to 95% in this embodiment) but above 90% (the recommended threshold is set to 90% in this embodiment), meeting the safety baseline but not reaching the recommended target.

[0251] Predicted poor bonding rate: 1.8% is lower than 2.0% (the recommended threshold in this embodiment is set to 2.0%), which is better than the recommended standard.

[0252] The predicted alignment error rate is 2.2% (A1) and 2.5% (A2), which is less than or equal to 2.5% (the recommended threshold in this embodiment is set to 2.5%), thus meeting the recommended standard.

[0253] Final assessment conclusion:

[0254] In this scenario, all indicators are better than the safe operation baseline and no management red lines have been violated. However, the peak resource utilization indicator did not reach the recommended threshold. Therefore, the simulation evaluation module outputs a postponed evaluation conclusion. The system adds this scenario to the evaluation queue, noting that the main concern is that HP-01's capacity utilization is at a critically high level, posing a certain risk. A further evaluation will be conducted when a better opportunity or conditions are met.

[0255] If the proposal is assessed as postponed, the following queuing and reassessment strategy will be implemented:

[0256] Continuously monitor the actual production schedule margin, planned changeover window time, and high-priority order insertion status in the manufacturing execution system;

[0257] Based on the preset optimization objectives, the optimal time for re-evaluation of the scheme is dynamically calculated;

[0258] When a management window that meets the preset conditions is detected, a re-evaluation reminder is automatically pushed to the administrator, or the simulation evaluation process is automatically re-executed according to the preset policy; if the re-evaluation is recommended, it is pushed to the management terminal again.

[0259] In this embodiment of the invention, the first candidate scheduling suggestion is: HP-02 is shut down for maintenance, order A1 is given priority, order A2 is tested with a new process, and HP-01 is assessed as being postponed because its capacity utilization is at a critically high level and has been added to the queue to be evaluated.

[0260] At time T4, the plan is marked as postponed. The system begins to implement the following queuing and re-evaluation strategy.

[0261] The system's backend continuously monitors real-time data from the Manufacturing Execution System (MES) to determine the appropriate time to implement the postponement plan. Key monitored metrics include:

[0262] Actual production schedule margin: The system calculates in real time whether the progress of all currently executing orders is ahead of schedule. For example, if a medium-priority order A3 is detected, its completion time is expected to be 2 hours ahead of schedule due to smooth upstream material supply. This 2-hour margin is a usable schedule margin window.

[0263] Planned changeover window time: The system query of the production calendar found that, according to the plan, 4 hours after T4, the production line will arrange a planned changeover downtime window of 1 hour to prepare for the new product model.

[0264] Priority order insertion status: The system detected that no new high-priority order insertion plans were made within 2 hours after time T4, indicating that the production line workload has been relatively stable recently.

[0265] The system analyzes the monitoring data based on the preset optimization goal (in this embodiment, the optimization goal is to minimize the interruption of the existing production plan) and dynamically calculates the best time to start the suspension plan.

[0266] Calculation Logic: The system weighs options based on preset decision rules. The main risk of this postponement plan lies in the initial impact on the high-load production line. Therefore, selecting a time when the production line load is low and there is a schedule margin as a buffer as the timing for initiating the plan can significantly reduce the initial impact risk, even if the full execution of the plan will continue into subsequent normal production periods.

[0267] Calculation results: After weighing the above factors, the system concludes that starting the execution plan at the time point when order A3 is completed ahead of schedule and immediately before the planned changeover window (e.g., 3.5 hours after time point T4) can utilize the existing schedule margin as a buffer, thereby minimizing the interruption of the existing plan.

[0268] When the system clock reaches T4+3.5 hours, the system detects that the conditions are met and triggers the re-evaluation process. According to the system's preset strategy, this embodiment executes the automatic re-execution of the simulation evaluation path.

[0269] Automatic re-execution of simulation evaluation: The system automatically obtains the latest production line global operation profile at T4+3.5 hours and inputs it into the simulation evaluation module again.

[0270] Reassessment conclusion: Due to the advancement of production planning (such as the early completion of order A3), the production line load conditions have changed, for example, causing the predicted peak resource utilization of HP-01 to decrease from 95% to 88%. This metric is now better than the recommended threshold of 90%.

