MES management system based on internet of things
By building an IoT-based MES management system, the problems of data lag and heterogeneous data processing were solved, enabling real-time data collection, analysis, and production scheduling, thereby improving the controllability and efficiency of the production process.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- BEIJING ZHONGHE YIYOU TECHNOLOGY CO LTD
- Filing Date
- 2026-02-27
- Publication Date
- 2026-06-02
AI Technical Summary
Existing IoT-based MES systems suffer from problems such as data lag, inaccuracy, and poor real-time performance in data acquisition. Furthermore, they lack effective processing and deep integration of massive heterogeneous data, making it difficult to achieve intelligent production scheduling and refined management.
The MES management system is constructed with a perception layer, network layer, platform layer, and application layer. It adopts multiple types of IoT sensing terminals, heterogeneous converged communication architecture, machine learning models, and digital twin technology to realize real-time data acquisition, processing, analysis, and visualization simulation. It also combines multi-objective optimization algorithms for production scheduling and resource allocation.
It achieves low-latency and high-reliability data transmission, equipment health prediction and production anomaly early warning, improves the controllability and scheduling efficiency of the production process, and reduces management costs.
Smart Images
Figure CN122134005A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the fields of industrial automation and information technology, and more specifically, to an Internet of Things (IoT) based MES management system. Background Technology
[0002] MES (Manufacturing Execution System) is a shop floor-oriented management information system that sits between the upper-level planning and management system and the lower-level industrial control system. Traditional MES systems often rely on manual data entry or semi-automated barcode scanning for data collection, resulting in problems such as data lag, inaccuracy, and poor real-time performance.
[0003] With the development of IoT technology, combining it with MES systems has become a trend. However, existing IoT-based MES systems typically only connect sensor data, lacking effective processing and analysis of massive amounts of heterogeneous data and deep integration with production processes, making it difficult to achieve truly intelligent production scheduling and refined management. Summary of the Invention
[0004] In order to overcome the above-mentioned deficiencies of the prior art, embodiments of the present invention provide an Internet of Things-based MES management system to solve the problems mentioned in the background art.
[0005] To achieve the above objectives, the present invention provides the following technical solution: The perception layer includes multiple IoT sensing terminals deployed on production equipment, materials, work-in-process and the environment, used to collect equipment status data, process parameters, material location information and environmental data in real time; The network layer, including industrial Ethernet, wireless LAN and 5G communication units, is used to establish data communication channels and upload the data collected by the sensing layer. The platform layer includes: The IoT access module is used to manage the connected IoT sensing terminals and perform protocol parsing and data preprocessing. The data platform, including time-series databases and relational databases, is used to store processed real-time data and business data. The data processing and analysis engine cleans, filters, and aggregates real-time data based on a stream processing framework, and combines machine learning models to predict equipment health and provide early warnings of production anomalies. The digital twin module constructs a virtual workshop that maps to the physical workshop based on 3D modeling technology, and synchronizes the perception layer data in real time to realize the visualization simulation and traceability of the production process; The application layer includes: The production scheduling module dynamically optimizes production scheduling and resource allocation based on real-time data and forecasts provided by the data analysis engine.
[0006] Preferably, the perception layer consists of multiple types of IoT sensing terminals deployed throughout the entire production process, forming a comprehensive data acquisition network: vibration sensors, current sensors, temperature sensors, and speed sensors are embedded in key parts of the production equipment (motors, transmission systems, control systems) to collect data on the vibration frequency (sampling rate up to 100Hz), operating current fluctuation range, core component temperature changes, and speed stability during equipment operation in real time, thereby obtaining early signs of hidden equipment failures; each batch of materials and work-in-process is assigned a dual-frequency RFID tag with a built-in NFC chip; ultra-high frequency RFID readers (identification distance 0-5 meters) are deployed at workshop entrances and exits, shelves, and production line workstations; and high-frequency reading terminals are set up at key process nodes to achieve full-process location tracking and status updates for material warehousing, requisition, circulation, and consumption, as well as simultaneous reading of multiple tags (more than 30 tags can be identified at a time).
[0007] Preferably, the network layer constructs a heterogeneous converged communication architecture of "wired high-reliability backbone + wireless high-flexibility coverage + 5G deterministic enhancement", adopts a three-level topology design of "core switching layer - edge access layer - terminal perception layer" to achieve millisecond-level, high-fidelity, secure and reliable transmission of all elements of data in industrial scenarios; in fixed high real-time production line areas (such as stamping and final assembly lines), an industrial Ethernet ring network supporting TSN (IEEE 802.1Qbv Time-Aware Shaping) is deployed, adopting a dual-protocol redundancy architecture of PROFINETIRT and EtherCAT, and the dual-ring topology (MRP / PRP protocol) ensures self-healing of link failures within 50ms, and the jitter of key control command transmission is ≤10μs.
