Management method and system for drive-by-wire MES system

Abstract

The invention discloses a management method and system for a drive-by-wire MES system, and relates to the technical field of MES systems, the system is composed of a plurality of functional modules, and the system comprises a data acquisition module for acquiring production line operation data in real time, the operation data comprising sensor data, production line data and process parameter data; the data processing module processes the operation data by deploying a lightweight algorithm at an edge node, and uploads the processed operation data to a cloud for deep analysis, and the deep analysis comprises trend analysis, bottleneck identification and prediction modeling; in combination with a deep analysis result, mapping a physical production environment in real time, and simulating different scheduling strategy responses; the task scheduling and optimizing module is used for dynamically adjusting task priorities by adopting a reinforcement learning algorithm on the basis of production line data, and introducing a buffer area capacity adjusting mechanism to cope with load fluctuation; and receiving process limitation link information predicted by the MES system, and dynamically adjusting a transportation path in combination with a distributed path planning strategy.

Description

Technical Field

[0001] The present invention relates to the technical field of MES systems, and in particular to a management method and system for a wire-controlled MES system. Background Art

[0002] With technological advancements and changing market demands, MES systems have continued to develop and improve, and are gradually being applied to various manufacturing industries. They help companies improve production efficiency and flexibility while ensuring product quality. Through real-time data collection, production scheduling, quality control, and equipment management, they achieve transparent and intelligent management of the production process. They serve as a data bridge between the enterprise resource planning system and the shop floor control layer, focusing on dynamic optimization at the execution layer. Through sensors, PLCs, RFID, and other technologies, they collect data such as equipment status, process parameters, and material flow, dynamically adjusting production plans to respond to urgent orders or equipment failures. Through a computerized management system for shop floor production processes, they serve as a link between the upper-level enterprise resource planning system and the lower-level production equipment control system. Line-controlled MES systems primarily focus on production site management and optimization to improve production efficiency, product quality, and resource utilization. They collect various data on the production line in real time, including equipment status, process parameters, and material usage, enabling real-time monitoring of the production process. They conduct reasonable production scheduling and scheduling based on order demand and actual production capacity, optimizing production processes and reducing waiting times.

[0003] In traditional line-controlled MES systems, management methods usually rely on preset fixed rules for task scheduling and decision-making. This model has obvious limitations in terms of real-time data-driven. At the same time, due to delays in data collection and processing, the system's real-time performance is insufficient, resulting in delayed updates to production status, affecting the timeliness and accuracy of decision-making. Data accuracy issues are also prominent, especially in equipment status monitoring and quality control. Traditional methods are difficult to avoid noise interference and data loss, further weakening the reliability of the system. Taking automobile manufacturing as an example, when an equipment failure or bottleneck occurs at a workstation on a production line, it is difficult for the existing MES system to reallocate production tasks to other available workstations through dynamic scheduling algorithms. Summary of the Invention

[0004] To achieve the above objectives, the present invention is implemented through the following technical solutions: A management system for a wire-controlled MES system, comprising: Data acquisition module, which collects production line operation data in real time. The operation data includes sensor data, production line data and process parameter data; The data processing module processes operational data by deploying lightweight algorithms at edge nodes and uploads the processed data to the cloud for in-depth analysis, including trend analysis, bottleneck identification, and predictive modeling. Combined with the in-depth analysis results, it maps the physical production environment in real time and simulates responses to different scheduling strategies. The task scheduling and optimization module uses a reinforcement learning algorithm to dynamically adjust task priorities based on production line data, while also introducing a buffer capacity adjustment mechanism to cope with load fluctuations. It receives process constraint information predicted by the MES system and, combined with a distributed path planning strategy, dynamically adjusts transportation routes. The prediction and control module analyzes historical equipment data based on the LSTM neural network, predicts potential failures, triggers preventive maintenance work orders, and monitors production quality in real time.

[0005] Furthermore, the process of collecting production line operation data in real time is as follows: By deploying near the production line, sensor data is collected; the sensor data includes temperature sensors, pressure sensors and displacement sensors; Production line data includes real-time material location, order urgency, and material inventory; Using standardized protocols, the data collected by sensors is transmitted to the MES system for real-time monitoring and adjustment, and saved as process parameter data.

[0006] Furthermore, the process of processing the operating data is as follows: Deploy low-complexity models at edge nodes to perform real-time cleaning and feature extraction of operational data. Leverage distributed computing to process multi-source data streams in parallel, combine LSTM neural networks to predict equipment failures, and dynamically adjust task priorities through reinforcement learning algorithms. Leverage event-driven architectures to trigger immediate responses. Dynamically allocate resources at edge nodes through containerized deployment, and upload critical data to the cloud for in-depth analysis based on the TSN time-sensitive network. Clean the operation data, unify the operation data from different sensor sources into a standard format, and perform structured analysis on unstructured operation data.

