Float glass cold end intelligent control system and method based on Internet of Things and big data

Abstract

The invention discloses a float glass cold end intelligent control system and method based on Internet of Things and big data, and belongs to the technical field of float glass production. The system comprises a global sensing layer, a data aggregation and edge processing layer, a cloud platform big data analysis and decision-making layer and a man-machine interaction and execution layer, full-process data is collected through the global sensing layer, preprocessing and real-time response are carried out through the edge processing layer, an intelligent model is constructed by the cloud platform layer to realize analysis and decision-making, and the human-machine interaction and execution layer is established. Finally, instruction landing and production execution are completed through a man-machine interaction and execution layer, and a'perception-analysis-decision-control 'closed loop is formed. According to the method, data acquisition, preprocessing, model operation, fault early warning, decision optimization, instruction issuing and execution feedback are sequentially completed based on the system. The problems that an existing cold end system is isolated in information, decision-making depends on manpower, fault response lags behind, quality control is passive and the like are solved, and production efficiency, product quality and equipment reliability are improved.

Description

Technical Field

[0001] This invention relates to the field of float glass production, specifically to an intelligent control system and method for the cold end of float glass based on the Internet of Things and big data. Background Technology

[0002] The cold-end system of float glass is a crucial link in the float glass production line, primarily responsible for cutting, inspecting, and stacking the annealed glass strips. Its operational status directly determines the production efficiency, product quality, and production cost of float glass. Currently, the cold-end systems of most domestic float glass manufacturers still suffer from numerous technical shortcomings, making it difficult to meet the demands of modern intelligent manufacturing. Specific problems include: 1. The problem of information silos is prominent: Core equipment in the cold-end production process, such as cutting equipment, quality inspection devices, and stacking robots, are mostly provided by different manufacturers. Each device uses an independent control system and data storage unit, resulting in inconsistent data formats and incompatible transmission protocols. For example, when the quality inspection device detects a local scratch defect in the glass strip, it cannot promptly synchronize information such as the defect location and size to the cutting equipment. The cutting equipment still cuts according to a preset fixed length, resulting in defective glass being cut into finished product specifications, generating a large amount of waste and wasting raw materials and energy.

[0003] 2. Production decisions rely heavily on human experience: Key decision-making processes in cold-end production, such as setting cutting length, adjusting quality judgment thresholds, and selecting stacking strategies, heavily depend on the individual experience of operators. Differences in experience levels and operating habits among operators lead to poor production process stability. For example, when adjusting cutting parameters, operators must manually adjust cutting pressure and speed based on experience when dealing with 3mm ultra-thin glass and 12mm thick glass. Improper parameter matching can easily result in chipped edges, uneven cross-sections, and other problems, affecting product quality and even causing glass breakage.

[0004] 3. Delayed Equipment Failure Response: Existing cold-end systems primarily rely on "reactive maintenance" or "fixed-cycle maintenance." "Reactive maintenance" requires repairs to be carried out after equipment failure and downtime, resulting in prolonged unplanned downtime and severely impacting production schedules. "Fixed-cycle maintenance" lacks consideration of the actual operating condition of the equipment, potentially leading to either "over-maintenance" (e.g., replacing parts before they reach their wear limits, increasing maintenance costs) or "under-maintenance" (e.g., failing to detect severely worn cutting wheels in time, resulting in decreased cutting quality or even equipment damage). Industry statistics indicate that cold-end systems using traditional maintenance methods typically experience an average of over 600 unused loads per month, severely hindering production efficiency.

[0005] 4. Quality control is reactive: Currently, quality inspection in cold-end systems is mostly conducted after the glass ribbon has been cut, which is a "post-inspection" model. When defects such as bubbles or tin contamination are found in the glass, the defective glass has already been cut and segmented, becoming scrap, making it impossible to provide early warning or process intervention. For example, if uneven local stress occurs in the glass ribbon during annealing due to improper temperature control, existing inspection methods cannot identify this problem before cutting. After cutting, the stress release can easily lead to glass breakage, further increasing the scrap rate.

[0006] Therefore, those skilled in the art have provided an intelligent control system and method for the cold end of float glass based on the Internet of Things and big data to solve the problems mentioned in the background art. Summary of the Invention

[0007] The purpose of this invention is to provide an intelligent control system and method for the cold end of float glass based on the Internet of Things and big data, so as to solve the problems mentioned in the background art.