[0008] Preferably, in the IoT access module, a terminal access authentication mechanism is used to verify the device identity of newly added IoT sensing terminals (based on two-way authentication of the device's unique identifier and digital certificate), automatically completing terminal registration, model identification, and permission configuration. It supports real-time monitoring of terminal online status, remote parameter configuration, and fault alarm push. To address the issue of fragmented device protocols in industrial scenarios, a multi-protocol parsing engine is built-in, compatible with industrial protocols and custom private protocols. Through a protocol conversion module, heterogeneous data from different terminals are uniformly converted into the system's standard data format, solving the data silo problem caused by incompatibility of traditional system protocols.
[0009] Preferably, in the data processing and analysis engine, the data processing and analysis engine adopts the Lambda architecture and integrates batch and stream processing technology to perform in-depth real-time calculation and mining analysis on the data cleaned by the IoT access module; the engine is based on the Apache Flink stream processing framework and constructs a complete computing chain from real-time data cleaning to complex event processing and machine learning model prediction, and finally outputs device health indicators and production anomaly warning signals. At the real-time stream processing level, the engine performs three levels of computational tasks on the standardized data stream pushed by the IoT access module: First, it performs windowing cleaning and filtering based on a sliding time window (e.g., the sliding window width). sliding step size The original data is smoothed using the Laida criterion (to eliminate gross errors, i.e., when a data point...). satisfy Time (of which, Represented as the mean within the window, (Expressed as standard deviation), these are identified as outliers and removed, while fill-in is performed using linear interpolation; secondly, multi-stream association and aggregation are implemented, connecting data streams with device ID and timestamp as keys to static business master data (such as device model and rated parameters) in real time, and calculating aggregation features according to preset statistical periods (such as per minute or per shift); for example, calculating the real-time approximate value of the overall equipment efficiency (OEE), the specific calculation method is as follows: in, Represented as availability, Represented as performance, Represented as mass; It is calculated based on the time window of the device status (running, standby, shutdown) collected in real time; After completing the basic feature engineering, the engine's core capability is manifested in predictive analysis based on machine learning models. For critical production equipment, the system adopts an architecture combining offline training and online inference: in the offline phase, it utilizes historical operational data (such as vibration characteristics) stored in a time-series database. Temperature characteristics Current characteristics Using the fault label Y and its corresponding fault label Y, a device health prediction model based on a Long Short-Term Memory (LSTM) network is trained. The model's objective is to predict the device's remaining useful life (RUL), and its loss function is the mean squared error between the predicted RUL and the actual RUL. The specific calculation method is as follows: in, Represented as a loss function, Represented as predicted RUL, Let N represent the true RUL, N represent the number of samples, and i represent the i-th sample in the sample set; During the online inference phase, the engine consumes real-time data streams and extracts the same feature vectors as during the training phase. Input the loaded model file and output the current device's health index in real time. (Value range 0~1, 1 represents complete health, 0 represents complete...) At that time, a yellow alert was triggered.
[0010] Preferably, in the digital twin module, at the level of 3D modeling and geometric mapping, the module first constructs the geometric skeleton of the virtual workshop through multi-source data fusion technology; for static elements such as factory structure, production line layout, and equipment body, laser point cloud scanning and reverse modeling technology are used to construct a high-precision 3D geometric model in the Unity 3D engine based on CAD (computer-aided design) drawings and on-site measurement data, ensuring that the geometric deviation between the virtual space and the physical space meets the following requirements: ,in, These are the spatial coordinates of key feature points in the virtual model. These are the measured coordinates of the points corresponding to the physical entity. A threshold for modeling accuracy is set. For dynamic entities such as AGVs, material pallets, and work-in-process, a parametric model library is established to support dynamic adjustment of the model's pose and state based on real-time data. At the level of real-time data synchronization and behavior mapping, the digital twin module subscribes to the real-time data streams of the IoT access module and the data platform through a high-speed data interface, driving the virtual model and physical entities to achieve time synchronization and behavioral consistency. Internally, the module constructs a unified object model, where each physical entity's corresponding agent in the virtual space is bound to its perception data channel. For example, for a CNC machine tool, its virtual model receives the spindle load in real time. Feed rate ,temperature and operating status (Run, Standby, Alarm, Shutdown), and drive 3D animation through a state machine engine: when During operation, the model's spindle component rotates at the calculated speed. Tool parameters) rotation, cutting fluid flow effects are triggered synchronously; the position and attitude update of the virtual model follow rigid body kinematics constraints, let the target position of a certain AGV at time t be... The current actual location is The position of the virtual model in the next rendering frame. The interpolation algorithm smoothly approximates the actual position, ensuring visual continuity. The specific calculation method is as follows: in, Represented as smoothing factor It is used to balance response speed and visual jitter.