[0271] Evaluation Conclusion Update and Push: Based on the new predicted data, the simulation evaluation module outputs recommended evaluation conclusions. The system then pushes the proposed solution (now updated to a recommended status) back to the management terminal for administrator review and adoption.

[0272] The iterative updates to the resource scheduling strategy model include:

[0273] Set the admission rules for training samples: the samples must cover a complete order cycle; the deviation between the predicted indicators and the actual performance must be within the preset allowable error range;

[0274] When the number of newly added compliant training samples reaches the preset scale or the system's built-in performance monitor detects that the performance of the resource scheduling strategy model has degraded to the performance drift threshold, the incremental learning process is automatically triggered.

[0275] The incremental learning process employs a transfer learning strategy, which uses new samples to fine-tune the model parameters while retaining the knowledge of the original model.

[0276] The updated model was validated using a reserved validation set. The core validation metrics were: an increase in the correlation coefficient between the predicted on-time delivery rate and the actual on-time delivery rate, and a decrease in the mean square error between the predicted resource utilization rate and the actual resource utilization rate.

[0277] If the verification passes, the new version of the model is deployed; if the verification fails, the model is rolled back to the previous version and the model version audit log is recorded.

[0278] In this embodiment of the invention, the production resource scheduling system has been running stably for 3 months, during which its core resource scheduling strategy model (initial version V1.0) has generated multiple successful automated scheduling schemes. The system continuously stores the actual operational performance data after the schemes are executed as potential training samples.

[0279] The training sample admission rule is that only samples that meet the following conditions will be marked as compliant training samples and stored in the training sample library:

[0280] Rule 1: Cover the entire order lifecycle. The sample must contain all data generated from the start of order execution to its complete completion. For example, a data segment containing only half of the order execution will be excluded.

[0281] Rule 2: Deviation between prediction and actual results must be within the allowable error range. The absolute deviation between the predicted indicator of the sample (e.g., the on-time delivery rate predicted by the simulation evaluation module is 95%) and the actual performance (actual on-time delivery rate is 93%) is 2%. The system's preset allowable error range is 5%, therefore, the deviation of this sample is within the allowable range and is permitted for inclusion in the database.

[0282] Rule 3: Balance the ratio of positive to negative examples. The system monitors the ratio of positive examples (samples of solutions that are ultimately evaluated as recommended and successfully implemented) to negative examples (samples of solutions that have poor actual implementation results or are evaluated as not recommended) in the sample library, and maintains the ratio at 100:1 through a sampling strategy to ensure sufficient negative examples and prevent bias in model training.

[0283] In another preferred embodiment, the ratio of positive to negative examples is 200:1 or 50:1, depending on the historical failure rate of the production line.

[0284] The system's built-in performance monitor continuously evaluates the performance of Model V1.0. The monitor primarily tracks key metrics between the scheduling schemes predicted by the model and the actual execution results.

[0285] Triggering condition 1: The number of new samples reaches a threshold. For example, the system is preset to automatically trigger an update when 1000 new compliant samples are added. Currently, there are 800 new samples, which does not reach the threshold.

[0286] Triggering condition two: The model performance degrades to the drift threshold.

[0287] Performance monitoring: The monitor found that in the past two weeks, the mean absolute error (MAE) between the predicted on-time delivery rate and the actual on-time delivery rate of the automated scheduling scheme recommended by Model V1.0 increased from 2.5% in the stable period to 4.5%.

[0288] Performance drift assessment: The system's preset performance drift threshold is a mean absolute error exceeding 4.0%. The current metric of 4.5% has exceeded the threshold.

[0289] Triggering action: The system automatically triggers the incremental learning process to prepare for the generation of model V2.0.

[0290] The system initiates an incremental learning process and uses a transfer learning strategy to update model V1.0.

[0291] Application of transfer learning strategies:

[0292] Preserve original model knowledge: The system loads the parameters of the V1.0 model and freezes the parameters of its underlying feature extraction layer.