[0007] Furthermore, the process of the in-depth analysis is as follows: Trend analysis: Using sliding average and wavelet transform to eliminate sensor noise, applying large-window sliding average to the denoised data to extract long-term trends, and determining trend stationarity through unit root tests. Combining autocorrelation and partial autocorrelation function graphs to identify cyclical components, the ARIMA model is used to predict cyclical fluctuations in the differenced stationary series. Trend and cyclic terms are superimposed to generate a complete trend curve, and model parameters are updated based on real-time MES data. Bottleneck identification: Prometheus monitors equipment OEE, and the Apriori algorithm is used to identify association rules between equipment failures and process parameters. Equipment utilization is calculated based on sensor data and MES system indicators. If the equipment idle rate exceeds the threshold, it is identified as a bottleneck; otherwise, it is not a bottleneck. RFID is used to monitor buffer inventory and, based on the production line topology, locate the source of accumulation. Predictive modeling: Extract characteristic frequencies of equipment failures and use random forests to predict equipment failures; model equipment aging trends based on long-term dependencies in time series, quantify trend direction and duration; and embed fault prediction models and bottleneck rules into the MES system.

[0008] Furthermore, the process of real-time mapping of the physical production environment is as follows: The digital twin platform deployed in the cloud creates corresponding virtual entities, uses industrial simulation software to build an operational production line model, defines event-driven mechanisms, sets state transition logic, and configures task scheduling strategies. The operating data collected and processed by edge computing nodes is uploaded to the cloud in real time. When new data is uploaded, the simulation engine updates the model status in real time.

[0009] Furthermore, the process of dynamically adjusting task priorities using the reinforcement learning algorithm is as follows: S301: Obtain historical operating data, use Plant Simulation to build a digital twin model, and simulate the dynamic behavior of the production line; S302: Balance production efficiency, cost, and quality through discrete action space and continuous action space algorithms combined with entropy weighted multi-objective decision-making methods; S303: Initialize the model, randomly generate a scheduling strategy, update the Q value according to the reward function, use experience replay, store historical state-action-reward data, and use the target network to stabilize the training process; stop training when the model reaches the preset performance indicators in the simulation environment; verify the robustness of the model in extreme scenarios.

[0010] Furthermore, the specific process of introducing the buffer capacity adjustment mechanism is as follows: A physical buffer is set up before and after each key workstation, and the release rhythm of upstream tasks is dynamically adjusted according to the current buffer queue length. If the buffer is close to full, the priority of the task before the process is lowered; otherwise, the release of tasks is accelerated.

[0011] Furthermore, the process of dynamically adjusting the transport path is as follows: The global task is broken down into multiple subtasks, and the subtasks are assigned to the optimal AGV route through a distributed algorithm; the AGV bids for the subtasks based on its own status and prioritizes the optimal path.

[0012] Furthermore, the process of predicting potential failures and triggering preventive maintenance work orders is as follows: The edge computing node receives sensor data in real time and inputs it into the LSTM model in a sliding window format. The sliding window length generates a prediction sequence, outputs the fault category, sets the fault probability threshold based on historical data, and dynamically adjusts the threshold. The prediction results are pushed to the MES system through the API interface, and maintenance work orders are automatically generated. The Bi-LSTM+Attention mechanism is used to capture long-term dependencies in the production process, and the attention weight is used to identify key anomalies. The model is updated regularly.

[0013] A management method for a wire-controlled MES system comprises the following steps: Step 1: Collect production line operation data in real time. The operation data includes sensor data, production line data, and process parameter data. Step 2: Deploy lightweight algorithms at edge nodes to process operational data and upload it to the cloud for in-depth analysis, including trend analysis, bottleneck identification, and predictive modeling. Combined with the in-depth analysis results, the physical production environment is mapped in real time to simulate responses to different scheduling strategies. Step 3: Based on production line data, a reinforcement learning algorithm is used to dynamically adjust task priorities. A buffer capacity adjustment mechanism is also introduced to cope with load fluctuations. The system receives process constraint information predicted by the MES system and combines it with a distributed path planning strategy to dynamically adjust transportation routes. Step 4: Analyze historical equipment data based on the LSTM neural network, predict potential failures, trigger preventive maintenance work orders, and monitor production quality in real time.

[0014] The present invention provides a management method and system for a wire-controlled MES system, which has the following beneficial effects: (1) By deploying lightweight algorithms at edge nodes, the present invention enables real-time cleaning and feature extraction of production line operation data, and utilizes distributed computing to process multi-source data streams in parallel. This architecture not only reduces the time delay for data transmission to the cloud, but also ensures that critical data can be uploaded to the cloud with minimal delay for in-depth analysis, thereby improving the overall system's responsiveness and real-time performance.

