Steel logistics whole-process real-time dynamic management method and device

Abstract

The invention relates to a steel logistics full-process real-time dynamic management method and device, and belongs to the technical field of steel logistics control methods and devices. According to the technical scheme of the invention, data acquisition and monitoring are carried out, and transport vehicle and cargo states are tracked; demand prediction and inventory dynamic adjustment are carried out, and high-frequency cargo response emergency orders are allocated according to shipment frequency; visual management of the supply chain is carried out, shared data of all links of the supply chain is established, a 3D logistics map is constructed, and inventory, transportation states and bottleneck nodes are displayed in real time; and risk control management: quickly starting the alternative scheme when the early warning is triggered. The method has the advantages that efficiency is optimized, limitation of a traditional single transportation mode is reduced, rolling updating and emergency response of demand prediction are achieved, data transparency and credibility are improved, rapid risk response is achieved, and logistics full-life-cycle management is achieved.

Description

Technical Field

[0001] The present invention relates to a method and device for real-time dynamic management of the entire process of steel logistics, belonging to the technical field of steel logistics control methods and devices. Background Art

[0002] Steel logistics uses "steel" as its carrier, "logistics" as its operation, and "information" as its core. It integrates steel trade, e-commerce, and third-party logistics. Capital flow, information flow, and logistics flow promote and integrate each other, covering the intersection of the four major industries of construction, metallurgy, information industry, and modern logistics.

[0003] A search revealed that a method and device for real-time dynamic management of the entire steel logistics process is disclosed in the invention patent with Chinese patent publication number CN118627991A. The method for dynamic management of steel logistics in this invention patent acquires and preprocesses historical steel logistics data, constructs a digital twin model to improve prediction accuracy, and updates the digital twin model with real-time steel logistics data, thereby ensuring the real-time nature of the digital twin model. Finally, based on the updated digital twin model and real-time steel logistics data, the method identifies risks and optimizes routes for steel logistics, thereby improving steel logistics efficiency and achieving real-time dynamic management of the entire steel logistics process. However, this dynamic management method for steel logistics only analyzes logistics data by inputting it into the digital twin model. There are certain one-sidedness and limitations in data analysis and processing. It cannot comprehensively consider the impact of factors such as cost, timeliness and carbon emissions on logistics and transportation data. In addition, its process management cannot cover multiple links such as transportation vehicles, warehousing environment and production connection, and it cannot realize supply chain visualization and refined risk control operations. Therefore, a real-time dynamic management method and device for the entire process of steel logistics are proposed. Summary of the Invention

[0004] The purpose of the present invention is to provide a real-time dynamic management method and device for the entire process of steel logistics, dynamically adjust the path strategy through reinforcement learning, design fitness functions and parallel computing acceleration modules to further optimize efficiency, reduce the limitations of the traditional single transportation mode by integrating transportation resources under different modes, and realize rolling updates of demand forecasts and emergency responses by combining weather data influence coefficients. By scanning codes to query full-process information, data transparency and trust are further improved, a dynamic early warning system is constructed to identify risk data in transportation, warehousing and supply chain links, and contingency plan templates and path replanning algorithms are designed to achieve rapid risk response. It has the advantage of realizing full life cycle management of logistics and effectively solves the above-mentioned problems existing in the background technology.

[0005] The technical solution of the present invention is: a real-time dynamic management method for the entire process of steel logistics, comprising the following steps: S1. Data collection and monitoring: Using GPS and RFID technology to track the status of transport vehicles and cargo, and using sensor modules to monitor equipment status in real time and provide fault warnings. S2. Intelligent scheduling and route optimization: combining algorithms to optimize transportation routes and integrating multi-mode transportation data to dynamically determine the optimal combination; S3. Demand forecasting and dynamic inventory adjustment: predicting future demand fluctuations to adjust safety stock thresholds, automatically allocating high-frequency goods to meet urgent orders based on shipping frequency, and achieving dynamic inventory allocation. S4. Visualize supply chain management, establish shared data across all supply chain links, ensure data transparency and traceability, and build a 3D logistics map based on digital twin technology to display inventory, transportation status, and bottleneck nodes in real time. S5. Risk control management: set key indicators to build a risk warning model, and quickly launch alternative plans based on model predictions when a warning is triggered.

[0006] The specific implementation steps of step S1 are as follows: L1. Data collection and deployment 1) Transport Monitoring: a. Track the location of transport vehicles in real time using their GPS locators, recording their travel paths and dwell times. b. Use load sensors to record changes in vehicle weight in real time. c. Select the appropriate RFID tag type based on the logistics scenario, assign a unique EPC to each item or package, and automatically scan incoming and outgoing items through RFID readers for rapid traceability. 2) Warehouse monitoring: a. Detecting abnormal heat generation in inventory stacks using infrared thermal imaging cameras; b. Monitoring the operating status of logistics vehicles within the warehouse using battery monitoring technology; 3) Production-logistics connection monitoring: Complete data connection with the steel mill through the MES system interface, and automatically obtain steel production batch, specification and quality inspection result data; L2. Network transmission architecture: 1) Transport vehicles and mobile devices transmit real-time data via cellular networks, and LoRaWAN is used to enable long-distance and low-power transmission of sensor data within the warehouse; 2) Deploy edge servers in the factory based on edge computing nodes to perform local pre-processing of vibration, temperature and other data for uploading abnormal data; L3. Data Integration and Visualization: 1) By integrating order data ERP, transportation data TMS, and inventory data WMS systems, using API tools to synchronize system data in real time, standardize data formats, and build a data platform; 2) Build a GIS map to display the real-time location of vehicles in transit, build a 3D digital twin warehouse to display shelf storage status in real time, and use the equipment health dashboard to dynamically update KPI data such as crane failure rate and forklift utilization; 3) Set alarm thresholds and push warning information in real time via a third-party platform based on the warning status; L4. Data security backup: Blockchain notarization of key data, storage of historical data for 7 days on the local server, storage of full historical data in the cloud, and data cached on the vehicle device in the event of network disconnection, automatically retransmitting data after network recovery.