[0293] Fine-tuning model parameters: The system uses 800 newly added compliant training samples accumulated recently to fine-tune the parameters of the top-level decision layer of the model.

[0294] Output: After incremental learning, a new candidate model V2.0 is generated.

[0295] The new model V2.0 must be validated before it can be deployed.

[0296] Validation process: The system uses a reserved, static benchmark validation set (containing 200 historical samples that have not been used in any model training) to validate the performance of the V2.0 model.

[0297] Core validation metric: Correlation coefficient between predicted on-time delivery rate and actual on-time delivery rate: For example, the V2.0 model performs at 0.86 on this metric, which is a significant improvement over the V1.0 model's 0.74 after the decline, and is better than the original stable period's 0.85.

[0298] Mean squared error between predicted and actual resource utilization rates: For example, the error value of the V2.0 model is 0.04, which is significantly lower than the current error of 0.06 in the V1.0 model.

[0299] Verification Conclusion: All core verification metrics indicate that the V2.0 model outperforms the declining V1.0 model and meets deployment criteria. Verification passed, and the system automatically deploys the V2.0 model as the new production version.

[0300] Version rollback mechanism: After the V2.0 model is deployed, the system continuously monitors the online performance of the V2.0 model (such as the mean absolute error of the predicted on-time delivery rate). If this metric exceeds the performance drift threshold (such as 4.0%) again within the preset observation period (such as 24 hours), the system will automatically roll back to version V1.0.

[0301] Audit log recording: The entire update process is recorded in detail in the model version audit log, which meets the requirements of quality management and traceability.

[0302] The process of triggering management review and strategy optimization includes:

[0303] Generate verification suggestions for sequentially performing production cycle time compliance verification, quality target achievement verification, and equipment health boundary verification;

[0304] Once the verification suggestion is executed, if any verification fails, the current implementation of the plan will be terminated, and a scheduling plan that is adopted and rolled back to the previous evaluation recommendation will be generated.

[0305] After implementing the rollback suggestion, a management review report is automatically generated. The management review report includes quantitative analysis of performance deviations, a snapshot of the overall operation profile of the production line, a comparative analysis of the implementation plan and simulation predictions, and optimization items for management strategies.

[0306] The management review report is pushed to the management terminal and stored in the historical case library.

[0307] In this embodiment of the invention, the first candidate scheduling suggestion that was recommended and adopted after simulation evaluation was: HP-02 was shut down for maintenance, order A1 was prioritized and produced on HP-01, and order A2 was tested on HP-03 with a new process. However, an unexpected deviation occurred during the actual execution.

[0308] For example, after the plan has been implemented for 8 hours, the actual operational performance data continuously monitored by the system shows that the actual value of the order delivery on-time rate deviates from the predicted value by 8%, far exceeding the preset tolerance range (such as 5%). The system automatically triggers the management review and strategy optimization process.

[0309] The system first generates a set of structured verification suggestions and drives relevant systems or personnel to execute them sequentially in order to locate the root cause of the deviation:

[0310] Production cycle time compliance verification:

[0311] Verification content: Verify whether the actual production cycle time is consistent with the cycle time preset in the plan.

[0312] Verification Result: This verification is performed first. For example, the system finds that due to a delay in the supply of a specific material required for order A1, the actual production cycle of HP-01 is 15% slower than the preset value, and the verification fails.

[0313] Quality target achievement verification:

[0314] Verification content: Verify whether the key quality indicators have achieved the expected goals.

[0315] Verification Result: Then perform this verification. For example, the actual bonding failure rate is 1.9%, which is within the allowable range compared to the predicted 1.8%. Verification passed.

[0316] Device health boundary verification:

[0317] Verification content: Verify whether the critical equipment is operating in a healthy state.

[0318] Verification result: This verification is performed last. For example, if the device operating parameters are normal and no health alarms are triggered, the verification passes.

[0319] If the production cycle time compliance check fails, the system will immediately execute the contingency plan:

[0320] Abort the current plan: The system sends a command to the MES to abort the continued implementation of the current scheduling plan.