[0008] To achieve the above objectives, the present invention provides the following technical solution: The intelligent control system and method for the cold end of float glass based on the Internet of Things and big data include a global perception layer, a data aggregation and edge processing layer, a cloud platform big data analysis and decision-making layer, and a human-machine interaction and execution layer. The global perception layer is deployed at key nodes in the cold end to collect glass strip quality information, operating parameters, and equipment status data. The data aggregation and edge processing layer receives data from the global perception layer, performs preprocessing and real-time local data processing, and uploads the data to the cloud platform big data analysis and decision-making layer. The cloud platform big data analysis and decision-making layer stores the data and generates decision instructions through intelligent models. The human-machine interaction and execution layer displays the data and decision suggestions, and executes the decision instructions to drive the equipment operation.

[0009] As a further aspect of the present invention: the global perception layer includes a machine vision module, a laser speed / length measurement module, and an equipment status monitoring sensor; the machine vision module is used to scan the glass strip online before cutting to detect and locate defects such as bubbles, scratches, and solder adhesion; the laser speed / length measurement module is used to accurately measure the running speed and length of the glass strip; the equipment status monitoring sensor includes a vibration sensor, a temperature sensor, and a current sensor, used to collect the operating parameters of the cutting wheel and the stacking robot.

[0010] As a further aspect of the present invention: the data aggregation and edge processing layer includes an IoT gateway and edge computing nodes; the IoT gateway performs filtering, compression and protocol conversion on the massive data collected by the global perception layer; the edge computing nodes perform local processing on data with high real-time requirements, including image analysis of the machine vision module, identifying defects and generating defect coordinate information, and sending it to the cutting controller to achieve defect avoidance.

[0011] As a further aspect of the present invention: the cloud platform big data analysis and decision-making layer includes a data lake and a big data analysis engine; the data lake stores historical and real-time data uploaded by the edge processing layer, and the data types cover process parameters, equipment status data, quality data and production order information; the big data analysis engine is used to build predictive maintenance models and adaptive optimization control models.

[0012] As a further aspect of the present invention: the predictive maintenance model is trained using machine learning algorithms such as time series analysis and anomaly detection based on vibration spectrum, temperature rise trend and other data collected by equipment condition monitoring sensors, and is used to provide early warning of faults such as cutting wheel wear and bearing failure, and to generate maintenance suggestions; the adaptive optimization control model receives production orders, real-time quality data and equipment status information, and generates control commands through reinforcement learning or deep learning algorithms.

[0013] As a further aspect of the present invention: the control instructions generated by the adaptive optimization control model include intelligent scheduling and cutting optimization instructions, process parameter self-tuning instructions, and collaborative scheduling instructions; the intelligent scheduling and cutting optimization instructions adjust the cutting scheme according to defect location information; the process parameter self-tuning instructions fine-tune the cutting pressure and speed parameters according to glass thickness and speed changes; and the collaborative scheduling instructions issue stacking mode and path instructions to the stacking robot.

[0014] As a further aspect of the present invention: the human-machine interaction and execution layer includes an intelligent control platform and a PLC / actuator; the intelligent control platform provides a visual interface to display real-time data of the entire process, equipment health status, quality reports and optimization decision suggestions; the PLC / actuator receives optimization instructions from the cloud platform big data analysis and decision layer or data aggregation and edge processing layer to drive the cutting machine and stacking robot to complete the action.

[0015] A smart control method for the cold end of float glass based on the Internet of Things and big data includes the following steps: S1: Through the machine vision module, laser speed / length measurement module and equipment status monitoring sensor of the global perception layer, information on glass strip defects, running speed and length data and equipment operating parameters are collected respectively; S2: The IoT gateway of the data aggregation and edge processing layer filters, compresses and converts the data collected by S1. The edge computing node analyzes the glass strip defect information locally in real time and issues instructions. At the same time, all the processed data is uploaded to the data lake of the cloud platform big data analysis and decision-making layer. S3: The big data analysis engine of the cloud platform's big data analysis and decision-making layer calls data lake data to run predictive maintenance models and adaptive optimization control models; S4: The predictive maintenance model outputs equipment health status warnings and maintenance suggestions to the intelligent management and control platform; S5: The adaptive optimization control model integrates production orders, real-time quality data, and equipment status information to generate the optimal cutting scheme and equipment control parameter set; S6: Send the control parameter set generated in S5 to the PLC / actuator; S7: The PLC / actuator drives the equipment to complete the production operation and feeds the execution results back to the data lake to form a closed-loop optimization.