[0011] Preferably, in the production scheduling module, the production scheduling module is driven by real-time production data output by the data processing and analysis engine, equipment health prediction results, and anomaly warning information. It dynamically adjusts production scheduling and resource allocation through a multi-objective optimization algorithm to achieve optimal synergy between production efficiency, cost control, and order delivery. The module first constructs a multi-dimensional scheduling constraint model, integrating constraints such as order priority (urgent orders, regular orders, and replenishment orders), equipment load rate, material inventory status, personnel skill matching degree, and process dependencies. Order priority is quantified through a weighted scoring mechanism, and the specific calculation method is as follows: in, The priority score for the a-th order (value range [0,10]), Due to the urgency of order delivery, The order batch weight (the larger the batch, the higher the weight, with values ranging from [0,3]) Customer level weights (core customers take values [0,4]), , , The weighting coefficients are used to quantify and rank orders based on their priority using this formula. The specific method for calculating order delivery urgency is as follows: in, For delivery deadline, For the current time, For planned production cycles; In the production scheduling stage, an improved genetic algorithm (GA) is used to optimize the scheduling scheme. The multi-objective optimization function is "minimizing order delivery delay time, maximizing equipment utilization, and minimizing changeover costs." The specific calculation method for the core objective function is as follows: in, Let this be the delay time for the a-th order. For order quantity, Let be the overall efficiency of the b-th device. The total number of devices. Let be the time cost of the k-th line switch. This represents the number of line swaps. The target weight is used; the algorithm represents the process order through binary encoding, and uses roulette wheel selection, single-point crossover and mutation operations to generate the next generation scheduling scheme. After 50-100 iterations, the optimal scheduling plan is output. The scheduling update cycle is set to 10-30 minutes according to the production rhythm to ensure synchronization with the real-time production status.
[0012] The specific method for calculating the delay time of the a-th order is as follows: In terms of resource allocation, the module dynamically allocates production tasks based on equipment health prediction results and real-time load rates: prioritizing high-precision, high-load production tasks. Health equipment, for Early warning equipment is assigned light load tasks and maintenance windows are reserved. The fault warning equipment automatically avoids task assignments while incorporating a personnel skills matrix (which records the operator's proficiency rating for each process). The optimal matching of personnel and workstations is achieved through the Hungarian algorithm to minimize work time loss. The matching optimization formula is as follows: in The estimated operating time for operator p on equipment b. The standard operating time for equipment b. (Avoid denominator of 0) Let g be a 0-1 decision variable (1 indicates a match, 0 indicates a mismatch), and g be the number of operators.
[0013] The technical effects and advantages of this invention are as follows: 1. This invention constructs a heterogeneous integrated communication architecture and a multi-protocol parsing engine to achieve low-latency and high-reliability data transmission, break down data silos caused by protocol fragmentation, and ensure the efficiency and security of data transmission through intelligent traffic scheduling and in-depth security protection. 2. This invention integrates technologies such as machine learning and digital twins to achieve equipment health prediction, early warning of production anomalies, and visualized simulation and full lifecycle traceability of the production process, upgrading MES from "post-event statistics" to "pre-event prediction and in-event control" and improving the controllability of the production process; 3. Based on real-time data analysis results, this invention achieves dynamic optimization of production scheduling and resource allocation through multi-objective optimization algorithms and intelligent matching algorithms. Combined with an emergency scheduling mechanism, it can quickly respond to sudden production situations, significantly improve production scheduling efficiency and resource utilization, and reduce production management costs. Attached Figure Description
[0014] Figure 1 This is a schematic diagram of the module connection of the present invention.
[0015] Figure 2 This is a schematic diagram of the data processing and analysis engine workflow of the present invention.
[0016] Figure 3 This is a schematic diagram of the operation process of the digital twin module of the present invention. Detailed Implementation
[0017] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.
[0018] Please see Figure 1 As shown, the present invention provides an Internet of Things (IoT) based MES management system, comprising: The perception layer includes multiple IoT sensing terminals deployed on production equipment, materials, work-in-process and the environment, used to collect equipment status data, process parameters, material location information and environmental data in real time; The perception layer consists of multiple types of IoT sensing terminals deployed throughout the entire production process, forming a comprehensive data acquisition network: vibration sensors, current sensors, temperature sensors, and speed sensors are embedded in key parts of production equipment (motors, transmission systems, control systems) to collect real-time data on vibration frequency (sampling rate up to 100Hz), operating current fluctuation range, core component temperature changes, and speed stability during equipment operation, thereby identifying early signs of hidden equipment failures; each batch of materials and work-in-process is assigned a dual-frequency RFID tag with a built-in NFC chip; ultra-high frequency RFID readers (reading distance 0-5 meters) are deployed at workshop entrances and exits, shelves, and production line workstations; and additional sensors are installed at key process nodes. High-frequency read / write terminals are installed to achieve full-process location tracking and status updates for materials during warehousing, requisition, circulation, and consumption, as well as simultaneous reading of multiple tags (more than 30 tags can be identified at a time). Visual inspection terminals and laser rangefinders are installed at production workstations to collect real-time data on product appearance dimensions, assembly accuracy, and process execution parameters (such as welding temperature, coating thickness, and processing time). The visual inspection terminals use AI image recognition algorithms to automatically identify appearance defects at the 0.1mm level. Temperature and humidity sensors, dust concentration sensors, gas sensors, and light sensors are deployed throughout the workshop to collect data on the temperature, humidity, dust concentration, and harmful gas content of the production environment, meeting the environmental compliance requirements of different industries.