[0321] Generate rollback suggestion: The system automatically generates a rollback suggestion, which reads: Due to the production cycle time being seriously deviated from the plan due to material supply delays, it is recommended to immediately roll back to the stable scheduling scheme with scheme number P1. This scheme was recommended in the previous simulation evaluation and its actual performance did not trigger the review process.

[0322] Once the rollback suggestion is confirmed, the system executes the rollback operation and automatically generates an in-depth management review report:

[0323] Execute rollback: The system rolls back the production process to the P1 scheme and monitors its stable operation.

[0324] Generate a debriefing report: The report is automatically generated and includes the following core contents:

[0325] Performance Deviation Quantitative Analysis: The analysis clearly indicates an 8% deviation in on-time order delivery rate, and quantifies that this deviation could lead to a 10-hour delay in the delivery of order A1. The root cause was identified as a mismatch between production takt time and material supply delays.

[0326] Snapshot of the associated production line global operation profile: Attached is the production line global operation profile data at the moment before the plan was implemented.

[0327] Comparative analysis of the implementation plan and simulation prediction: A detailed comparison was made between the planned production cycle and the actual cycle, highlighting that the difference in the material supply process is an external risk that the simulation model did not fully consider.

[0328] Management strategy optimization items: Based on the analysis results, specific optimization suggestions are proposed. For example:

[0329] Short-term strategy: Strengthen information sharing with material suppliers and include the real-time delivery time of suppliers in the data collection scope.

[0330] Long-term strategy: Optimize the simulation assessment module, increase the simulation of supply chain risk dimensions, or add a supply chain stability scoring item to the scoring model.

[0331] Push to management terminal: The system pushes this complete management review report to the production scheduling administrator and supply chain management manager.

[0332] Stored in the historical case repository: This report and all associated data are stored in the historical case repository as a complete case of scheduling failure caused by external supply chain issues. This case can be used in the future to support human decision-making by scheduling administrators, or as a negative sample for iterative training of resource scheduling strategy models, thereby achieving continuous optimization of the system.

[0333] A liquid crystal display module production resource scheduling system, comprising:

[0334] The data fusion and risk assessment module is configured to collect multi-source operational data from the hot pressing production line, generate a global operational profile of the production line based on the multi-source operational data, and match the global operational profile of the production line with multiple preset management risk assessment rules. If any rule is matched, a manual intervention scheduling decision suggestion is generated; otherwise, an automated resource scheduling suggestion scheme is generated by a pre-trained resource scheduling strategy model.

[0335] The decision interaction and solution generation module is configured to respond to scheduling decision suggestions from human intervention, automatically generate decision support reports and push them to the management terminal, receive and convert scheduling adjustment intentions confirmed by the scheduling administrator, and generate the first candidate scheduling suggestion solution.

[0336] The simulation evaluation module is configured to virtually run the first candidate scheduling suggestion or the automated scheduling scheme, and output a judgment of recommendation, postponement or non-recommendation as the evaluation conclusion.

[0337] The solution execution and monitoring module is configured to push the corresponding first candidate scheduling suggestion solution to the operation interface of the manufacturing execution system for execution if the evaluation conclusion is recommended, and obtain the actual operational performance data of the solution after execution; if the deviation between the actual indicators and the predicted indicators of the solution exceeds the preset tolerance range, the management review and strategy optimization process is triggered.

[0338] The model iteration module is configured as a training sample for management decision optimization, triggering and executing incremental training and version verification of the resource scheduling strategy model.

[0339] In this embodiment, the system is deployed on a server in the workshop data center. The server interacts with the equipment PLC controllers, temperature or pressure sensors of the hot pressing production line in real time via the OPCUA industrial communication protocol; at the same time, it interfaces with the upper-level Manufacturing Execution System (MES) and Enterprise Resource Planning System (ERP) through RESTful API interfaces to synchronize order and resource status.

[0340] For example, when the system detects a maintenance risk with the HP-02 hot press.

[0341] Data fusion and risk assessment module:

[0342] Data Acquisition: Timed (e.g., every minute) collection of multi-source operational data from ERP, MES, equipment controllers, and sensors. For example, it was found that HP-02 triggered three temperature over-limit alarms in the past hour.