[0007] The specific implementation steps of step S2 are as follows: L1. Multi-source data fusion processing: 1) Acquire multi-source data based on data collection deployment and divide the multi-source data into real-time data and business data. Real-time data includes vehicle GPS location, traffic congestion index API, weather warning data, and warehouse loading and unloading progress. Business data includes order urgency, steel specifications and types, and customer time windows. 2) Processing abnormal data based on GPS positioning drift, specifically: a. Import data into pandas and matplotlib databases for data processing and visualization, respectively; b. Load GPS data from a CSV file and remove invalid data containing null values; c. Calculate drift point criteria based on distance and time difference, and set distance and time thresholds; d. Remove drift points that exceed the set threshold range and perform visual comparison; L2. Dynamic Path Optimization: 1) Set transportation cost and timeliness as the objective function and build priority rules, including: a. For urgent orders, set the timeliness weight to 70% and the cost weight to 30%; b. For regular orders, set the cost weight to 60%, the timeliness weight to 30%, and the carbon emission weight to 10%; 2) Constructing constraints includes: a. Setting vehicle load and cargo stacking limits as physical constraints; b. Setting environmentally friendly restricted areas and loading and unloading time windows as policy constraints; 3) Multi-objective genetic algorithm design: a. Sequential encoding is used to represent the path sequence of a vehicle on each chromosome. Vehicle type and loading method are added as additional genes. A penalty term is set to significantly reduce fitness when the constraint is violated. The fitness function is expressed as: Fitness = α*(cost / baseline cost) + β*(time / baseline time) + γ*(empty rate / baseline empty rate), where α+β+γ=1; b. Use sequential crossover while retaining parent path segments, randomly swap two client nodes or insert new nodes for mutation, define the state space and design the dynamic space, and set the reward function for dynamic adjustment of reinforcement learning; c. Access data input interfaces through external APIs and internal systems, set up event-driven mechanisms, perform hard constraint checks on load and time windows, and perform soft constraint optimization on detour cost trade-offs; d. Use Spark or GPU to accelerate the population iteration of the genetic algorithm, perform parallel computing optimization, output several groups of non-dominated solutions for Pareto optimal solution screening, and mark the recommended paths, risk points, and alternative route comparisons on the GIS map for visualization output; L3. Intermodal transport collaborative optimization: This integrates road, rail, and waterway transport data to compare costs across transport modes, provides advance carriage reservation information, and uses automated guided vehicles (AGVs) to automatically align train carriages with truck platforms, reducing loading and unloading wait times. Railway terminal resources are dispatched, and steel storage areas are dynamically allocated based on ship arrival times, enabling port yard resource management and optimized node connectivity. L4. Vehicle Resource Scheduling: a. Build a hierarchical resource library, prioritize self-operated fleets for high-value-added orders, automatically merge less-than-truckload orders heading in the same direction for intelligent dispatch, and achieve flexible capacity pool management; b. Optimize driver behavior based on a driving rating system. L5. Visual Decision Support: 1) Digital Twin Simulation: After inputting orders for a future time period, the decision-making system simulates vehicle movement across the network, predicts bottlenecks, and automatically generates comparison plans. It then compiles historical transportation data to generate a map of frequently delayed sections, guiding long-term route planning and conducting heat map analysis. 2) Collaborative decision-making: The algorithm is adjusted to recommend an optimization plan. The system automatically re-optimizes the remaining tasks based on the optimization plan. The system also marks defaulted orders with a red alert and provides compensation plans.

[0008] The specific implementation steps of step S3 are as follows: L1. Data integration and feature engineering: a. Divide the data input scope according to data type, including historical sales data, market dynamics data, customer order data, production plan data, and supply chain data; b. Count historical sales data using a sliding window and mark holidays to obtain time series features. The infrastructure policy text is run through the BERT model to extract semantic vectors, which serve as prediction inputs. Weather data is converted into impact coefficients to embed external features. L2. Demand forecast model construction: Model training is performed, with weekly updates of forecast data for the next 13 weeks. The latest market data is incorporated each time to achieve rolling forecasts. Detection of abnormal signals triggers model retraining, and regional demand forecasts are adjusted to respond to emergencies. L3. Dynamic Inventory Optimization: Inventory is divided into tiers based on safety stock, dynamic replenishment points, and expired inventory. When inventory drops to ROP, production instructions are automatically sent to the MES system, triggering JIT replenishment. Based on geographic location and transportation cost optimization, AGVs are dispatched within the warehouse at night to move frequently shipped items to the loading area. Level 4. Automated Equipment Collaboration: Unmanned overhead cranes (AGVs) use laser scanning of RFID tags to automatically grab items based on outbound priorities and dynamically adjust stacking plans. Based on real-time order heat maps, AGVs are dynamically assigned task paths, enabling cluster scheduling. AGVs automatically queue for charging during off-peak hours, optimizing charging strategies to ensure full daytime operation. L5. Digital twin monitoring: Real-time display of storage capacity rate in each area, detailed information of different steel models, and setting of abnormal warning dashboards.

[0009] The specific implementation steps of step S4 are as follows: L1. Data Sharing Platform: 1) Data interface standardization: API and electronic data interchange (EDI) are used to unify the data formats of steel mill MES, logistics provider TMS, warehouse WMS, and customer CRM systems, and to standardize the coding of steel product categories and timestamps; 2) Blockchain Data Storage: Build a Hyperledger Fabric consortium chain, set up steel mills, logistics providers, and quality inspection agencies as nodes, and use hashing of key fields in purchase contracts, quality inspection reports, and transport receipts to store them on the chain to prevent data tampering. Customers can scan codes to trace steel production batches and logistics tracks, while logistics providers only have access to relevant waybills. 3) Cloud-edge collaborative architecture: By deploying edge servers in steel mills and warehouses, we pre-process real-time data and integrate data using Alibaba Cloud DataWorks, supporting petabyte-level storage and real-time analysis. L2. Visual monitoring system: 1) 3D digital twin modeling: A laser scanner is used to generate a high-precision warehouse model, which is then processed and mapped to the location of steel coils in real time. Vehicle GPS is connected to the model to display vehicle location, speed, temperature and humidity, and weight change data in real time on a GIS map. 2) Multi-layer data dashboard: Build a global overview of the supply chain, display real-time inventory in transit and capacity utilization data, create risk heat maps, and mark frequently delayed sections based on historical data; L3. Early Warning Decision-Making: Configure a rules engine to trigger an inventory warning when the safety stock level falls below 80%. It provides replenishment suggestions, recommends suppliers and delivery times, predicts transportation risks during catastrophic weather conditions based on weather conditions, automatically pushes rerouting plans, and implements automated collaborative process management.

[0010] The specific implementation steps of step S5 are as follows: L1. Risk identification and classification: 1) Identify risks across the entire steel logistics chain and scan and address risk sources, including: a. Transportation: Disaster weather causes road interruptions, vehicle breakdowns, and loose and shifting cargo.

[0011] b. Warehousing: Fire, shelf collisions, and excessive moisture in warehouses can lead to steel corrosion. c. Supply chain: iron ore supplier disruptions, port strikes, and policy changes; 2) Prioritize based on probability of occurrence and impact, set the evaluation criteria on a scale of 1-5, and build a risk knowledge base; L2. Monitoring and early warning system: 1) Sensor network deployment: Multi-parameter sensors are installed on transport vehicles and thresholds are set. If the temperature is >50°C, a temperature threshold alarm is triggered. If three consecutive sudden brakings occur per minute, the current driving behavior is determined to be high-risk: Smoke sensors and infrared thermal imagers are installed in the warehouse environment. If the air particle concentration is greater than 5% LEL, a smoke concentration alarm is triggered. If the local temperature rise is greater than 10°C / h, a fire alarm is triggered. 2) External Data Access: Access the Central Meteorological Observatory API to obtain real-time typhoon paths and rainfall forecasts. Use NLP to crawl government websites and identify keywords to automatically generate risk alerts. 3) Configure the warning rule engine to process warnings in a graded manner, set trigger conditions, and push warning information through a third-party platform; L3. Dynamic Decision-Making: 1) Design a contingency plan template to handle transport disruptions by deploying backup vehicles to pick up cargo before the disruption point and initiating multimodal transport. At the same time, notify the customer and negotiate a compensation plan. 2) Build a route replanning algorithm. By inputting the coordinates of the interruption point, the location of available vehicles, and the customer's time window, it outputs three alternative routes based on cost, timeliness, and risk comparison. Using a greedy algorithm, it quickly matches the nearest available resources to achieve resource scheduling optimization.