[0019] The network layer, including industrial Ethernet, wireless LAN and 5G communication units, is used to establish data communication channels and upload the data collected by the sensing layer. The network layer constructs a heterogeneous converged communication architecture of "wired high-reliability backbone + wireless high-flexibility coverage + 5G deterministic enhancement", and adopts a three-level topology design of "core switching layer - edge access layer - terminal perception layer" to achieve millisecond-level, high-fidelity, secure and reliable transmission of all elements of data in industrial scenarios. In fixed high real-time production line areas (such as stamping and final assembly lines), an industrial Ethernet ring network supporting TSN (IEEE 802.1Qbv time-aware shaping) is deployed, adopting a dual-protocol redundancy architecture of PROFINET IRT and EtherCAT. The dual-ring topology (MRP / PRP protocol) ensures self-healing of link failures within 50ms, and the jitter of critical control command transmission is ≤10μs. In areas with high mobile device density (AGV paths, flexible assembly stations), deploy Wi-Fi 6E industrial-grade AP arrays, enabling 160MHz channel bonding, OFDMA resource scheduling, and 802.11k / v / r fast roaming protocols, combined with RF environment self-optimization algorithms; for high mobility and ultra-low latency scenarios (such as collaborative robot clusters, AR remote operation and maintenance), deploy 5G private network standalone (SA) mode, configure a down-to-the-line UPF (user plane function) to the edge equipment room based on uRLLC network slicing, and stably control end-to-end latency within 8ms (air interface <1ms), and use 5G LAN technology to achieve seamless IP interoperability with OT networks; The network layer features an AI-driven intelligent traffic scheduling engine: based on data semantic tags (e.g., "equipment emergency stop command" is tagged with CS7 priority, "4K quality inspection video stream" is tagged with AF41), QoS policies are dynamically allocated—critical control flows are guaranteed to be transmitted at the microsecond level through TSN time gating, video streams are uploaded in staggered shifts after being compressed using H.265+ROI regional encoding and transmitted via Wi-Fi 6's TWT (Target Wake-up Time) mechanism, and massive sensor data is aggregated at edge nodes and then transmitted in batches by 5G mMTC slices; simultaneously, a "network digital twin" unit is integrated to mirror the physical network topology, traffic heatmap, and link quality in real time, and to predict congestion risks and pre-adjust routing policies using graph neural networks; the security system implements "end-pipe-cloud" in-depth protection: terminal access uses 802.1X+EAP-TLS two-way authentication and MACsec national cryptographic SM4 encryption, service flows are divided into VLANs for isolation according to security levels, industrial firewalls and LSTM-based abnormal traffic detection systems are deployed (real-time identification of DDoS and scanning attacks), and all cross-domain data transmissions use dynamic keys generated by quantum random numbers.