[0343] Profile Generation: The data is cleaned and integrated, and a global operational profile of the production line is generated based on a pre-set scoring model. The calculated equipment maintenance risk score for HP-02 is 35 points.

[0344] Rule matching and decision-making: Match the profile with preset rules. If the matching rule's equipment maintenance risk score is below 40, manual intervention is triggered. Since 35 is below 40, the match is successful.

[0345] Output: This module generates a manual intervention scheduling decision suggestion with the message: "HP-02 equipment has a high maintenance risk." This suggestion is automatically pushed to the decision interaction and solution generation module via the system's internal message bus. The path of generating an automated solution by the resource scheduling strategy model is not triggered. If no rule is matched, this module will call the pre-trained resource scheduling strategy model to directly generate an automated resource scheduling suggestion and output it to the simulation evaluation module.

[0346] Decision interaction and solution generation module:

[0347] Response Recommendation: In response to the above human intervention recommendation, an automatic decision support report is generated, which includes the triggering reason, data snapshot and recommended solution (such as transferring the HP-02 task), and pushed to the administrator terminal.

[0348] Receiving Intent: Based on the report, the dispatch administrator confirms the dispatch adjustment intention on the terminal interface, such as transferring the task of order A1 on HP-02 to HP-01.

[0349] Generating a scheme: This intention transformation generates a structured first candidate scheduling suggestion scheme, which clarifies the new equipment allocation and order sequence.

[0350] Output: The generated first candidate scheduling proposal will be output to the simulation evaluation module.

[0351] Simulation evaluation module:

[0352] Virtual operation: After receiving the first candidate scheduling suggestion, it is virtually operated based on the production line digital twin model.

[0353] Predictive assessment: Through two-level simulation, predict the production cycle (e.g., 28 hours) and key quality indicators (e.g., bonding defect rate of 1.8%) after its execution.

[0354] Comprehensive assessment: The prediction results are compared with multi-level thresholds. It is assumed that all indicators are better than the safety baseline, but some do not meet the recommended targets.

[0355] Output: Output a deferred evaluation conclusion and add the proposed solution to the evaluation queue. Simultaneously, this conclusion is fed back to the decision-making interaction and solution generation module to notify the administrator.

[0356] Solution Execution and Monitoring Module:

[0357] When a solution is recommended: If the simulation evaluation concludes that it is recommended, this module will push the solution to the MES operation interface for administrators to adopt and implement. Subsequently, it will monitor the actual operational performance data (such as the actual on-time delivery rate) after the solution is implemented in real time.

[0358] When the deviation exceeds the limit: For example, if another executed plan subsequently deviates, and the plan execution and monitoring module detects that the deviation between the actual indicator and the predicted indicator exceeds the preset tolerance range (e.g., 5%), the plan execution and monitoring module will automatically trigger the management review and strategy optimization process.

[0359] Model Iteration Module:

[0360] Sample Management: Continuously manage the actual performance data acquired by the solution execution and monitoring module. Validate the data and only store compliant training samples (e.g., covering the entire period and with deviations within acceptable limits) in the sample repository.

[0361] Performance monitoring and triggering: Monitor the online performance of the resource scheduling strategy model. Automatically trigger incremental learning when the number of new samples reaches a certain scale, or when a decline in model performance is detected.

[0362] Iterative update: Using a transfer learning strategy, the model parameters are fine-tuned with new samples to generate a new version of the candidate model (such as V2.0).

[0363] Validation and Deployment: Validate the new model using a reserved validation set. If validation passes, deploy the new version; if it fails, roll back to the old version and log the changes.

[0364] Output: The updated, higher-performance resource scheduling strategy model is deployed and then used by the data fusion and risk assessment module in subsequent automated decision-making.

[0365] The above description represents the preferred embodiments of the present invention. It should be noted that those skilled in the art can make various improvements and modifications without departing from the principles of the present invention, and these improvements and modifications should also be considered within the scope of protection of the present invention.