[0012] A real-time dynamic management device for the entire process of steel logistics includes a data acquisition module, a route optimization module, a demand forecasting module, a visualization management module, and a risk control decision module. The four modules are connected in sequence. The data acquisition module is used to acquire multi-source data and perform data fusion processing; the route optimization module is used to optimize the transportation route and provide the optimal route combination decision; the demand forecasting module is used to predict demand fluctuations and adjust the safety stock according to the fluctuation range to achieve dynamic allocation; the visualization management module is used to visualize the data of the entire supply chain; the risk control decision module monitors risk data according to a risk warning model and provides decision support.

[0013] The beneficial effects of the present invention are: dynamically adjusting the path strategy through reinforcement learning, designing the fitness function and parallel computing acceleration module to further optimize the efficiency, reducing the limitations of the traditional single transportation mode by integrating transportation resources under different modes, and realizing rolling updates of demand forecasts and emergency responses by combining the influence coefficient of weather data. By scanning the code to query the whole process information, the transparency and trust of the data are further improved, a dynamic early warning system is constructed to identify the risk data of transportation, warehousing and supply chain links, and contingency plan templates and path replanning algorithms are designed to achieve rapid response to risks. It has the advantage of realizing the management of the entire logistics life cycle. BRIEF DESCRIPTION OF THE DRAWINGS

[0014] Figure 1 It is a flow chart of the method of the present invention. DETAILED DESCRIPTION

[0015] In order to make the purpose, technical solutions and advantages of the invention implementation cases clearer, the technical solutions in the invention implementation cases will be clearly and completely described below in conjunction with the drawings in the implementation cases. Obviously, the implementation cases described are only a small part of the implementation cases of the present invention, rather than all the implementation cases. Based on the implementation cases in the present invention, all other implementation cases obtained by ordinary technicians in this field without making creative work are within the scope of protection of the present invention.

[0016] A real-time dynamic management method for the entire process of steel logistics, comprising the following steps: S1. Data collection and monitoring: Using GPS and RFID technology to track the status of transport vehicles and cargo, and using sensor modules to monitor equipment status in real time and provide fault warnings. S2. Intelligent scheduling and route optimization: combining algorithms to optimize transportation routes and integrating multi-mode transportation data to dynamically determine the optimal combination; S3. Demand forecasting and dynamic inventory adjustment: predicting future demand fluctuations to adjust safety stock thresholds, automatically allocating high-frequency goods to meet urgent orders based on shipping frequency, and achieving dynamic inventory allocation. S4. Visualize supply chain management, establish shared data across all supply chain links, ensure data transparency and traceability, and build a 3D logistics map based on digital twin technology to display inventory, transportation status, and bottleneck nodes in real time. S5. Risk control management: set key indicators to build a risk warning model, and quickly launch alternative plans based on model predictions when a warning is triggered.

[0017] The specific implementation steps of step S1 are as follows: L1. Data collection and deployment 1) Transport Monitoring: a. Track the location of transport vehicles in real time using their GPS locators, recording their travel paths and dwell times. b. Use load sensors to record changes in vehicle weight in real time. c. Select the appropriate RFID tag type based on the logistics scenario, assign a unique EPC to each item or package, and automatically scan incoming and outgoing items through RFID readers for rapid traceability. 2) Warehouse monitoring: a. Detecting abnormal heat generation in inventory stacks using infrared thermal imaging cameras; b. Monitoring the operating status of logistics vehicles within the warehouse using battery monitoring technology; 3) Production-logistics connection monitoring: Complete data connection with the steel mill through the MES system interface, and automatically obtain steel production batch, specification and quality inspection result data; L2. Network transmission architecture: 1) Transport vehicles and mobile devices transmit real-time data via cellular networks, and LoRaWAN is used to enable long-distance and low-power transmission of sensor data within the warehouse; 2) Deploy edge servers in the factory based on edge computing nodes to perform local pre-processing of vibration, temperature and other data for uploading abnormal data; L3. Data Integration and Visualization: 1) By integrating order data ERP, transportation data TMS, and inventory data WMS systems, using API tools to synchronize system data in real time, standardize data formats, and build a data platform; 2) Build a GIS map to display the real-time location of vehicles in transit, build a 3D digital twin warehouse to display shelf storage status in real time, and use the equipment health dashboard to dynamically update KPI data such as crane failure rate and forklift utilization; 3) Set alarm thresholds and push warning information in real time via a third-party platform based on the warning status; L4. Data security backup: Blockchain notarization of key data, storage of historical data for 7 days on the local server, storage of full historical data in the cloud, and data cached on the vehicle device in the event of network disconnection, automatically retransmitting data after network recovery.