[0020] The platform layer includes: The IoT access module is used to manage the connected IoT sensing terminals and perform protocol parsing and data preprocessing. The IoT access module verifies the identity of newly added IoT sensing terminals through a terminal access authentication mechanism (based on two-way authentication of unique device identifier and digital certificate), automatically completes terminal registration, model identification, and permission configuration, and supports real-time monitoring of terminal online status, remote parameter configuration, and fault alarm push. Addressing the fragmentation of device protocols in industrial scenarios, it incorporates a multi-protocol parsing engine, compatible with industrial protocols and custom private protocols. A protocol conversion module unifies heterogeneous data from different terminals into the system's standard data format, resolving the data silo problem caused by incompatible traditional system protocols. In the data preprocessing stage, an "edge computing + intelligent filtering" strategy is adopted. First, an outlier detection algorithm (based on the 3σ principle and sliding window model) removes invalid data caused by sensor drift and transmission interference. Then, high-frequency collected data undergoes downsampling (dynamically adjusting the sampling frequency according to data importance, retaining the original accuracy of key parameters, and using mean compression for non-key parameters). Simultaneously, data format standardization, field completion, and timestamp synchronization are completed. The data platform, including time-series databases and relational databases, is used to store processed real-time data and business data. In the data platform, which serves as the core of the system's data storage and management, a hybrid storage architecture of "time-series database + relational database" is adopted to achieve efficient and categorized storage of real-time data and business data. The time-series database (deployed using an InfluxDB cluster) stores real-time data collected by the sensing layer, including equipment operating status parameters, environmental monitoring data, and process data. Through time indexing optimization technology, historical time-series data within any time period can be retrieved, meeting the query needs of scenarios such as equipment fault tracing and production trend analysis. The storage capacity of a single node can be expanded to the petabyte level, and it supports automatic tiered storage of data according to its lifecycle to balance storage performance. The data platform utilizes a relational database (using a MySQL master-slave architecture) to store structured business data, including order information, material ledgers, personnel files, production plans, quality inspection reports, and equipment files. Through data table partitioning, index optimization, and transaction management mechanisms, it ensures data read / write consistency and reliability, supports complex relational queries and multi-condition filtering, and meets the data analysis needs of business scenarios such as production scheduling, cost accounting, and report statistics. Furthermore, the data platform has a built-in data synchronization and backup module, ensuring no data loss through scheduled incremental backups and real-time off-site disaster recovery mechanisms. It also supports flexible data interaction with upper-layer application modules and provides standardized data access interfaces.
[0021] The data processing and analysis engine cleans, filters, and aggregates real-time data based on a stream processing framework, and combines machine learning models to predict equipment health and provide early warnings of production anomalies. The data processing and analysis engine adopts the Lambda architecture and integrates batch and stream processing technology to perform in-depth real-time calculation and mining analysis on the data cleaned by the IoT access module. Based on the Apache Flink stream processing framework, the engine has built a complete computing chain from real-time data cleaning to complex event processing and machine learning model prediction, and finally outputs equipment health indicators and production anomaly warning signals. At the real-time stream processing level, the engine performs three levels of computational tasks on the standardized data stream pushed by the IoT access module: First, it performs windowing cleaning and filtering based on a sliding time window (e.g., the sliding window width). sliding step size The original data is smoothed using the Laida criterion (to eliminate gross errors, i.e., when a data point...). satisfy Time (of which, Represented as the mean within the window, (Expressed as standard deviation), these are identified as outliers and removed, while fill-in is performed using linear interpolation; secondly, multi-stream association and aggregation are implemented, connecting data streams with device ID and timestamp as keys to static business master data (such as device model and rated parameters) in real time, and calculating aggregation features according to preset statistical periods (such as per minute or per shift); for example, calculating the real-time approximate value of the overall equipment efficiency (OEE), the specific calculation method is as follows: in, Represented as availability, Represented as performance, Represented as mass; It is calculated based on the time window of the device status (running, standby, shutdown) collected in real time; After completing the basic feature engineering, the engine's core capability is manifested in predictive analysis based on machine learning models. For critical production equipment, the system adopts an architecture combining offline training and online inference: in the offline phase, it utilizes historical operational data (such as vibration characteristics) stored in a time-series database. Temperature characteristics Current characteristics Using the fault label Y and its corresponding fault label Y, a device health prediction model based on a Long Short-Term Memory (LSTM) network is trained. The model's objective is to predict the device's remaining useful life (RUL), and its loss function is the mean squared error between the predicted RUL and the actual RUL. The specific calculation method is as follows: in, Represented as a loss function, Represented as predicted RUL, Let N represent the true RUL, N represent the number of samples, and i represent the i-th sample in the sample set; During the online inference phase, the engine consumes real-time data streams and extracts the same feature vectors as during the training phase. Input the loaded model file and output the current device's health index in real time. (Values range from 0 to 1, where 1 represents perfect health and 0 represents complete ill health) and remaining lifespan. ; At the equipment health status assessment level, the module deeply integrates the predictive analysis results output by the data processing and analysis engine to establish a multi-dimensional equipment health evaluation system. For each piece of equipment, the module defines a health index. (Value range 0~1), taking into account multiple indicators: in, Vibration characteristic index, based on real-time effective vibration value Compared with the benchmark value calculate: This is expressed as a temperature deviation index; This is expressed as a load balancing metric. The normalized function representing the remaining effective lifetime is defined as follows: RUL These are expressed as weighting coefficients, determined using the Analytic Hierarchy Process (AHP) combined with expert experience; Meanwhile, the engine integrates a Complex Event Processing (CEP) unit for production anomaly warnings; this unit monitors the data stream after model scoring in real time based on predefined business rules and pattern matching logic; in quality control scenarios, it uses statistical process control (SPC) anomaly detection criteria to calculate a certain process parameter in real time. Motion range and process capability index The specific method for calculating the process capability index is as follows: USL and LSL are process design standards, where USL represents the upper limit of the specification and LSL represents the lower limit of the specification; when calculated in real time... If 7 consecutive points appear on the same side of the mean, the CEP unit immediately generates a quality anomaly event; The specific method for calculating the moving range is as follows: in, Let be the mean of the i-th sample group. This represents the measured value of the h-th specific process parameter in the i-th sample group; Let be the range of the i-th sample group; Then calculate the population mean and average range to determine the upper control limit (USL) and lower control limit (LSL).