[0015] (2) The present invention adopts a reinforcement learning algorithm to dynamically adjust task priorities and combines it with a buffer capacity adjustment mechanism to cope with load fluctuations. The MES system can flexibly adjust the task allocation strategy according to the real-time production status. Based on the historical data analysis of the LSTM neural network, the system can predict potential equipment failures and trigger preventive maintenance work orders.

[0016] (3) Through the Bi-LSTM+Attention mechanism, the system can capture long-term dependencies in the production process, identify key anomalies, and perform modeling and analysis on quality inspection data. Through the digital twin platform deployed in the cloud, a virtual entity of the physical production environment is created, the model status is updated in real time, and the response of different scheduling strategies is simulated. This allows managers to test the effects of various production plans in a virtual environment without actually interfering with the production process, providing support for the formulation of better production plans and further improving the accuracy and efficiency of decision-making. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] Figure 1 Schematic diagram of the system flow of the present invention; Figure 2 Schematic diagram of the overall method of the present invention. DETAILED DESCRIPTION

[0018] The following will provide a clear and complete description of the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. All other embodiments obtained by ordinary technicians in this field based on the embodiments of the present invention without making any creative efforts shall fall within the scope of protection of the present invention. Example 1

[0019] See also Figure 1 , Embodiment 1 of the present application provides a management system for a wire-controlled MES system, the system comprising: Data acquisition module, which collects production line operation data in real time. The operation data includes sensor data, production line data and process parameter data; Sensor data: Sensor data includes temperature sensors, pressure sensors, and displacement sensors. These sensors are deployed near the production line and are responsible for collecting sensor data and making quick decisions, such as adjusting the gripping force to avoid damaging the material. Temperature sensor: Determines the surface temperature of an object by measuring the amount of infrared radiation emitted by the object. It is installed near the welding equipment without direct contact with the object being measured and monitors temperature changes during the welding process. Pressure sensor: This sensor detects changes in the distance between two plates using a capacitive sensor. Pressure is usually measured by capacitance changes caused by pressure. It has high sensitivity and good linearity. It is integrated into the robot's gripping device to measure gripping force and ensure safe handling. Displacement sensor: Calculates the distance between the target object and the sensor by emitting a laser beam and receiving the reflected light signal, providing high-precision measurement without contact. By being installed at the joints of the robot, it monitors the position and movement distance of the robot arm to ensure the accuracy of the robot arm's operation. Production line data: Production line data includes real-time material location, order urgency, and material inventory; it is acquired through visual cameras, RFID equipment, and PLCs. Real-time material location: Industrial vision cameras installed at key production line nodes or on robots capture the image information of materials needed for the production line. Computer vision algorithms analyze the image content and determine the material's location coordinates. For example, on an automotive parts assembly line, vision cameras can be used to identify the specific location of parts on a conveyor belt to guide robotic arms for accurate grasping. For moving materials, continuous imaging combined with motion tracking algorithms can be used to update their location in real time. Order urgency: Order information, including delivery date and customer priority, is extracted from the ERP system as a criterion for determining order urgency. A series of rules are set based on this extracted information. The MES system automatically calculates the urgency of each order based on these rules and adjusts production plans and task allocation strategies accordingly. Material Inventory: RFID equipment and PLCs are used to attach RFID tags to materials or pallets to store information such as material type, quantity, and batch. RFID readers are deployed at key locations on the warehouse entrance and exit, as well as on the production line. When RFID-tagged items pass through, the readers automatically identify and record their entry and exit information, thereby updating inventory status in real time. PLCs are also used to connect to various sensors on the production line to monitor changes in material consumption and, combined with data from the RFID system, to determine inventory levels. Process parameter data: Using standardized protocols, the data collected by sensors is transmitted to the MES system, which monitors and adjusts these process parameters in real time, effectively improving the automation and intelligence level of the production line and ensuring an efficient and stable production process. For example: On a car body assembly line in a car manufacturer, multiple robots are responsible for moving different types of parts to designated workstations for assembly. The MES system performs the following operations based on data obtained from various process parameter sensors: When a car door needs to be moved to the welding workstation, the vision camera first confirms its position on the conveyor belt; the MES system guides the robot arm to the correct position for grasping; during the grasping process, the pressure sensor continuously monitors the force applied to the car door to ensure that it is not damaged by excessive squeezing; at the same time, the MES system decides whether to prioritize this moving task based on the urgency of other tasks on the current production line; if the temperature sensor in the welding area detects an abnormally high temperature, the MES system will automatically reduce the welding speed or temporarily stop the welding operation to prevent damage to the equipment; throughout the entire process, all relevant data is recorded by the MES system for subsequent analysis and quality traceability.