[0016] As a further aspect of the present invention: the predictive maintenance model in step S3 determines the wear degree of equipment components by analyzing the vibration spectrum, identifies abnormal heating of equipment by tracking the temperature rise trend, and optimizes the early warning accuracy by combining historical fault data; the adaptive optimization control model prioritizes avoiding defective areas when generating cutting schemes to ensure that the yield of high-quality glass is maximized.

[0017] As a further aspect of the present invention: the execution results in step S7 include the actual operating parameters of the equipment, the glass cutting quality inspection results, and the stacking completion status; after the feedback data is stored in the data lake, it is used to update the training data of the predictive maintenance model and the adaptive optimization control model, thereby improving the accuracy of model decision-making.

[0018] Compared with the prior art, the beneficial effects of the present invention are: 1. Achieve global collaborative optimization and improve production efficiency: By breaking down information silos between cutting, testing and stacking equipment through IoT technology, data exchange and collaborative operation among multiple devices can be achieved; by dynamically adjusting production parameters and strategies with the help of an adaptive optimization control model, the glass yield can be increased by 3%-5%, the length of the production line that each operator can control can be increased by more than 100%, significantly improving production efficiency and per capita output value, and reducing enterprise production costs.

[0019] 2. Predictive maintenance reduces costs and increases efficiency, minimizing ineffective operations: Based on equipment status data and machine learning algorithms, it provides early warnings of equipment failures 24-72 hours in advance, transforming traditional "reactive maintenance" and "fixed maintenance" into "preventive maintenance." Unplanned downtime is reduced by more than 80%, and the average monthly ineffective equipment operations are reduced from more than 600 heavy boxes to less than 400 heavy boxes, directly reducing equipment wear and production losses and creating more economic benefits for enterprises.

[0020] 3. Closed-loop quality control to improve product consistency: By using a machine vision module to detect glass defects before cutting and combining edge computing to achieve real-time defect avoidance, the quality control point is shifted from "post-process inspection" to "pre-process intervention". At the same time, through process parameter self-tuning, the cutting quality of glass of different specifications and batches is ensured to be consistent, the product qualification rate is increased by 3%-4%, the generation of waste products is reduced, and the waste of raw materials is reduced.

[0021] 4. Reduce reliance on manual labor and improve process stability: Transform the experience of skilled operators (such as cutting parameters and stacking strategies for different glass thicknesses) into intelligent algorithm models. The models automatically generate optimization decisions, avoiding process fluctuations caused by differences in operator experience and making the production process more stable. At the same time, reduce the difficulty of training new employees, shortening the training cycle from the traditional 1 month to 2 weeks, and reducing the company's human resource training costs.

[0022] 5. The system is highly scalable and adaptable to different needs: The system adopts a modular design, which can flexibly adjust the number of sensors, model parameters and control logic according to the scale of the enterprise's production line (such as an annual output of 3 million or 8 million weight boxes) and product type (such as ordinary float glass, photovoltaic glass, and ultra-clear glass), adapting to the personalized production needs of different enterprises and having broad application prospects. Attached Figure Description

[0023] Figure 1 This is a schematic diagram of the overall architecture of the present invention; Figure 2 This is a flowchart of the intelligent control method of the present invention. Detailed Implementation

[0024] 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.

[0025] Please see Figures 1-2 In this embodiment of the invention, an intelligent control system and method for the cold end of float glass based on the Internet of Things and big data are described. This system, from data acquisition to instruction execution, is divided into a four-layer architecture, with each layer having clearly defined functions and working collaboratively to achieve intelligent control of the entire cold end production process. The specific structure is as follows: (1) Global perception layer The global perception layer, serving as the system's data acquisition source, consists of a sensor network deployed at key nodes of the cold-end production line. It enables comprehensive, real-time perception of glass ribbon quality, operating status, and equipment health, specifically including: Machine vision module: Employing an industrial high-definition camera (resolution no less than 20 megapixels) and image recognition algorithms, this module performs online scanning (scanning frame rate no less than 30fps) on the glass strip before it enters the cutting process. This module can detect and locate defects on the glass strip surface in real time, such as bubbles (minimum identifiable diameter 0.1mm), scratches (minimum identifiable length 1mm, width 0.05mm), and solder adhesion (minimum identifiable area 0.5mm²), while simultaneously recording the location coordinates and size of these defects, providing data support for subsequent cutting scheme optimization.