Claims

1. A method for scheduling production resources for liquid crystal display modules, characterized in that, The production resource scheduling system executes the following steps: Step S1: Obtain multi-source operation data of the hot pressing production line, including production order information and equipment status data. Generate a global operation profile of the production line based on the multi-source operation data. Match the global operation profile of the production line with multiple preset management risk assessment rules. If any rule is matched, generate a scheduling decision suggestion for manual intervention. Otherwise, generate an automated scheduling scheme from a pre-trained resource scheduling strategy model. Step S2: In response to the scheduling decision suggestion made by human intervention, a decision support report is automatically generated and pushed to the management terminal; Receiving the scheduling adjustment intention confirmed by the scheduling administrator based on the decision support report, and converting the scheduling adjustment intention into a first candidate scheduling suggestion scheme, including: In response to the scheduling decision suggestions made by human intervention, the production resource scheduling system automatically generates a decision support report containing the triggering cause, related data snapshots, impact scope analysis and recommended solutions, and pushes it to the management terminal; Based on the decision support report, the dispatch administrator provides the following optional operations in the management terminal interface: The automated scheduling scheme recommended by the resource scheduling strategy model shall be adopted; Retrieve and select the parameter set that is most similar to the current overall operation profile of the production line from the historical success case database; Manually enter the new scheduling adjustment intention; The content of the scheduling adjustment intention includes: adjusting the priority of the order queue, reserving capacity quotas for specific orders, and approving the trial use of new process parameter combinations within a limited quantity; The production resource scheduling system converts the final confirmed scheduling adjustment intention into a structured first candidate scheduling suggestion and attaches a management constraint label, which includes: applicable product batch range, suggestion validity period, and performance deviation review trigger threshold. Step S3: Input the first candidate scheduling suggestion or the automated scheduling scheme into the simulation evaluation module for virtual operation, predict its order delivery on-time rate and equipment efficiency indicators, and output an evaluation conclusion of recommendation, postponement, or non-recommendation based on the prediction results; the evaluation process of the simulation evaluation module includes: Level 1 rapid simulation: Based on the historical average cycle time of the production line and the capacity model of bottleneck equipment, estimate the theoretical production cycle, peak resource consumption, and whether there are resource conflicts of the proposed scheme; The second level of fine simulation: Based on historical quality data and process parameters, a multiple regression model is used to predict key quality indicators after the implementation of the scheme. The key quality indicators include the predicted bonding failure rate and the predicted proportion of batches with misalignment. Based on the combined prediction results of the two-level simulations, a final evaluation is made on the first candidate scheduling proposal or the automated scheduling scheme: if all indicators are better than the preset recommended thresholds, it is evaluated as recommended; if the indicators are better than the safe operation baseline but some indicators do not reach the target thresholds, it is evaluated as suspended and added to the queue for evaluation; if any indicator violates the management red line threshold, it is evaluated as not recommended. Step S4: If the evaluation conclusion is a recommendation, the corresponding first candidate scheduling suggestion or the automated scheduling scheme is pushed to the production management terminal for execution, and the actual operational performance data after the scheme is executed is obtained; if the deviation between the indicators in the actual operational performance data of the scheme and the predicted indicators exceeds the preset tolerance range, the management review and strategy optimization process is triggered. Step S5: Store the actual operational performance data and the corresponding production line global operational profile and scheduling scheme as training samples for iterative updates of the resource scheduling strategy model.

2. The liquid crystal display module production resource scheduling method according to claim 1, characterized in that, In step S1: The production order information and equipment operating status data are cleaned and fused. The production line global operation profile is calculated through a pre-set scoring model. The quantitative dimensions of the production line global operation profile include at least: equipment maintenance risk score calculated based on equipment alarm frequency and performance degradation trend, delivery risk score calculated based on remaining order delivery time and work-in-process quantity, and process stability score calculated based on process parameter volatility.

3. The liquid crystal display module production resource scheduling method according to claim 2, characterized in that, The management risk assessment rules include: If the equipment maintenance risk score is lower than the preset safety threshold, equipment maintenance decision recommendations will be generated. If the delivery risk score exceeds a preset tolerance threshold, a decision suggestion for adjusting order priority will be generated. If the process stability score exceeds the preset warning threshold, a process review decision recommendation will be generated. If the planned occupancy rate of any critical resource exceeds the preset saturation threshold for an extended period, a decision suggestion to generate resource reallocation or production cycle adjustment will be triggered. If no rules are triggered, the current global operation profile of the production line is input into the resource scheduling strategy model, and the automated scheduling scheme is output. The automated scheduling scheme includes the order execution priority sequence, equipment resource allocation ratio, and process parameter adjustment range.