[0020] The data processing module processes operational data by deploying lightweight algorithms at edge nodes and uploads the processed data to the cloud for in-depth analysis, including trend analysis, bottleneck identification, and predictive modeling. Combined with the in-depth analysis results, it maps the physical production environment in real time and simulates responses to different scheduling strategies. Run data for processing: Through lightweight algorithm optimization, low-complexity models are deployed at edge nodes to perform real-time cleaning and feature extraction of raw sensor data. Distributed computing is used to process multi-source data streams in parallel, combined with LSTM neural networks to predict equipment failures, and reinforcement learning algorithms are used to dynamically adjust task priorities. An event-driven architecture is used to trigger immediate responses, such as emergency shutdowns when pressure sensors exceed their limits, ensuring low-latency decision-making. Containerized deployment is used to dynamically allocate resources at edge nodes, and critical data is uploaded to the cloud for in-depth analysis based on the TSN time-sensitive network. Low-complexity model: For resource-constrained edge devices, a low-complexity algorithm is used to adjust task priorities and dynamically assign robot task priorities. For example, material handling for urgent orders is prioritized. The Voronoi diagram algorithm is used for real-time obstacle avoidance, combined with environmental data provided by the MES system, such as obstacle locations, to generate the optimal path. Cleaning and feature extraction: Utilize the edge node's rule engine to filter redundant data. For example, only upload alarm data indicating that the temperature exceeds a threshold, rather than all raw data. High-frequency data is aggregated over time windows, such as merging vibration data once per second into minute-by-minute statistics. Smooth signals using methods such as sliding window averaging and low-pass filters, and perform frequency domain filtering on high-frequency noise using wavelet transforms or Fourier transforms. Detect whether there is packet loss or excessive sampling intervals, and fill in missing values using interpolation. Identify outliers based on statistical methods, and use machine learning models to identify sudden abnormal behaviors, eliminating or replacing outliers with reasonable estimates. Unify data from different sensor sources into a standard format, and perform structured analysis on unstructured data. After completing the operation data cleaning, further extract representative and operational features from the raw data for subsequent modeling and analysis; perform time domain feature extraction, use fast Fourier transform to convert the signal to the frequency domain, extract the main frequency amplitude, frequency band energy distribution, harmonic components, etc.; use methods such as short-time Fourier transform and wavelet transform to analyze the frequency components of the signal that change over time, which is applicable to non-stationary signals; and judge the equipment operation status, such as startup, operation, shutdown, fault, etc., based on sensor data; and identify specific events by combining logical rules or classification models; Upload critical data to the cloud: Use lightweight encoding formats to compress data volume and reduce transmission bandwidth requirements; upload processed data to the cloud via 5G / TSN networks, supporting OPC UA / MQTT protocols to ensure seamless integration with the MES system; In-depth analysis: Trend analysis: Long-term trend identification uses sliding average or wavelet transform to eliminate sensor noise; ARIMA model is used to analyze cyclical fluctuations; large window sliding average is applied to the denoised data again to filter out short-term fluctuations and extract the long-term trend curve; unit root test is used to determine whether the trend term is stationary; cyclical components are identified through autocorrelation function and partial autocorrelation function graphs, and the optimal parameter combination is selected in combination with the AIC / BIC criteria. The ARIMA model is used to predict the stationary series after difference and infer the cyclical fluctuation pattern of the original data; the long-term trend term is superimposed with the cyclical fluctuation term to generate a complete trend analysis result, and the ARIMA model parameters are regularly updated based on the real-time data feedback from the MES system. Bottleneck identification: Identify performance bottlenecks in the production line, such as low equipment utilization and material accumulation. Combine operational data with the Apriori algorithm to identify correlations between equipment failures and specific process parameters. Use Prometheus + Grafana to monitor equipment OEE in real time and identify inefficient links. Sensors collect data such as equipment operating status and material flow rate, and combine this with the MES system's work order completion rate to calculate the overall equipment utilization rate. For example, if the equipment idle rate at a certain workstation consistently exceeds 30%, it is identified as a potential bottleneck. Use RFID or visual sensors to monitor buffer inventory levels. When inventory exceeds a preset threshold, an alert is triggered, and the source of the accumulation is located using the production line topology. Statistical analysis is used to identify inefficient links. For example, if the throughput of a process is significantly lower than that of upstream and downstream processes, and its CPU / memory utilization exceeds 85% for a long period of time, it is determined to be a performance bottleneck. Fault types are classified based on association rules, and the effectiveness of the rules is verified by combining them with equipment maintenance logs. Association rules are embedded in the MES system to monitor parameter combination anomalies in real time and trigger adaptive adjustments. For example, if the temperature exceeds a threshold, the equipment operating speed is automatically reduced to avoid overload failures. Predictive modeling: Extract characteristic frequencies of equipment failures through fast Fourier transforms; use random forests or XGBoost to predict equipment failures; extract trend components from noise and quantify the trend's direction, speed, and duration; capture long-term dependencies in time series through