[0018] The specific implementation steps of step S2 are as follows: L1. Multi-source data fusion processing: 1) Acquire multi-source data based on data collection deployment and divide the multi-source data into real-time data and business data. Real-time data includes vehicle GPS location, traffic congestion index API, weather warning data, and warehouse loading and unloading progress. Business data includes order urgency, steel specifications and types, and customer time windows. 2) Processing abnormal data based on GPS positioning drift, specifically: a. Import data into pandas and matplotlib databases for data processing and visualization, respectively; b. Load GPS data from a CSV file and remove invalid data containing null values; c. Calculate drift point criteria based on distance and time difference, and set distance and time thresholds; d. Remove drift points that exceed the set threshold range and perform visual comparison; L2. Dynamic Path Optimization: 1) Set transportation cost and timeliness as the objective function and build priority rules, including: a. For urgent orders, set the timeliness weight to 70% and the cost weight to 30%; b. For regular orders, set the cost weight to 60%, the timeliness weight to 30%, and the carbon emission weight to 10%; 2) Constructing constraints includes: a. Setting vehicle load and cargo stacking limits as physical constraints; b. Setting environmentally friendly restricted areas and loading and unloading time windows as policy constraints; 3) Multi-objective genetic algorithm design: a. Sequential encoding is used to represent the path sequence of a vehicle on each chromosome. Vehicle type and loading method are added as additional genes. A penalty term is set to significantly reduce fitness when the constraint is violated. The fitness function is expressed as: Fitness = α*(cost / baseline cost) + β*(time / baseline time) + γ*(empty rate / baseline empty rate), where α+β+γ=1; b. Use sequential crossover while retaining parent path segments, randomly swap two client nodes or insert new nodes for mutation, define the state space and design the dynamic space, and set the reward function for dynamic adjustment of reinforcement learning; c. Access data input interfaces through external APIs and internal systems, set up event-driven mechanisms, perform hard constraint checks on load and time windows, and perform soft constraint optimization on detour cost trade-offs; d. Use Spark or GPU to accelerate the population iteration of the genetic algorithm, perform parallel computing optimization, output several groups of non-dominated solutions for Pareto optimal solution screening, and mark the recommended paths, risk points, and alternative route comparisons on the GIS map for visualization output; L3. Intermodal transport collaborative optimization: This integrates road, rail, and waterway transport data to compare costs across transport modes, provides advance carriage reservation information, and uses automated guided vehicles (AGVs) to automatically align train carriages with truck platforms, reducing loading and unloading wait times. Railway terminal resources are dispatched, and steel storage areas are dynamically allocated based on ship arrival times, enabling port yard resource management and optimized node connectivity. L4. Vehicle Resource Scheduling: a. Build a hierarchical resource library, prioritize self-operated fleets for high-value-added orders, automatically merge less-than-truckload orders heading in the same direction for intelligent dispatch, and achieve flexible capacity pool management; b. Optimize driver behavior based on a driving rating system. L5. Visual Decision Support: 1) Digital Twin Simulation: After inputting orders for a future time period, the decision-making system simulates vehicle movement across the network, predicts bottlenecks, and automatically generates comparison plans. It then compiles historical transportation data to generate a map of frequently delayed sections, guiding long-term route planning and conducting heat map analysis. 2) Collaborative decision-making: The algorithm is adjusted to recommend an optimization plan. The system automatically re-optimizes the remaining tasks based on the optimization plan. The system also marks defaulted orders with a red alert and provides compensation plans.

[0019] The specific implementation steps of step S3 are as follows: L1. Data integration and feature engineering: a. Divide the data input scope according to data type, including historical sales data, market dynamics data, customer order data, production plan data, and supply chain data; b. Count historical sales data using a sliding window and mark holidays to obtain time series features. The infrastructure policy text is run through the BERT model to extract semantic vectors, which serve as prediction inputs. Weather data is converted into impact coefficients to embed external features. L2. Demand forecast model construction: Model training is performed, with weekly updates of forecast data for the next 13 weeks. The latest market data is incorporated each time to achieve rolling forecasts. Detection of abnormal signals triggers model retraining, and regional demand forecasts are adjusted to respond to emergencies. L3. Dynamic Inventory Optimization: Inventory is divided into tiers based on safety stock, dynamic replenishment points, and expired inventory. When inventory drops to ROP, production instructions are automatically sent to the MES system, triggering JIT replenishment. Based on geographic location and transportation cost optimization, AGVs are dispatched within the warehouse at night to move frequently shipped items to the loading area. Level 4. Automated Equipment Collaboration: Unmanned overhead cranes (AGVs) use laser scanning of RFID tags to automatically grab items based on outbound priorities and dynamically adjust stacking plans. Based on real-time order heat maps, AGVs are dynamically assigned task paths, enabling cluster scheduling. AGVs automatically queue for charging during off-peak hours, optimizing charging strategies to ensure full daytime operation. L5. Digital twin monitoring: Real-time display of storage capacity rate in each area, detailed information of different steel models, and setting of abnormal warning dashboards.

[0020] The specific implementation steps of step S4 are as follows: L1. Data Sharing Platform: 1) Data interface standardization: API and electronic data interchange (EDI) are used to unify the data formats of steel mill MES, logistics provider TMS, warehouse WMS, and customer CRM systems, and to standardize the coding of steel product categories and timestamps; 2) Blockchain Data Storage: Build a Hyperledger Fabric consortium chain, set up steel mills, logistics providers, and quality inspection agencies as nodes, and use hashing of key fields in purchase contracts, quality inspection reports, and transport receipts to store them on the chain to prevent data tampering. Customers can scan codes to trace steel production batches and logistics tracks, while logistics providers only have access to relevant waybills. 3) Cloud-edge collaborative architecture: By deploying edge servers in steel mills and warehouses, we pre-process real-time data and integrate data using Alibaba Cloud DataWorks, supporting petabyte-level storage and real-time analysis. L2. Visual monitoring system: 1) 3D digital twin modeling: A laser scanner is used to generate a high-precision warehouse model, which is then processed and mapped to the location of steel coils in real time. Vehicle GPS is connected to the model to display vehicle location, speed, temperature and humidity, and weight change data in real time on a GIS map. 2) Multi-layer data dashboard: Build a global overview of the supply chain, display real-time inventory in transit and capacity utilization data, create risk heat maps, and mark frequently delayed sections based on historical data; L3. Early Warning Decision-Making: Configure a rules engine to trigger an inventory warning when the safety stock level falls below 80%. It provides replenishment suggestions, recommends suppliers and delivery times, predicts transportation risks during catastrophic weather conditions based on weather conditions, automatically pushes rerouting plans, and implements automated collaborative process management.

[0021] The specific implementation steps of step S5 are as follows: L1. Risk identification and classification: 1) Identify risks across the entire steel logistics chain and scan and address risk sources, including: a. Transportation: Disaster weather causes road interruptions, vehicle breakdowns, and loose and shifting cargo.

[0022] b. Warehousing: Fire, shelf collisions, and excessive moisture in warehouses can lead to steel corrosion. c. Supply chain: iron ore supplier disruptions, port strikes, and policy changes; 2) Prioritize based on probability of occurrence and impact, set the evaluation criteria on a scale of 1-5, and build a risk knowledge base; L2. Monitoring and early warning system: 1) Sensor network deployment: Multi-parameter sensors are installed on transport vehicles and thresholds are set. If the temperature is >50°C, a temperature threshold alarm is triggered. If three consecutive sudden brakings occur per minute, the current driving behavior is determined to be high-risk: Smoke sensors and infrared thermal imagers are installed in the warehouse environment. If the air particle concentration is greater than 5% LEL, a smoke concentration alarm is triggered. If the local temperature rise is greater than 10°C / h, a fire alarm is triggered. 2) External Data Access: Access the Central Meteorological Observatory API to obtain real-time typhoon paths and rainfall forecasts. Use NLP to crawl government websites and identify keywords to automatically generate risk alerts. 3) Configure the warning rule engine to process warnings in a graded manner, set trigger conditions, and push warning information through a third-party platform; L3. Dynamic Decision-Making: 1) Design a contingency plan template to handle transport disruptions by deploying backup vehicles to pick up cargo before the disruption point and initiating multimodal transport. At the same time, notify the customer and negotiate a compensation plan. 2) Build a route replanning algorithm. By inputting the coordinates of the interruption point, the location of available vehicles, and the customer's time window, it outputs three alternative routes based on cost, timeliness, and risk comparison. Using a greedy algorithm, it quickly matches the nearest available resources to achieve resource scheduling optimization.