[0022] The digital twin module constructs a virtual workshop that maps to the physical workshop based on 3D modeling technology, and synchronizes the perception layer data in real time to realize the visualization simulation and traceability of the production process; In the digital twin module, at the 3D modeling and geometric mapping level, the module first constructs the geometric skeleton of the virtual workshop through multi-source data fusion technology. For static elements such as factory structure, production line layout, and equipment, laser point cloud scanning and reverse modeling technology are used to construct a high-precision 3D geometric model in the Unity 3D engine based on CAD (computer-aided design) drawings and on-site measurement data, ensuring that the geometric deviation between the virtual space and the physical space meets the following requirements: ,in, These are the spatial coordinates of key feature points in the virtual model. These are the measured coordinates of the points corresponding to the physical entity. A threshold for modeling accuracy is set. For dynamic entities such as AGVs, material pallets, and work-in-process, a parametric model library is established to support dynamic adjustment of the model's pose and state based on real-time data. At the level of real-time data synchronization and behavior mapping, the digital twin module subscribes to the real-time data streams of the IoT access module and the data platform through a high-speed data interface, driving the virtual model and physical entities to achieve time synchronization and behavioral consistency. Internally, the module constructs a unified object model, where each physical entity's corresponding agent in the virtual space is bound to its perception data channel. For example, for a CNC machine tool, its virtual model receives the spindle load in real time. Feed rate ,temperature and operating status (Run, Standby, Alarm, Shutdown), and drive 3D animation through a state machine engine: when During operation, the model's spindle component rotates at the calculated speed. Tool parameters) rotation, cutting fluid flow effects are triggered synchronously; the position and attitude update of the virtual model follow rigid body kinematics constraints, let the target position of a certain AGV at time t be... The current actual location is The position of the virtual model in the next rendering frame. The interpolation algorithm smoothly approximates the actual position, ensuring visual continuity. The specific calculation method is as follows: in, Represented as smoothing factor This is used to balance response speed and visual jitter. At the level of production process visualization and simulation, prediction data (such as predicted equipment remaining lifespan) is provided by the data processing and analysis engine. This allows for the simulation of future production states; for example, when the engine predicts that a certain machine tool will be... If a malfunction occurs, the digital twin module highlights the virtual model of the machine tool in yellow and simulates the logistics congestion following the malfunction, presenting the congestion index through visual coding such as color gradients and flowing lines. The spatial distribution of is calculated using the following method: in, The current work-in-process quantity at the workstation is represented by BufferCapacity, which represents the workstation's rated buffer capacity. When this happens, the corresponding area model displays an orange warning to help managers intuitively identify production bottlenecks; In terms of full lifecycle traceability of the production process, the module integrates a spatiotemporal backtracking engine; this engine indexes and stores all sensing data and the operational status of the twin model according to a timeline; users can select any historical moment in the 3D virtual workshop interface by using the timeline slider. The system quickly retrieves all equipment status, material location, and environmental parameters at a given moment from the time-series database and drives the virtual workshop to recreate the 3D scene at that time. This process involves spatiotemporal correlation queries of massive amounts of data. Through joint optimization of spatial and temporal indexes, the smoothness of the backtracking process is ensured. During the tracing process, users can click on any virtual device or material to retrieve its detailed information panel at that moment. For example, for a work-in-process workpiece j, users can query its details at time j. Processing parameters: Processing parameters Device ID, Program Number, Spindle Load feed rate Quality inspection results This enables one-click backtracking from final product quality issues to historical processing procedures, providing visualized data support for quality analysis and process optimization.