complex nonlinear trends, extract local features, and analyze long-term trends in vibration and temperature data to predict equipment aging, identify bottlenecks, and improve overall efficiency. Real-time mapping of physical production environments: Obtain process path information in the MES system, using process number, name, sequence, equipment resources required for each process, working time standards, bill of materials, as well as quality control points and key parameter requirements; In the digital twin platform deployed in the cloud, corresponding virtual entities are created for each device, each transport chain, and each process node. Using industrial simulation software, an operational production line model is constructed, event-driven mechanisms are defined, state transition logic is set, and task scheduling strategies are configured. Operational data collected and processed by edge computing nodes is uploaded to the cloud in real time to update the state of the digital twin model and unify multi-source heterogeneous data into a structured format. Whenever new data is uploaded, the simulation engine updates the model state, device status, material location and flow status, refreshes task queues and scheduling priorities, modifies process parameters, and triggers abnormal events according to the following logic: The task scheduling and optimization module uses a reinforcement learning algorithm to dynamically adjust task priorities based on production line data, while also introducing a buffer capacity adjustment mechanism to cope with load fluctuations. It receives process constraint information predicted by the MES system and, combined with a distributed path planning strategy, dynamically adjusts transportation routes. By inputting production line data, it outputs task priority sorting, AGV transportation route planning, and resource allocation suggestions between processes; Use reinforcement learning algorithms to dynamically adjust task priorities: Abstract the task scheduling problem into a reinforcement learning environment: element describe state The status of all current tasks (waiting, processing, completed), equipment occupancy, buffer queue length, AGV distribution, etc. action Adjust the priority of a task (increase / decrease), and assign an AGV to perform a transportation task award Positive feedback includes reduced time to complete tasks, eased bottlenecks, improved OEE, and shortened waiting times. Conversely, negative rewards are given. Strategy The reinforcement learning model determines the next action based on the current state Reinforcement learning models rely on simulation environments, reward function design, and policy optimization to make optimal scheduling decisions in complex production environments; Training process: S301: Acquire historical operating data, such as equipment failure records, task completion times, process paths and bottleneck predictions provided by the MES system, and real-time sensor data uploaded from the edge. Use Plant Simulation to build a digital twin model to simulate the dynamic behavior of the production line. S302: Using discrete action space and continuous action space algorithms, combined with entropy weighted multi-objective decision-making methods, balance production efficiency, cost, quality and other objectives; S303: Initialize the model and randomly generate a scheduling strategy, such as randomly assigning task priorities; execute actions in the Agen simulation environment, observe environmental feedback, update Q values or policy network parameters based on the reward function, use experience replay, store historical state-action-reward data to avoid overfitting, and use the target network to stabilize the training process; stop training when the model reaches the preset performance indicators in the simulation environment; verify the robustness of the model in extreme scenarios; Preset performance indicators: Based on historical data analysis, industry standards, and the company's own goal setting, by collecting and analyzing existing data, setting specific improvement goals based on business needs, selecting appropriate KPIs as measurement standards, and using Plant Simulation or other tools to simulate the model's performance in different scenarios, making adjustments until the preset performance indicators are met. Save as a deployable AI service to extract the model's optimal strategy for specific scenarios, such as "when the welding station buffer is >80%, reduce the priority of upstream tasks"; Introducing a buffer capacity adjustment mechanism: Set up virtual / physical buffers before and after each key workstation. Dynamically adjust the release rhythm of upstream tasks based on the current buffer queue length. If the buffer is close to full, reduce the priority of the task before the process or guide the AGV to other paths. The code formula is: ```Python if buffer_level > threshold_high: reduce_priority_of_upstream_tasks() elif buffer_level < threshold_low: increase_priority_of_upstream_tasks() ``` Key workstations: These are the slowest steps in the production process, and their speed determines the output capacity of the entire production line. For example, in automotive manufacturing, welding robots may have slower welding speeds than other workstations due to aging equipment, making them key workstations. Dynamically adjust transportation routes: Break down global tasks into multiple subtasks, such as "transport materials from the warehouse to the welding station" or "AGV 1 departs from the warehouse → AGV 1 passes through the buffer zone → AGV 1 arrives at the welding station." A distributed algorithm assigns subtasks to the most suitable AGVs or robots. AGVs bid for subtasks based on their own status, such as battery level and load capacity, prioritizing the optimal path. AGVs reach consensus through negotiation to avoid resource contention. When an emergency occurs on the production line, such as equipment failure or a change in task priority, the AGV needs to adjust its path in real time: If the welding station is marked as a bottleneck by the MES due to a full buffer, the AGV can be temporarily rerouted to the spraying station to store materials; If a path is blocked by an obstacle, the AGV needs to find an alternative route through local replanning; AGVs adjust their trajectories by exchanging position and gradient information to maximize group efficiency, such as distance to the target and path cost. AGVs predict the future behavior of other devices to avoid