[0023] A real-time dynamic management device for the entire process of steel logistics includes a data acquisition module, a route optimization module, a demand forecasting module, a visualization management module, and a risk control decision module. The four modules are connected in sequence. The data acquisition module is used to acquire multi-source data and perform data fusion processing; the route optimization module is used to optimize the transportation route and provide the optimal route combination decision; the demand forecasting module is used to predict demand fluctuations and adjust the safety stock according to the fluctuation range to achieve dynamic allocation; the visualization management module is used to visualize the data of the entire supply chain; the risk control decision module monitors risk data according to a risk warning model and provides decision support. Example

[0024] The specific steps of data collection and monitoring in step S1 include: L1. Data collection and deployment: 1) Transport Monitoring: a. Track the location of transport vehicles in real time using their GPS locators, recording their travel paths and dwell times. b. Use load sensors to record changes in vehicle weight in real time. c. Select the appropriate RFID tag type based on the logistics scenario, assign a unique EPC to each item or package, and automatically scan incoming and outgoing items through RFID readers for rapid traceability. 2) Warehouse monitoring: a. Detecting abnormal heat generation in inventory stacks using infrared thermal imaging cameras; b. Monitoring the operating status of logistics vehicles within the warehouse using battery monitoring technology; 3) Production-logistics connection monitoring: Complete data connection with the steel mill through the MES system interface, and automatically obtain steel production batch, specification and quality inspection result data; L2. Network transmission architecture: 1) Transport vehicles and mobile devices transmit real-time data via cellular networks, and LoRaWAN is used to enable long-distance and low-power transmission of sensor data within the warehouse; 2) Deploy edge servers in the factory based on edge computing nodes to perform local pre-processing of vibration, temperature and other data for uploading abnormal data; L3. Data integration and visualization: 1) By integrating order data ERP, transportation data TMS, and inventory data WMS systems, using API tools to synchronize system data in real time, standardize data formats, and build a data platform; 2) Build a GIS map to display the real-time location of vehicles in transit, build a 3D digital twin warehouse to display shelf storage status in real time, and use the equipment health dashboard to dynamically update KPI data such as crane failure rate and forklift utilization; 3) Set alarm thresholds and push warning information in real time via a third-party platform based on the warning status; L4. Data security backup: Blockchain notarization of key data, storage of historical data for 7 days on the local server, storage of full historical data in the cloud, and data cached on the vehicle device in the event of network disconnection, automatically retransmitting data after network recovery.

[0025] Specifically, different data collection methods are used for different links, and multi-dimensional data such as location, weight, status, and production batches are covered. This can fully reflect the actual situation of each link in the logistics process and provide a rich data foundation for subsequent analysis and decision-making. Selecting the appropriate transmission method (cellular network and LoRaWAN) based on different scenarios (transportation and warehouse) can balance the real-time performance, low power consumption, and long-distance requirements of data transmission. At the same time, the use of edge computing nodes can improve data processing efficiency. Integrate data from multiple systems for real-time synchronization and standardization, build a data platform, and facilitate unified data management and analysis. Combined with visualization tools such as GIS maps, 3D digital twin warehouses, and equipment health dashboards, it can intuitively display the status and key indicators of each logistics link, while also setting alarm thresholds and delivering early warning information in real time. The use of blockchain evidence can ensure the security and non-tamperability of key data. The storage method that combines local and cloud storage can not only meet the needs of fast access to recent data, but also preserve the full amount of data for a long time; the offline cache and automatic re-upload functions ensure the integrity and continuity of the data.

[0026] The specific steps of intelligent scheduling and path optimization in step S2 include: L1. Multi-source data fusion processing: 1) Acquire multi-source data based on data collection deployment and divide the multi-source data into real-time data and business data. Real-time data includes vehicle GPS location, traffic congestion index API, weather warning data, and warehouse loading and unloading progress. Business data includes order urgency, steel specifications and types, and customer time windows. 2) Processing abnormal data based on GPS positioning drift, specifically: a. Import data into pandas and matplotlib databases for data processing and visualization, respectively; b. Load GPS data from a CSV file and remove invalid data containing null values; c. Calculate drift point criteria based on distance and time difference, and set distance and time thresholds; d. Remove drift points that exceed the set threshold range and perform visual comparison; L2. Dynamic Path Optimization: 1) Set transportation cost and timeliness as the objective function and build priority rules, including: a. For urgent orders, set the timeliness weight to 70% and the cost weight to 30%; b. For regular orders, set the cost weight to 60%, the timeliness weight to 30%, and the carbon emission weight to 10%; 2) Constructing constraints includes: a. Setting vehicle load and cargo stacking limits as physical constraints; b. Setting environmentally friendly restricted areas and loading and unloading time windows as policy constraints; 3) Multi-objective genetic algorithm design: a. Sequential encoding is used to represent the path sequence of a vehicle on each chromosome. Vehicle type and loading method are added as additional genes. A penalty term is set to significantly reduce fitness when the constraint is violated. The fitness function is expressed as: Fitness = α*(cost / baseline cost) + β*(time / baseline time) + γ*(empty rate / baseline empty rate), where α+β+γ=1; b. Use sequential crossover while retaining parent path segments, randomly swap two client nodes or insert new nodes for mutation, define the state space and design the dynamic space, and set the reward function for dynamic adjustment of reinforcement learning; c. Access data input interfaces through external APIs and internal systems, set up event-driven mechanisms, perform hard constraint checks on load and time windows, and perform soft constraint optimization on detour cost trade-offs; d. Use Spark or GPU to accelerate the population iteration of the genetic algorithm, perform parallel computing optimization, output several groups of non-dominated solutions for Pareto optimal solution screening, and mark the recommended paths, risk points, and alternative route comparisons on the GIS map for visualization output; L3. Intermodal transport collaborative optimization: This integrates road, rail, and waterway transport data to compare costs across transport modes, provides advance carriage reservation information, and uses automated guided vehicles (AGVs) to automatically align train carriages with truck platforms, reducing loading and unloading wait times. Railway terminal resources are dispatched, and steel storage areas are dynamically allocated based on ship arrival times, enabling port yard resource management and optimized node connectivity. L4. Vehicle Resource Scheduling: a. Build a hierarchical resource library, prioritize self-operated fleets for high-value-added orders, automatically merge less-than-truckload orders heading in the same direction for intelligent dispatch, and achieve flexible capacity pool management; b. Optimize driver behavior based on a driving rating system. L5. Visual Decision Support: 1) Digital Twin Simulation: After inputting orders for a future time period, the decision-making system simulates vehicle movement across the network, predicts bottlenecks, and automatically generates comparison plans. It then compiles historical transportation data to generate a map of frequently delayed sections, guiding long-term route planning and conducting heat map analysis. 2) Collaborative decision-making: The algorithm is adjusted to recommend an optimization plan. The system automatically re-optimizes the remaining tasks based on the optimization plan. The system also marks defaulted orders with a red alert and provides compensation plans.