[0026] Laser velocimetry / length measurement module: Based on the principle of laser Doppler velocimetry, this module accurately measures the real-time running speed (measurement accuracy ±0.01m / min) and cumulative length (measurement accuracy ±0.1mm) of the glass strip. This module can synchronize the glass strip's position information in real time, ensuring that the cutting equipment can adjust the cutting rhythm according to the actual running state of the glass strip, avoiding deviations in cutting length caused by fluctuations in the glass strip's speed.

[0027] Equipment status monitoring sensors: Dedicated sensors are deployed for core components of critical equipment such as the cutting wheel and stacking robot. Among them, vibration sensors (measurement range 0-5000Hz, accuracy ±0.1g) are used to collect vibration data from the cutting wheel spindle and stacking robot joints; temperature sensors (measurement range -40℃-200℃, accuracy ±0.5℃) are used to monitor the operating temperature of the stacking robot bearings and cutting motor; and current sensors (measurement range 0-100A, accuracy ±0.1A) are used to collect the operating current of the cutting motor and stacking robot drive motor. These sensors obtain real-time equipment operating parameters, providing a data foundation for equipment fault prediction.

[0028] (2) Data aggregation and edge processing layer The data aggregation and edge processing layer undertakes the dual functions of "data preprocessing" and "real-time control response," resolving the contradiction between the latency of massive data transmission and the demand for real-time control, and ensuring the system's rapid response to critical production scenarios. Specifically, this includes: IoT Gateway: Employing an industrial-grade IoT gateway (supporting multiple industrial communication protocols such as RS485, EtherNet / IP, and Profinet), it receives massive amounts of data from various sensors in the global sensing layer (data transmission rate up to 100Mbps). The IoT gateway first removes environmental noise (such as workshop vibration interference in vibration data) from the data using a Kalman filter algorithm, then compresses the data volume by more than 50% using the LZ77 compression algorithm to reduce data transmission bandwidth usage. Finally, it converts raw data in different formats (such as analog outputs from sensors and binary format image data) into standard JSON format to ensure data compatibility in subsequent stages.

[0029] Edge computing nodes: Embedded edge computing devices (equipped with quad-core ARM Cortex-A72 processors, with a computing power of 2 TOPS) are used to process real-time data (such as defect image data acquired by machine vision modules) locally. For example, a lightweight target detection algorithm is used to analyze defect images in real time (inference speed up to 50fps), quickly identify the defect type and location, generate defect coordinate information (positioning accuracy ±1mm), and send it directly to the cutting controller to control the cutting equipment to avoid the defect area, prevent defective glass from entering subsequent processes, and reduce waste.

[0030] (3) Cloud platform big data analysis and decision-making layer The cloud platform's big data analysis and decision-making layer is the system's "intelligent core." Through big data storage technology and intelligent algorithm models, it enables equipment fault prediction and production process optimization decisions, providing scientific data support for cold-end production. Specifically, this includes: The data lake, built on the Hadoop Distributed File System (HDFS), stores various types of data uploaded from the edge processing layer, including process parameters (cutting pressure, cutting speed, stacking robot speed, etc.), equipment status data (vibration spectrum, component temperature, motor current, etc.), quality data (defect type, defect quantity, pass rate, etc.), and production order information (glass specifications, thickness, order quantity, quality grade requirements, etc.). The data lake supports petabyte-level data storage and establishes data relationships through key fields such as timestamps and equipment numbers to ensure data traceability, providing a complete data foundation for subsequent model training and data analysis.

[0031] Big Data Analytics Engine: Built on the Spark distributed computing framework, its core function is to construct two interconnected intelligent models—a predictive maintenance model and an adaptive optimization control model. Specific functions are as follows: Predictive Maintenance Model: This model takes vibration spectrum, temperature rise trend, and current fluctuation data collected by equipment condition monitoring sensors as input and is trained using a machine learning algorithm that combines time series analysis (ARIMA model) and anomaly detection (Isolation Forest algorithm). By analyzing historical fault data (such as the vibration frequency change pattern caused by cutter wheel wear and the temperature rise trend before bearing failure), the model can establish the correlation between equipment operating parameters and fault states, providing early warnings of faults such as cutter wheel wear and stacking robot bearing failure 24-72 hours in advance. Simultaneously, it generates specific maintenance recommendations based on the remaining equipment lifespan (e.g., "Cutter wheel remaining lifespan 12 hours, recommended replacement after the current order is completed (estimated 2 hours)"), guiding operators to conduct preventative maintenance.