potential conflicts in advance, such as predicting a potential collision when an AGV is about to turn. Combining centralized global planning with distributed local obstacle avoidance, AGVs balance efficiency and flexibility. For example: In automotive manufacturing production lines, aging equipment leads to reduced processing capacity at welding stations and a serious backlog of tasks. When buffer capacity approaches its upper limit, AGVs frequently wait, causing congestion in transport routes. Traditional fixed path planning cannot adapt to dynamic load changes, creating traffic jams around welding stations. Bottlenecks at welding stations increase waiting times for downstream processes such as painting and assembly, delaying overall production cycles. The RRT algorithm explores the environment through random sampling, quickly generating feasible paths that avoid obstacles. This algorithm is particularly suitable for complex and dynamic production line environments. When the buffer capacity of a welding station exceeds a threshold, the AGV uses the RRT algorithm to dynamically plan a detour to avoid congested areas of the welding station. For example: Path 1: The AGV starts from the warehouse, goes around to the spraying station to temporarily store materials, and then returns after the welding station load is reduced. Path 2: The AGV drives directly to other non-bottleneck workstations, such as the stamping station, to complete tasks and reduce dependence on the welding station. The MES system predicts the load changes of the welding station in the next 30 minutes based on historical data and real-time sensor input. When the MES predicts that the welding station will be overloaded, the system proactively issues a "reroute instruction" to the AGV, guiding it to go to the spraying station or stamping station first. The feasibility of the path adjustment is verified through the digital twin model to ensure that the new path will not cause congestion in other workstations. Market mechanism algorithms, such as auction allocation, are used to dynamically assign task priorities to avoid multiple AGVs competing for welding station resources at the same time. The AGV uses laser SLAM and visual sensors to perceive the environment in real time, and combines the RRT algorithm to perform local path replanning to avoid sudden obstacles. Through dynamic path adjustment and load balancing, the backlog of tasks at the welding station is reduced by 40%, the overall production line OEE, and the comprehensive equipment efficiency are increased from 82% to 95%, and the production cycle is increased from the original 50 JPH (pieces / hour) to 57. JPH, meeting the production capacity requirements of the M254 model; AGV waiting time around the welding station was reduced from an average of 8 minutes to 5 minutes, reducing congestion by 30%. Through route optimization, AGV energy consumption was reduced by 12%, further reducing operating costs; The prediction and control module uses an LSTM neural network to analyze historical equipment data, predict potential failures, trigger preventive maintenance work orders, and monitor production quality in real time. The hidden layer uses the Bi-LSTM layer to capture the previous and next dependencies, and the Dropout layer prevents overfitting. The output layer uses the Softmax activation function for classification tasks and the linear activation function for regression tasks. The code example is as follows: ```Python from keras.models import Sequential from keras.layers import LSTM, Bidirectional, Dense, Dropout model = Sequential() model.add(Bidirectional(LSTM(64, return_sequences=False), input_shape=(60, 6))) model.add(Dropout(0.2)) model.add(Dense(1, activation='sigmoid')) # Classification task (fault / normal) model.compile(optimizer='adam', loss='binary_crossentropy', metrics=['accuracy']) ``` The edge computing node receives sensor data in real time and inputs it into the LSTM model in a sliding window format. The AGV's motor temperature is uploaded once per second with a sliding window length of 60 seconds to generate a prediction sequence. The node also outputs fault categories such as overheating, bearing wear, and electrical short circuits. It also outputs the remaining service life or failure probability. A fault probability threshold is set based on historical data and dynamically adjusted. The prediction results are pushed to the MES system through the API interface, automatically generating maintenance work orders. If the fault probability at the welding station is predicted to be >90% after 30 minutes, the MES system automatically issues a "Check the welding gun cooling system" work order. The Bi-LSTM+Attention mechanism is used to capture long-term dependencies in the production process, such as weld defects caused by welding current fluctuations. Key anomalies are identified through attention weights, and the model is regularly updated to adapt to changes in process parameters. Bi-LSTM modeling is performed on quality inspection data to identify abnormal patterns, and control limits are set in combination with statistical process control methods. Set a failure probability threshold based on historical maintenance records of similar equipment, such as the typical failure time of a welding gun cooling system. If the failure probability of a welding gun cooling system typically increases from 80% to 90% 30 minutes before failure, then the failure probability threshold is set at 90%. Feedback actual maintenance results, such as whether a fault has occurred, into the training set to retrain the model. Use MLflow or DVC to track model versions to ensure traceability. Feedback the maintenance work order status (completed / deferred) into the prediction model to optimize future predictions. Synchronize the prediction results into the digital twin model to verify the effectiveness of the maintenance strategy. For example, in the fault prediction and quality monitoring of automobile welding stations, the weld quality fluctuates due to equipment aging, and it is necessary to predict faults in advance and optimize process parameters. The welding current, voltage, and weld width sensor data are collected once a second, and the Bi-LSTM+Attention model is used to predict the weld defect probability. When the predicted defect probability is >70%, the welding current is automatically adjusted and a maintenance work order is generated. The weld defect rate is reduced from 5% to 1.2%, and the equipment failure downtime is reduced by 40%.