[0027] Specifically, we set objective function weights for different types of orders, build physical and policy constraints, and make path optimization more aligned with actual business needs and real-world limitations. We use a multi-objective genetic algorithm combined with data interfaces and event-driven mechanisms to improve algorithm efficiency and optimization results, while also outputting high-quality path solutions. Integrate data from multiple modes of transportation, optimize node connections, reduce loading and unloading waiting time, achieve effective resource management and allocation, improve transportation efficiency and reduce costs, and realize visual decision support through digital twin simulation and collaborative decision-making, thereby improving the scientific nature and accuracy of decision-making.

[0028] The specific steps of demand forecasting and inventory dynamic adjustment in step S3 include: L1. Data integration and feature engineering: a. Divide the data input scope according to data type, including historical sales data, market dynamics data, customer order data, production plan data, and supply chain data; b. Count historical sales data using a sliding window and mark holidays to obtain time series features. The infrastructure policy text is run through the BERT model to extract semantic vectors, which serve as prediction inputs. Weather data is converted into impact coefficients to embed external features. L2. Demand forecast model construction: Model training is performed, with weekly updates of forecast data for the next 13 weeks. The latest market data is incorporated each time to achieve rolling forecasts. Detection of abnormal signals triggers model retraining, and regional demand forecasts are adjusted to respond to emergencies. L3. Dynamic Inventory Optimization: Inventory is divided into tiers based on safety stock, dynamic replenishment points, and expired inventory. When inventory drops to ROP, production instructions are automatically sent to the MES system, triggering JIT replenishment. Based on geographic location and transportation cost optimization, AGVs are dispatched within the warehouse at night to move frequently shipped items to the loading area. Level 4. Automated Equipment Collaboration: Unmanned overhead cranes (AGVs) use laser scanning of RFID tags to automatically grab items based on outbound priorities and dynamically adjust stacking plans. Based on real-time order heat maps, AGVs are dynamically assigned task paths, enabling cluster scheduling. AGVs automatically queue for charging during off-peak hours, optimizing charging strategies to ensure full daytime operation. L5. Digital twin monitoring: Real-time display of storage capacity rate in each area, detailed information of different steel models, and setting of abnormal warning dashboards.

[0029] Specifically, a rolling forecast approach is adopted to continuously incorporate the latest market data. When abnormal signals are detected, the model can be retrained and the forecast value adjusted. This further improves the responsiveness of demand forecasts to market changes and emergencies. The system divides inventory into tiers and combines reorder points and just-in-time replenishment mechanisms to achieve dynamic inventory optimization. Unmanned overhead cranes and automated equipment such as AGVs can work collaboratively, and through reasonable task allocation and scheduling strategies, they can improve operational efficiency and automation levels. Based on digital twin monitoring, the storage capacity rate and detailed information of steel models in each area are displayed in real time, and abnormal warning dashboards are set up to provide visual support for inventory management decisions.

[0030] The specific steps of supply chain visualization management in step S4 include: L1. Data Sharing Platform: 1) Data interface standardization: API and electronic data interchange (EDI) are used to unify the data formats of steel mill MES, logistics provider TMS, warehouse WMS, and customer CRM systems, and to standardize the coding of steel product categories and timestamps; 2) Blockchain Data Storage: Build a Hyperledger Fabric consortium chain, set up steel mills, logistics providers, and quality inspection agencies as nodes, and use hashing of key fields in purchase contracts, quality inspection reports, and transport receipts to store them on the chain to prevent data tampering. Customers can scan codes to trace steel production batches and logistics tracks, while logistics providers only have access to relevant waybills. 3) Cloud-edge collaborative architecture: By deploying edge servers in steel mills and warehouses, we pre-process real-time data and integrate data using Alibaba Cloud DataWorks, supporting petabyte-level storage and real-time analysis. L2. Visual monitoring system: 1) 3D digital twin modeling: A laser scanner is used to generate a high-precision warehouse model, which is then processed and mapped to the location of steel coils in real time. Vehicle GPS is connected to the model to display vehicle location, speed, temperature and humidity, and weight change data in real time on a GIS map. 2) Multi-layer data dashboard: Build a global overview of the supply chain, display real-time inventory in transit and capacity utilization data, create risk heat maps, and mark frequently delayed sections based on historical data; L3. Early Warning Decision-Making: Configure a rules engine to trigger an inventory warning when the safety stock level falls below 80%. It provides replenishment suggestions, recommends suppliers and delivery times, predicts transportation risks during catastrophic weather conditions based on weather conditions, automatically pushes rerouting plans, and implements automated collaborative process management.

[0031] The specific steps of risk control management in step S5 include: L1. Risk identification and classification: 1) Identify risks across the entire steel logistics chain and scan and address risk sources, including: a. Transportation: Disaster weather causes road interruptions, vehicle breakdowns, and loose and shifting cargo.

[0032] b. Warehousing: Fire, shelf collisions, and excessive moisture in warehouses can lead to steel corrosion. c. Supply chain: iron ore supplier disruptions, port strikes, and policy changes; 2) Prioritize based on probability of occurrence and impact, set the evaluation criteria on a scale of 1-5, and build a risk knowledge base; L2. Monitoring and early warning system: 1) Sensor network deployment: Multi-parameter sensors are installed on transport vehicles and thresholds are set. If the temperature is >50°C, a temperature threshold alarm is triggered. If three consecutive sudden brakings occur per minute, the current driving behavior is determined to be high-risk: Smoke sensors and infrared thermal imagers are installed in the warehouse environment. If the air particle concentration is greater than 5% LEL, a smoke concentration alarm is triggered. If the local temperature rise is greater than 10°C / h, a fire alarm is triggered. 2) External Data Access: Access the Central Meteorological Observatory API to obtain real-time typhoon paths and rainfall forecasts. Use NLP to crawl government websites and identify keywords to automatically generate risk alerts. 3) Configure the warning rule engine to process warnings in a graded manner, set trigger conditions, and push warning information through a third-party platform; L3. Dynamic Decision-Making: 1) Design a contingency plan template to handle transport disruptions by deploying backup vehicles to pick up cargo before the disruption point and initiating multimodal transport. At the same time, notify the customer and negotiate a compensation plan. 2) Build a route replanning algorithm. By inputting the coordinates of the interruption point, the location of available vehicles, and the customer's time window, it outputs three alternative routes based on cost, timeliness, and risk comparison. Using a greedy algorithm, it quickly matches the nearest available resources to achieve resource scheduling optimization.