[0032] Adaptive Optimization Control Model: As the "decision center" of the system, this model receives three types of input information: production order information (glass thickness, specifications, quality grade requirements, order quantity), real-time quality data (defect distribution and defect types detected by the machine vision module), and equipment status information (wear level of the cutting wheel, load rate of the stacking robot, and equipment operating temperature). Through reinforcement learning (DQN deep Q-network) or deep learning (CNN-LSTM hybrid neural network) algorithms, the model comprehensively analyzes the input information and dynamically generates three types of control commands: Intelligent scheduling and cutting optimization instructions: Based on the coordinate information of glass strip defects, the cutting length and cutting position are dynamically adjusted. For example, when there are continuous scratches in a certain area of ​​the glass strip, the model can automatically extend or shorten the cutting length to avoid the defective area, ensuring that the cut glass products are defect-free and maximizing the yield of high-quality glass. Process parameter self-tuning command: Based on changes in glass thickness (e.g., 3mm, 5mm, 8mm) and glass belt running speed, automatically fine-tune the cutting pressure (adjustment range 0.1-0.5MPa) and cutting speed (adjustment range 0.5-2m / min). For example, when producing 3mm ultra-thin glass, automatically reduce the cutting pressure and speed to prevent glass breakage; when producing 12mm thick glass, appropriately increase the cutting pressure to ensure a smooth cut surface. Collaborative scheduling instructions: Based on the cutting rhythm of the cutting equipment (e.g., cutting 10 pieces of glass per minute) and the current load status of the stacking robot (e.g., current stacking layers and remaining stacking space), the optimal stacking mode (e.g., "crisscross stacking" is suitable for thick glass, and "layer stacking" is suitable for thin glass) and path instructions are issued to avoid waiting or congestion of the stacking robot, so as to keep the stacking efficiency and cutting efficiency in match and improve the smoothness of the overall production process.

[0033] (4) Human-computer interaction and execution layer The human-computer interaction and execution layer serves as the interface between the system and operators / production equipment. It enables "visualized display of decisions" and "implementation of optimization instructions," ensuring that system decisions are effectively translated into actual production actions. Specifically, this includes: The intelligent management and control platform, built on web and mobile platforms, provides an intuitive visual interface (supporting 2D / 3D configuration display) to show operators and managers key data for the entire cold-end production process: production progress (order completion rate, real-time output, remaining production tasks), equipment health status (equipment health score, fault warning information, maintenance plan), quality reports (real-time defect rate, finished product pass rate, percentage of different defect types), and optimization decision suggestions (such as "Current cutting pressure is 0.2MPa, it is recommended to adjust it to 0.3MPa to reduce edge breakage rate"). Operators can view data and confirm maintenance plans through the platform; managers can obtain production reports through the platform to assist in formulating production plans and management decisions.

[0034] PLC / Actuator: Utilizing a high-performance industrial PLC (such as the Siemens S7-1500 series) as the control core, it receives non-real-time optimization instructions (such as process parameter self-tuning instructions and intelligent scheduling instructions) from the cloud platform's big data analysis and decision-making layer, or real-time instructions (such as defect avoidance instructions) from the data aggregation and edge processing layer. Through servo motors, pneumatic actuators, and other execution components, it drives equipment such as cutting machines and stacking robots to complete precise movements. The cutting equipment achieves an execution accuracy of ±0.05mm, and the stacking robot's positioning accuracy reaches ±0.1mm, ensuring precise control of the production process.

[0035] 2. Intelligent control method for the cold end of float glass based on the above system This method is based on the aforementioned system and achieves intelligent control of the entire cold-end production process through collaborative work at each layer. The specific steps are as follows: S1: Full-process data acquisition After the system is started, each sensor module in the global perception layer collects data on the entire cold-end production process in real time at a preset frequency: The machine vision module continuously scans the glass strip before it enters the cutting process. Each frame of image acquisition takes no more than 30ms. It simultaneously identifies and records the type of defects on the surface of the glass strip (bubbles, scratches, tin contamination, etc.), the coordinates of the defect location (establishing an XY coordinate system with the starting end of the glass strip as the origin), and the size of the defect. The laser speed / length measurement module updates the real-time running speed and cumulative length data of the glass strip every 100ms to ensure real-time monitoring of the glass strip's movement status. The equipment status monitoring sensor collects equipment operating parameters such as the vibration frequency of the cutting wheel, the temperature of the stacking robot bearing, and the operating current of the cutting motor every 500ms, forming a sequence of equipment status data.