[0021] Example 2

[0022] See also Figure 2 Based on Example 1, Example 2 of the present application further provides a management method for a wire-controlled MES system, including the following specific steps: Step 1: Collect production line operation data in real time. The operation data includes sensor data, production line data, and process parameter data. Step 2: Deploy lightweight algorithms at edge nodes to process operational data and upload it to the cloud for in-depth analysis, including trend analysis, bottleneck identification, and predictive modeling. Combined with the in-depth analysis results, the physical production environment is mapped in real time to simulate responses to different scheduling strategies. Step 3: Based on production line data, a reinforcement learning algorithm is used to dynamically adjust task priorities. A buffer capacity adjustment mechanism is also introduced to cope with load fluctuations. The system receives process constraint information predicted by the MES system and combines it with a distributed path planning strategy to dynamically adjust transportation routes. Step 4: Analyze historical equipment data based on the LSTM neural network, predict potential failures, trigger preventive maintenance work orders, and monitor production quality in real time.

[0023] The above embodiments can be implemented in whole or in part by software, hardware, firmware, or any other combination thereof. When implemented using software, the above embodiments can be implemented in whole or in part in the form of a computer program product. Those skilled in the art will appreciate that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution.

[0024] The units described as separate components may or may not be physically separate, and the components shown as units may or may not be physical units, and may be located in one place or distributed across multiple network units. Some or all of these units may be selected to achieve the purpose of this embodiment as needed.

[0025] The above is only a specific implementation method of the present application, but the scope of protection of the present application is not limited thereto. Any technician familiar with this technical field can easily think of changes or replacements within the technical scope disclosed in this application, which should be covered by the scope of protection of the present application.

Claims

1. A management system for a wire-controlled MES system, characterized in that: The system includes: Data acquisition module, which collects production line operation data in real time. The operation data includes sensor data, production line data and process parameter data; The data processing module processes operational data by deploying lightweight algorithms at edge nodes and uploads the processed data to the cloud for in-depth analysis, including trend analysis, bottleneck identification, and predictive modeling. Combined with the in-depth analysis results, it maps the physical production environment in real time and simulates responses to different scheduling strategies. The task scheduling and optimization module uses a reinforcement learning algorithm to dynamically adjust task priorities based on production line data, while also introducing a buffer capacity adjustment mechanism to cope with load fluctuations. It receives process constraint information predicted by the MES system and, combined with a distributed path planning strategy, dynamically adjusts transportation routes. The prediction and control module analyzes historical equipment data based on the LSTM neural network, predicts potential failures, triggers preventive maintenance work orders, and monitors production quality in real time.

2. A management system for a wire-controlled MES system according to claim 1, characterized in that: The process of real-time collection of production line operation data is as follows: By deploying near the production line, sensor data is collected; the sensor data includes temperature sensors, pressure sensors and displacement sensors; Production line data includes real-time material location, order urgency, and material inventory; Using standardized protocols, the data collected by sensors is transmitted to the MES system for real-time monitoring and adjustment, and saved as process parameter data.

3. A management system for a wire-controlled MES system according to claim 1, characterized in that: The process of processing the operating data is as follows: Deploy low-complexity models at edge nodes to perform real-time cleaning and feature extraction of operational data. Leverage distributed computing to process multi-source data streams in parallel, combine LSTM neural networks to predict equipment failures, and dynamically adjust task priorities through reinforcement learning algorithms. Leverage event-driven architecture to trigger immediate responses; Dynamically allocate resources on edge nodes through containerized deployment, and upload key data to the cloud for in-depth analysis based on the TSN time-sensitive network; Clean the operation data, unify the operation data from different sensor sources into a standard format, and perform structured analysis on unstructured operation data.