4. The liquid crystal display module production resource scheduling method according to claim 3, characterized in that, If the plan is assessed as postponed, the following queuing and reassessment strategy will be implemented: Continuously monitor the actual production schedule margin, planned changeover window time, and high-priority order insertion status in the manufacturing execution system; Based on the preset optimization objectives, the optimal time for re-evaluation of the proposed solution is dynamically calculated; When a management window that meets the preset conditions is detected, a re-evaluation reminder is automatically pushed to the administrator, or the simulation evaluation process is automatically re-executed according to the preset policy. If the reassessment determines it is recommended, it will be pushed to the management terminal again.

5. The liquid crystal display module production resource scheduling method according to claim 1, characterized in that, The iterative update of the resource scheduling strategy model includes: Set the admission rules for training samples: the samples must cover a complete order cycle; the deviation between the predicted indicators and the actual performance must be within the preset allowable error range; When the number of newly added compliant training samples reaches the preset scale or the system's built-in performance monitor detects that the performance of the resource scheduling strategy model has degraded to the performance drift threshold, the incremental learning process is automatically triggered. The incremental learning process employs a transfer learning strategy, which uses new samples to fine-tune the model parameters while retaining the knowledge of the original model. The updated model was validated using a reserved validation set. The core validation metrics were: improved correlation coefficient between predicted on-time delivery rate and actual on-time delivery rate, and reduced mean square error between predicted resource utilization rate and actual resource utilization rate. If the verification passes, the new version of the model is deployed; if the verification fails, the model is rolled back to the previous version and the model version audit log is recorded.

6. The liquid crystal display module production resource scheduling method according to claim 1, characterized in that, The process of triggering management review and strategy optimization includes: Generate verification suggestions for sequentially performing production cycle time compliance verification, quality target achievement verification, and equipment health boundary verification; Once the verification suggestion is executed, if any verification fails, the current implementation of the scheme will be terminated, and a scheduling scheme that is adopted and rolled back to the previous evaluation recommendation will be generated. After the rollback suggestion is implemented, a management review report is automatically generated. The management review report includes quantitative analysis of performance deviation, a snapshot of the associated production line global operation profile, a comparative analysis of the implementation plan and simulation prediction, and management strategy optimization items. The management review report is pushed to the management terminal and stored in the historical case library.

7. A liquid crystal display module production resource scheduling system, used to implement the liquid crystal display module production resource scheduling method according to any one of claims 1 to 6, characterized in that, include: The data fusion and risk assessment module is configured to collect multi-source operational data of the hot pressing production line, generate a global operational profile of the production line based on the multi-source operational data, and match the global operational profile of the production line with multiple preset management risk assessment rules. If any rule is matched, a manual intervention scheduling decision suggestion is generated; otherwise, an automated resource scheduling suggestion scheme is generated by a pre-trained resource scheduling strategy model. The decision interaction and solution generation module is configured to respond to the scheduling decision suggestions provided by the human intervention, automatically generate a decision support report and push it to the management terminal, receive and convert the scheduling adjustment intention confirmed by the scheduling administrator, and generate a first candidate scheduling suggestion solution. The simulation evaluation module is configured to virtually run the first candidate scheduling suggestion scheme or the automated scheduling scheme, and output a judgment of recommendation, postponement or non-recommendation as the evaluation conclusion. The scheme execution and monitoring module is configured to push the corresponding first candidate scheduling suggestion scheme to the operation interface of the manufacturing execution system for execution if the evaluation conclusion is recommended, and obtain the actual operational performance data of the scheme after execution; if the deviation between the actual indicators and the predicted indicators of the scheme exceeds the preset tolerance range, the management review and strategy optimization process is triggered. The model iteration module is configured as a training sample for management decision optimization, triggering and executing incremental training and version verification of the resource scheduling strategy model.

Images