[0033] Specifically, by prioritizing risks based on probability of occurrence and impact and building a knowledge base, high-priority risks can be prioritized, improving the relevance and efficiency of risk management. By deploying a sensor network to monitor transport vehicles and warehouse environments in real time, and accessing external data to obtain weather and policy information, a comprehensive risk monitoring system can be established. Configuring an early warning rule engine to process early warnings in a hierarchical manner and push them through third-party platforms facilitates rapid response to early warning information. Building a path replanning algorithm and a greedy algorithm can quickly output alternative routes and match available resources based on actual conditions, achieving optimal resource scheduling and reducing risk losses, further improving the efficiency and stability of logistics operations.

[0034] The present invention integrates GPS and RFID technologies and combines multiple equipment monitoring sensors to cover multiple links such as transportation vehicles, storage environment and production connection, further refining the data collection granularity and improving data analysis accuracy; By comprehensively considering cost, timeliness, and carbon emissions, the system dynamically adjusts routing strategies through reinforcement learning, designs fitness functions and parallel computing acceleration modules to further optimize efficiency, and integrates transportation resources under different modes to achieve seamless cross-transportation integration, thus reducing the limitations of traditional single transportation modes. By combining the impact coefficient of weather data, we can achieve rolling updates of demand forecasts and emergency response. By scanning codes to query information throughout the entire process, we can further improve data transparency and trust. We can build a dynamic early warning system to identify risk data in transportation, warehousing, and supply chain links, and design contingency plan templates and path replanning algorithms to achieve rapid risk response.

[0035] While embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions, and variations may be made to these embodiments without departing from the principles and spirit of the invention, and that the scope of the invention is defined by the appended claims and their equivalents.

Claims

1. A real-time dynamic management method for the entire process of steel logistics, characterized in that The following steps are involved: S1. Data collection and monitoring: Using GPS and RFID technology to track the status of transport vehicles and cargo, and using sensor modules to monitor equipment status in real time and provide fault warnings. S2. Intelligent scheduling and route optimization: combining algorithms to optimize transportation routes and integrating multi-mode transportation data to dynamically determine the optimal combination; S3. Demand forecasting and dynamic inventory adjustment: predicting future demand fluctuations to adjust safety stock thresholds, automatically allocating high-frequency goods to meet urgent orders based on shipping frequency, and achieving dynamic inventory allocation. S4. Visualize supply chain management, establish shared data across all supply chain links, ensure data transparency and traceability, and build a 3D logistics map based on digital twin technology to display inventory, transportation status, and bottleneck nodes in real time. S5. Risk control management: set key indicators to build a risk warning model, and quickly launch alternative plans based on model predictions when a warning is triggered.

2. A method for real-time dynamic management of the entire steel logistics process according to claim 1, characterized in that: The specific implementation steps of step S1 are as follows: L1. Data collection and deployment 1) Transport monitoring: a. Track the location of transport vehicles in real time based on their GPS locators, and record their travel trajectories and dwell times; b. Real-time recording of transport vehicle weight changes based on load sensors; c. Select the appropriate RFID tag type based on the logistics scenario, assign a unique EPC to each item or package, and automatically scan incoming and outgoing items through RFID readers for rapid traceability; 2) Warehouse monitoring: a. Detecting abnormal heat generation in inventory stacks using infrared thermal imaging cameras; b. Monitoring the operating status of logistics vehicles within the warehouse using battery monitoring technology; 3) Production-logistics connection monitoring: Complete data connection with the steel mill through the MES system interface, and automatically obtain steel production batch, specification and quality inspection result data; L2. Network transmission architecture: 1) Transport vehicles and mobile devices transmit real-time data via cellular networks, and LoRaWAN is used to enable long-distance and low-power transmission of sensor data within the warehouse; 2) Deploy edge servers in the factory based on edge computing nodes to perform local pre-processing of vibration, temperature and other data for uploading abnormal data; L3. Data integration and visualization: 1) By integrating order data ERP, transportation data TMS, and inventory data WMS systems, using API tools to synchronize system data in real time, standardize data formats, and build a data platform; 2) Build a GIS map to display the real-time location of vehicles in transit, build a 3D digital twin warehouse to display shelf storage status in real time, and use the equipment health dashboard to dynamically update KPI data such as crane failure rate and forklift utilization; 3) Set alarm thresholds and push warning information in real time via a third-party platform based on the warning status; L4. Data security backup: Blockchain notarization of key data, storage of historical data for 7 days on the local server, storage of full historical data in the cloud, and data cached on the vehicle device in the event of network disconnection, automatically retransmitting data after network recovery.

3. The method for real-time dynamic management of the entire steel logistics process according to claim 1, characterized in that: The specific implementation steps of step S2 are as follows: L1. Multi-source data fusion processing: 1) Acquire multi-source data based on data collection deployment and divide the multi-source data into real-time data and business data. Real-time data includes vehicle GPS location, traffic congestion index API, weather warning data, and warehouse loading and unloading progress. Business data includes order urgency, steel specifications and types, and customer time windows. 2) Processing abnormal data based on GPS positioning drift, specifically: a. Import data into pandas and matplotlib databases for data processing and visualization, respectively; b. Load GPS data from a CSV file and remove invalid data containing null values; c. Calculate drift point criteria based on distance and time difference, and set distance and time thresholds; d. Remove drift points that exceed the set threshold range and perform visual comparison; L2. Dynamic Path Optimization: 1) Set transportation cost and timeliness as the objective function and build priority rules, including: a. For urgent orders, set the timeliness weight to 70% and the cost weight to 30%; b. For regular orders, set the cost weight to 60%, the timeliness weight to 30%, and the carbon emission weight to 10%; 2) Constructing constraints includes: a. Setting vehicle load and cargo stacking limits as physical constraints; b. Setting environmentally friendly restricted areas and loading and unloading time windows as policy constraints; 3) Multi-objective genetic algorithm design: a. Sequential encoding is used to represent the path sequence of a vehicle on each chromosome. Vehicle type and loading method are added as additional genes. A penalty term is set to significantly reduce fitness when the constraint is violated. The fitness function is expressed as: Fitness = α*(cost / baseline cost) + β*(time / baseline time) + γ*(empty rate / baseline empty rate), where α+β+γ=1; b. Use sequential crossover while retaining parent path segments, randomly swap two client nodes or insert new nodes for mutation, define the state space and design the dynamic space, and set the reward function for dynamic adjustment of reinforcement learning; c. Access data input interfaces through external APIs and internal systems, set up event-driven mechanisms, perform hard constraint checks on load and time windows, and perform soft constraint optimization on detour cost trade-offs; d. Use Spark or GPU to accelerate the population iteration of the genetic algorithm, perform parallel computing optimization, output several groups of non-dominated solutions for Pareto optimal solution screening, and mark the recommended paths, risk points, and alternative route comparisons on the GIS map for visualization output; L3. Intermodal transport collaborative optimization: This integrates road, rail, and waterway transport data to compare costs across transport modes, provides advance carriage reservation information, and uses automated guided vehicles (AGVs) to automatically align train carriages with truck platforms, reducing loading and unloading wait times. Railway terminal resources are dispatched, and steel storage areas are dynamically allocated based on ship arrival times, enabling port yard resource management and optimized node connectivity. L4. Vehicle Resource Scheduling: a. Build a hierarchical resource library, prioritize self-operated fleets for high-value-added orders, automatically merge less-than-truckload orders heading in the same direction for intelligent dispatch, and achieve flexible capacity pool management; b. Optimize driver behavior based on a driving rating system. L5. Visual Decision Support: 1) Digital Twin Simulation: After inputting orders for a future time period, the decision-making system simulates vehicle movement across the network, predicts bottlenecks, and automatically generates comparison plans. It then compiles historical transportation data to generate a map of frequently delayed sections, guiding long-term route planning and conducting heat map analysis. 2) Collaborative decision-making: The algorithm is adjusted to recommend an optimization plan. The system automatically re-optimizes the remaining tasks based on the optimization plan. The system also marks defaulted orders with a red alert and provides compensation plans.