[0023] The application layer includes: The production scheduling module dynamically optimizes production scheduling and resource allocation based on real-time data and forecast results provided by the data analysis engine. In the production scheduling module, driven by real-time production data output by the data processing and analysis engine, equipment health prediction results, and anomaly warning information, the module dynamically adjusts production scheduling and resource allocation through a multi-objective optimization algorithm to achieve optimal synergy between production efficiency, cost control, and order delivery. The module first constructs a multi-dimensional scheduling constraint model, integrating constraints such as order priority (urgent orders, regular orders, and replenishment orders), equipment load rate, material inventory status, personnel skill matching, and process dependencies. Order priority is quantified through a weighted scoring mechanism, specifically calculated as follows:
[0024] in, The priority score for the a-th order (value range [0,10]), Due to the urgency of order delivery, The order batch weight (the larger the batch, the higher the weight, with values ranging from [0,3]) Customer level weights (core customers take values [0,4]), , , The weighting coefficients are used to quantify and rank orders based on their priority using this formula. The specific method for calculating order delivery urgency is as follows: in, For delivery deadline, For the current time, For planned production cycles; In the production scheduling stage, an improved genetic algorithm (GA) is used to optimize the scheduling scheme. The multi-objective optimization function is "minimizing order delivery delay time, maximizing equipment utilization, and minimizing changeover costs." The specific calculation method for the core objective function is as follows: in, Let this be the delay time for the a-th order. For order quantity, Let be the overall efficiency of the b-th device. The total number of devices. Let be the time cost of the k-th line switch. This represents the number of line swaps. The target weight is used; the algorithm represents the process order through binary encoding, and uses roulette wheel selection, single-point crossover and mutation operations to generate the next generation scheduling scheme. After 50-100 iterations, the optimal scheduling plan is output. The scheduling update cycle is set to 10-30 minutes according to the production rhythm to ensure synchronization with the real-time production status.
[0025] The specific method for calculating the delay time of the a-th order is as follows: In terms of resource allocation, the module dynamically allocates production tasks based on equipment health prediction results and real-time load rates: prioritizing high-precision, high-load production tasks. Health equipment, for Early warning equipment is assigned light load tasks and maintenance windows are reserved. The fault warning equipment automatically avoids task assignments while incorporating a personnel skills matrix (which records the operator's proficiency rating for each process). The optimal matching of personnel and workstations is achieved through the Hungarian algorithm to minimize work time loss. The matching optimization formula is as follows: in The estimated operating time for operator p on equipment b. The standard operating time for equipment b. (Avoid denominator of 0) Let g be a 0-1 decision variable (1 indicates a match, 0 indicates a mismatch), and g be the number of operators. When the data processing and analysis engine pushes early warning information such as equipment failure, material shortage, and quality abnormality, the module automatically triggers the emergency dispatch mechanism: if it is an equipment failure, it quickly calculates the task replacement plan for the faulty equipment, transfers the unfinished tasks to the backup equipment or adjusts the production sequence, and optimizes the changeover time to within 10 minutes; if it is a material shortage, based on the material replenishment forecast time of the cross-system collaboration module, it adjusts the production priority of related orders or splits production tasks to ensure that critical orders are not affected. All scheduling adjustment instructions are simultaneously pushed to relevant personnel through the Andon system, workstation terminals, and mobile APP. At the same time, the virtual scheduling status is updated in real time in the digital twin module, enabling visualized communication and execution tracking of scheduling instructions. Furthermore, the module supports real-time monitoring and deviation analysis of production progress, calculating production achievement rate (…). With process pass rate ( When AR < 90% or PR < 95%, the system automatically analyzes the causes of deviations (equipment bottlenecks, personnel efficiency, material delays, etc.) and generates optimization suggestions.
[0026] In conclusion, the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.
Claims
1. A MES management system based on the Internet of Things, characterized in that, include: The perception layer includes multiple IoT sensing terminals deployed on production equipment, materials, work-in-process and the environment, used to collect equipment status data, process parameters, material location information and environmental data in real time; The network layer, including industrial Ethernet, wireless LAN and 5G communication units, is used to establish data communication channels and upload the data collected by the sensing layer. The platform layer includes: The IoT access module is used to manage the connected IoT sensing terminals and perform protocol parsing and data preprocessing. The data platform includes time-series databases and relational databases, used to store processed real-time data and business data; The data processing and analysis engine cleans, filters, and aggregates real-time data based on a stream processing framework, and combines machine learning models to predict equipment health and provide early warnings of production anomalies. The digital twin module, based on 3D modeling technology, constructs a virtual workshop that maps to the physical workshop, and synchronizes the perception layer data in real time to achieve visualized simulation and traceability of the production process; The application layer includes: The production scheduling module dynamically optimizes production scheduling and resource allocation based on real-time data and forecasts provided by the data analysis engine.
2. The MES management system based on the Internet of Things according to claim 1, characterized in that: The sensing layer equips materials and work-in-process with dual-frequency RFID tags with built-in NFC chips, and deploys ultra-high frequency RFID readers and high frequency reading terminals to perform tag identification and complete the entire process of material location tracking and status updates.
3. The MES management system based on the Internet of Things according to claim 1, characterized in that: In the network layer, a heterogeneous converged communication architecture of "wired high-reliability backbone + wireless high-flexibility coverage + 5G deterministic enhancement" is constructed, and a three-level topology design of "core switching layer - edge access layer - terminal perception layer" is adopted to transmit all elements of data in industrial scenarios.