4. A management system for a wire-controlled MES system according to claim 1, characterized in that: The process of the in-depth analysis is as follows: Trend analysis: Using sliding average and wavelet transform to eliminate sensor noise, applying large-window sliding average to the denoised data to extract long-term trends, and determining trend stationarity through unit root tests. Combining autocorrelation and partial autocorrelation function graphs to identify cyclical components, the ARIMA model is used to predict cyclical fluctuations in the differenced stationary series. Trend and cyclic terms are superimposed to generate a complete trend curve, and model parameters are updated based on real-time MES data. Bottleneck identification: Prometheus monitors equipment OEE, and the Apriori algorithm is used to identify association rules between equipment failures and process parameters. Equipment utilization is calculated based on sensor data and MES system indicators. If the equipment idle rate exceeds the threshold, it is identified as a bottleneck; otherwise, it is not a bottleneck. RFID is used to monitor buffer inventory and, based on the production line topology, locate the source of accumulation. Predictive modeling: Extract characteristic frequencies of equipment failures and use random forests to predict equipment failures; model equipment aging trends based on long-term dependencies in time series and quantify trend direction and duration; Embed the fault prediction model and bottleneck rules into the MES system.

5. A management system for a wire-controlled MES system according to claim 1, characterized in that: The process of real-time mapping of the physical production environment is as follows: The digital twin platform deployed in the cloud creates corresponding virtual entities, uses industrial simulation software to build an operational production line model, defines event-driven mechanisms, sets state transition logic, and configures task scheduling strategies. The operating data collected and processed by the edge computing nodes is uploaded to the cloud in real time. When new data is uploaded, the simulation engine updates the model status in real time.

6. A management system for a wire-controlled MES system according to claim 5, characterized in that: The process of dynamically adjusting task priorities using the reinforcement learning algorithm is as follows: S301: Obtain historical operating data, use Plant Simulation to build a digital twin model, and simulate the dynamic behavior of the production line; S302: Balance production efficiency, cost, and quality through discrete action space and continuous action space algorithms combined with entropy weighted multi-objective decision-making methods; S303: Initialize the model, randomly generate a scheduling strategy, update the Q value according to the reward function, use experience replay, store historical state-action-reward data, and use the target network to stabilize the training process; When the model reaches the preset performance indicators in the simulation environment, stop training and verify the robustness of the model in extreme scenarios.

7. A management system for a wire-controlled MES system according to claim 6, characterized in that: The specific process of introducing the buffer capacity adjustment mechanism is as follows: A physical buffer is set up before and after each key workstation, and the release rhythm of upstream tasks is dynamically adjusted according to the current buffer queue length. If the buffer is close to full, the priority of the task before the process is lowered; otherwise, the release of tasks is accelerated.

8. A management system for a wire-controlled MES system according to claim 7, characterized in that: The process of dynamically adjusting the transport path is as follows: The global task is broken down into multiple subtasks, and the subtasks are assigned to the optimal AGV route through a distributed algorithm; the AGV bids for the subtasks based on its own status and prioritizes the optimal path.

9. The management system for a wire-controlled MES system according to claim 1, characterized in that: The process of predicting potential failures and triggering preventive maintenance work orders is as follows: The edge computing node receives sensor data in real time and inputs it into the LSTM model in a sliding window format. The sliding window length generates a prediction sequence, outputs the fault category, sets the fault probability threshold based on historical data, and dynamically adjusts the threshold. The prediction results are pushed to the MES system through the API interface to automatically generate maintenance work orders. The Bi-LSTM+Attention mechanism is used to capture long-term dependencies in the production process, identify key anomalies with attention weights, and regularly update the model.

10. A management method for a wire-controlled MES system, characterized in that: The steps include: Step 1: Collect production line operation data in real time. The operation data includes sensor data, production line data, and process parameter data. Step 2: Deploy lightweight algorithms at edge nodes to process operational data and upload it to the cloud for in-depth analysis, including trend analysis, bottleneck identification, and predictive modeling. Combined with the in-depth analysis results, the physical production environment is mapped in real time to simulate responses to different scheduling strategies. Step 3: Based on production line data, a reinforcement learning algorithm is used to dynamically adjust task priorities. A buffer capacity adjustment mechanism is also introduced to cope with load fluctuations. The system receives process constraint information predicted by the MES system and combines it with a distributed path planning strategy to dynamically adjust transportation routes. Step 4: Analyze historical equipment data based on the LSTM neural network, predict potential failures, trigger preventive maintenance work orders, and monitor production quality in real time.

Images