[0036] S2: Data Preprocessing and Real-Time Response The IoT gateway receives raw data collected by various sensors in S1. First, it uses a Kalman filter algorithm to filter environmental interference signals in vibration and temperature data. Then, it uses the LZ77 compression algorithm to compress defect image data from 10MB / frame to less than 5MB / frame to reduce data transmission volume. Finally, it converts all data into standard JSON format to ensure data compatibility. Edge computing nodes perform local real-time analysis on defect image data collected by machine vision modules, identify defect types and locations through lightweight target detection algorithms, generate defect avoidance instructions (such as "do not cut in the glass strip X coordinate 1200-1300mm range"), and send them to the cutting controller within 100ms to control the cutting equipment to adjust the cutting position and achieve defect avoidance. The IoT gateway uploads all pre-processed data (including defect information, glass strip speed / length data, and equipment status data) to the data lake of the cloud platform's big data analysis and decision-making layer via 5G or industrial Ethernet to complete data storage.

[0037] S3: Intelligent Model Calculation The big data analytics engine in the cloud platform's big data analytics and decision-making layer retrieves historical and real-time data from the data lake at preset intervals (e.g., every 5 minutes), while simultaneously running predictive maintenance and adaptive optimization control models. The predictive maintenance model takes into account the vibration spectrum data of the cutting wheel over the past hour and the temperature change trend data of the stacking robot bearing over the past two hours. It uses the ARIMA time series analysis model to predict the remaining life of the core components of the equipment and the isolated forest algorithm to detect whether there are any abnormalities in the current equipment operating parameters (such as a sudden increase in bearing temperature of more than 5°C). The adaptive optimization control model takes into account the current production order information (such as "produce 1000 pieces of 5mm thick, 2440×1830mm glass, quality grade A"), the glass strip defect distribution data collected in S1, and the current operating status data of the equipment (such as the cutting wheel has been used for 8 hours and the wear level is 30%). It iteratively calculates the optimal control parameters through the DQN reinforcement learning algorithm to generate a control instruction set.

[0038] S4: Equipment Failure Warning Based on the calculation results, the predictive maintenance model generates equipment health status reports and outputs them to the intelligent management and control platform. If the equipment operating parameters are normal, the equipment health score (out of 100, a score of ≥80 is considered normal) will be displayed on the intelligent management platform, such as "Cut wheel health score 92, stacking robot health score 88". If abnormal equipment parameters are detected or potential faults are predicted, a fault warning will pop up on the intelligent management platform (e.g., "Abnormal temperature of stacking robot bearing, predicted to fail in 36 hours"). At the same time, detailed maintenance suggestions will be generated (e.g., "It is recommended to replace the bearing during the next order switchover interval (estimated in 4 hours), and the required spare part model is XXX"). The suggestions will also be pushed to the equipment maintenance personnel via the mobile APP.

[0039] S5: Production Decision Optimization The adaptive optimization control model generates the optimal decision-making scheme for the current production scenario based on the comprehensive analysis results, specifically including: Intelligent scheduling and cutting optimization solution: Adjust the cutting length and cutting sequence according to the defect distribution of the glass strip. For example, for glass strips with local defects, the original fixed cutting length of 2440mm is adjusted to 2300mm and 2200mm to avoid the defective area and ensure that all cut glass products are of high quality. Process parameter scheme: Based on the current glass thickness (e.g., 5mm) and glass belt running speed (e.g., 1.5m / min), determine the optimal cutting pressure (e.g., 0.3MPa) and cutting speed (e.g., 1.2m / min) to ensure cutting quality; Stacking scheduling scheme: Based on the cutting rhythm of the cutting equipment (e.g., cutting 8 pieces of glass per minute) and the current load status of the stacking robot, the stacking mode is determined to be "crisscross stacking", and the shortest stacking path is planned to avoid the robot running idle.

[0040] S6: Optimize command issuance The cloud platform's big data analysis and decision-making layer sends non-real-time optimization instructions (such as process parameter schemes, intelligent scheduling and cutting optimization schemes) to the PLC via industrial Ethernet; The data aggregation and edge processing layer directly sends real-time instructions (such as temporary defect avoidance instructions) to the cutting controller; After receiving the instruction, the PLC parses and converts the instruction to generate control signals that the equipment can recognize (such as pulse signals of servo motors and pneumatic control signals of pneumatic actuators).