4. The MES management system based on the Internet of Things according to claim 1, characterized in that: The data processing and analysis engine adopts the Lambda architecture and integrates batch and stream processing technology to perform in-depth real-time calculation and mining analysis on the data cleaned by the IoT access module. At the equipment health status assessment level, the module deeply integrates the predictive analysis results output by the data processing and analysis engine to establish a multi-dimensional equipment health evaluation system; for each piece of equipment, the module defines a health index. (Value range 0~1), taking into account multiple indicators, the specific calculation method is as follows: in, Vibration characteristic index, based on real-time effective vibration value Compared with the benchmark value calculate: ; This is expressed as a temperature deviation index; This is expressed as a load balancing metric. The normalized function representing the remaining effective lifetime is defined as follows: RUL ; These are expressed as weighting coefficients, determined through the analytic hierarchy process combined with expert experience.
5. The MES management system based on the Internet of Things according to claim 4, characterized in that: Meanwhile, the engine integrates a complex event processing unit for production anomaly early warning; this unit monitors the data stream after model scoring in real time based on predefined business rules and pattern matching logic; in quality control scenarios, it uses statistical process control anomaly criteria to calculate a certain process parameter in real time. Motion range and process capability index The specific method for calculating the process capability index is as follows: USL and LSL are process design standards, where USL represents the upper limit of the specification and LSL represents the lower limit of the specification; when calculated in real time... If 7 consecutive points appear on the same side of the mean, the CEP unit immediately generates a quality anomaly event.
6. The MES management system based on the Internet of Things according to claim 1, characterized in that: In the digital twin module, at the 3D modeling and geometric mapping level, the module first constructs the geometric skeleton of the virtual workshop through multi-source data fusion technology. For static elements such as factory structure, production line layout, and equipment, laser point cloud scanning and reverse modeling technology are used to construct a high-precision 3D geometric model in the Unity 3D engine based on CAD (computer-aided design) drawings and on-site measurement data, ensuring that the geometric deviation between the virtual space and the physical space meets the following requirements: ,in, These are the spatial coordinates of key feature points in the virtual model. These are the measured coordinates of the points corresponding to the physical entity. This serves as a threshold for modeling accuracy. At the real-time data synchronization and behavior mapping level, the digital twin module subscribes to the real-time data streams of the IoT access module and the data platform through a high-speed data interface, driving the virtual model and the physical entity to achieve time synchronization and behavioral consistency. The position and attitude updates of the virtual model follow rigid body kinematics constraints. Let the target position of a certain AGV at time t be... The current actual location is The position of the virtual model in the next rendering frame. The interpolation algorithm smoothly approximates the actual position, ensuring visual continuity. The specific calculation method is as follows: in, It is represented as a smoothing factor.
7. The MES management system based on the Internet of Things according to claim 1, characterized in that: In the production scheduling module, driven by real-time production data output by the data processing and analysis engine, equipment health prediction results, and anomaly warning information, the module dynamically adjusts production scheduling and resource allocation through a multi-objective optimization algorithm. The module first constructs a multi-dimensional scheduling constraint model, integrating constraints. The specific method for calculating order priority scores is as follows: in, The priority score for the a-th order. Due to the urgency of order delivery, For order batch weight, Customer level weights (core customers take values [0,4]), , , These are the weighting coefficients.
8. The MES management system based on the Internet of Things according to claim 7, characterized in that: In the production scheduling stage, an improved genetic algorithm is used to optimize the scheduling scheme. The multi-objective optimization function is to minimize order delivery delay time, maximize equipment utilization, and minimize changeover cost. The specific calculation method of the core objective function is as follows: in, Let be the delay time for the a-th order. For order quantity, Let be the overall efficiency of the b-th device. The total number of devices. Let be the time cost of the k-th line switch. This represents the number of line swaps. The target weight.
9. The MES management system based on the Internet of Things according to claim 7, characterized in that: In terms of resource allocation, the module dynamically allocates production tasks based on equipment health prediction results and real-time load rates: assigning high-precision, high-load production tasks to... Health equipment, for Early warning equipment is assigned light load tasks and maintenance windows are reserved. The fault warning equipment automatically avoids task assignments and, combined with the personnel skill matrix, uses the Hungarian algorithm to achieve optimal matching between personnel and workstations, minimizing work time loss. The matching optimization formula is: in The estimated operating time for operator p on equipment b. This is the standard operating time for equipment b. (Avoid denominator of 0) Let g be a 0-1 decision variable (1 indicates a match, 0 indicates a mismatch), and g be the number of operators. When the data processing and analysis engine pushes an early warning message, the module automatically triggers the emergency dispatch mechanism. All dispatch adjustment instructions are simultaneously pushed to relevant personnel through the Andon system, workstation terminals, and mobile APP. At the same time, the virtual scheduling status is updated in real time in the digital twin module to visualize the communication and execution tracking of dispatch instructions.