4. The method for real-time dynamic management of the entire steel logistics process according to claim 1, characterized in that: The specific implementation steps of step S3 are as follows: L1. Data integration and feature engineering: a. Divide the data input scope according to data type, including historical sales data, market dynamics data, customer order data, production plan data, and supply chain data; b. Count historical sales data using a sliding window and mark holidays to obtain time series features. The infrastructure policy text is run through the BERT model to extract semantic vectors, which serve as prediction inputs. Weather data is converted into impact coefficients to embed external features. L2. Demand forecast model construction: Model training is performed, with weekly updates of forecast data for the next 13 weeks. The latest market data is incorporated each time to achieve rolling forecasts. Detection of abnormal signals triggers model retraining, and regional demand forecasts are adjusted to respond to emergencies. L3. Dynamic Inventory Optimization: Inventory is divided into tiers based on safety stock, dynamic replenishment points, and expired inventory. When inventory drops to ROP, production instructions are automatically sent to the MES system, triggering JIT replenishment. Based on geographic location and transportation cost optimization, AGVs are dispatched within the warehouse at night to move frequently shipped items to the loading area. Level 4. Automated Equipment Collaboration: Unmanned overhead cranes (AGVs) use laser scanning of RFID tags to automatically grab items based on outbound priorities and dynamically adjust stacking plans. Based on real-time order heat maps, AGVs are dynamically assigned task paths, enabling cluster scheduling. AGVs automatically queue for charging during off-peak hours, optimizing charging strategies to ensure full daytime operation. L5. Digital twin monitoring: Real-time display of storage capacity rate in each area, detailed information of different steel models, and setting of abnormal warning dashboards.

5. The method for real-time dynamic management of the entire steel logistics process according to claim 1, characterized in that: The specific implementation steps of step S4 are as follows: L1. Data Sharing Platform: 1) Data interface standardization: API and electronic data interchange (EDI) are used to unify the data formats of steel mill MES, logistics provider TMS, warehouse WMS, and customer CRM systems, and to standardize the coding of steel product categories and timestamps; 2) Blockchain Data Storage: Build a Hyperledger Fabric consortium chain, set up steel mills, logistics providers, and quality inspection agencies as nodes, and use hashing of key fields in purchase contracts, quality inspection reports, and transport receipts to store them on the chain to prevent data tampering. Customers can scan codes to trace steel production batches and logistics tracks, while logistics providers only have access to relevant waybills. 3) Cloud-edge collaborative architecture: By deploying edge servers in steel mills and warehouses, we pre-process real-time data and integrate data using Alibaba Cloud DataWorks, supporting petabyte-level storage and real-time analysis. L2. Visual monitoring system: 1) 3D digital twin modeling: A laser scanner is used to generate a high-precision warehouse model, which is then processed and mapped to the location of steel coils in real time. Vehicle GPS is connected to the model to display vehicle location, speed, temperature and humidity, and weight change data in real time on a GIS map. 2) Multi-layer data dashboard: Build a global overview of the supply chain, display real-time inventory in transit and capacity utilization data, create risk heat maps, and mark frequently delayed sections based on historical data; L3. Early Warning Decision-Making: Configure a rules engine to trigger an inventory warning when the safety stock level falls below 80%. It provides replenishment suggestions, recommends suppliers and delivery times, predicts transportation risks during catastrophic weather conditions based on weather conditions, automatically pushes rerouting plans, and implements automated collaborative process management.

6. The method for real-time dynamic management of the entire steel logistics process according to claim 1, characterized in that: The specific implementation steps of step S5 are as follows: L1. Risk identification and classification: 1) Identify risks across the entire steel logistics chain and scan and address risk sources, including: a. Transportation: Disaster weather causes road interruptions, vehicle breakdowns, and loose and shifting cargo; b. Warehousing: Fire, shelf collisions, and excessive moisture in warehouses can lead to steel corrosion. c. Supply chain: iron ore supplier disruptions, port strikes, and policy changes; 2) Prioritize based on probability of occurrence and impact, set the evaluation criteria on a scale of 1-5, and build a risk knowledge base; L2. Monitoring and early warning system: 1) Sensor network deployment: Multi-parameter sensors are installed on transport vehicles and thresholds are set. If the temperature is >50°C, a temperature threshold alarm is triggered. If three consecutive sudden brakings occur per minute, the current driving behavior is determined to be high-risk: Smoke sensors and infrared thermal imagers are installed in the warehouse environment. If the air particle concentration is greater than 5% LEL, a smoke concentration alarm is triggered. If the local temperature rise is greater than 10°C / h, a fire alarm is triggered. 2) External Data Access: Access the Central Meteorological Observatory API to obtain real-time typhoon paths and rainfall forecasts. Use NLP to crawl government websites and identify keywords to automatically generate risk alerts. 3) Configure the warning rule engine to process warnings in a graded manner, set trigger conditions, and push warning information through a third-party platform; L3. Dynamic Decision-Making: 1) Design a contingency plan template to handle transport disruptions by deploying backup vehicles to pick up cargo before the disruption point and initiating multimodal transport. At the same time, notify the customer and negotiate a compensation plan. 2) Build a route replanning algorithm. By inputting the coordinates of the interruption point, the location of available vehicles, and the customer's time window, it outputs three alternative routes based on cost, timeliness, and risk comparison. Using a greedy algorithm, it quickly matches the nearest available resources to achieve resource scheduling optimization.

7. A real-time dynamic management device for the entire process of steel logistics, characterized by: It includes a data acquisition module, a route optimization module, a demand forecasting module, a visualization management module and a risk control decision module. The four are connected in sequence. The data acquisition module is used to obtain multi-source data and fuse the data; the route optimization module is used to optimize the transportation route and provide the optimal route combination decision; the demand forecasting module is used to predict demand fluctuations and adjust the safety stock according to the fluctuation range to achieve dynamic allocation; the visualization management module is used to visualize the data of the entire supply chain; the risk control decision module monitors risk data according to the risk warning model and provides decision support.

Images