[0041] S7: Execution and Closed-Loop Optimization The PLC sends control signals to the actuators (servo motors, pneumatic actuators, etc.) to drive the cutting machine to cut according to the optimized parameters, and the stacking robot to perform stacking operations according to the optimized stacking mode and path. During execution, the actuator collects the actual operating parameters of the equipment in real time (such as actual cutting pressure and actual operating speed of the stacking robot) and feeds these execution results back to the data lake of the cloud platform's big data analysis and decision-making layer; The cloud platform uses big data analysis and the decision-making layer to call feedback data, compare it with the optimal parameters generated by the model, calculate the deviation (e.g., the deviation between the actual cutting pressure of 0.29MPa and the optimal value of 0.3MPa is 0.01MPa), and adjust the control commands in the next model calculation to achieve dynamic optimization of the production process, forming a closed-loop system of "collection-analysis-decision-execution-feedback-optimization". Example 1

[0042] To verify the practical application effect of this invention, a three-month application test was conducted on the cold-end production line (annual production of 5 million weight boxes of float glass) of a large float glass manufacturing enterprise. During the test, the control system and method of this invention were used, and the production data were compared with those before the test (using a traditional cold-end system). The specific implementation process and results are as follows: 1. Implementation Preparation Equipment deployment: Along the glass conveyor path on the cold end production line, deploy 2 industrial high-definition cameras (24 million pixels resolution) as machine vision modules, 1 laser speed / length measuring device (measurement accuracy ±0.01m / min), 8 vibration sensors at key parts such as the cutting wheel spindle and stacking robot joints, 6 temperature sensors at parts such as the stacking robot bearings and cutting motors, and 4 current sensors next to the cutting motor and stacking robot drive motor; System setup: Deploy 2 industrial IoT gateways (supporting multi-protocol conversion), 1 edge computing node (equipped with a quad-core ARM Cortex-A72 processor), build a cloud platform based on Alibaba Cloud servers (using HDFS data lake and Spark analysis engine), and develop an intelligent management and control platform (supporting web and mobile access). Model training: Collect equipment failure data (such as wear of cutting wheel, failure of stacking robot bearings and other failure records and corresponding equipment parameters) and quality data (defect type, cutting pass rate, and yield) of the production line in the past year, and train the predictive maintenance model and the adaptive optimization control model to achieve a failure prediction accuracy of 96% and a cutting scheme optimization accuracy of 93%.

[0043] 2. Implementation process During the testing period, the production line mainly produced float glass of three commonly used thicknesses: 3mm, 5mm, and 8mm, and the control method of this invention was used for production. Data acquisition: The machine vision module collects approximately 50GB of defect image data of the glass strip daily; the laser velocity / length measurement module updates the velocity and length data every 100ms; and the equipment status sensor collects vibration, temperature, and current data every 500ms to ensure the comprehensiveness and real-time nature of the data acquisition. Edge processing: The average analysis time of the edge computing node for defect image data is 80ms. After generating the defect avoidance command, it is sent to the cutting controller within 90ms, realizing the precise avoidance of defective glass. No defective glass cutting problems caused by command delay occurred. Cloud analytics: The cloud platform processes approximately 200GB of various types of data daily, and the predictive maintenance model generates one equipment health report daily. During the test, three early warnings of faults were issued (all of which were due to wear of the cutting wheel). Maintenance personnel replaced the cutting wheel between orders based on the recommendations, and no unplanned downtime occurred. Optimized execution: The adaptive optimization control model dynamically adjusts the cutting parameters according to different glass thicknesses. For example, when producing 3mm ultra-thin glass, the cutting pressure is automatically reduced from the traditional 0.4MPa to 0.2MPa, and the cutting speed is reduced from 1.8m / min to 1.0m / min, effectively avoiding edge chipping during cutting. When producing 8mm thick glass, the cutting pressure is adjusted to 0.4MPa, and the cutting speed is adjusted to 1.5m / min to ensure a smooth cut surface.

[0044] 3. Implementation Results Compared to before testing (using a traditional cold-end system), the production data after adopting this invention was significantly optimized, as shown in the table below:

[0045] As can be seen from the above data, the present invention effectively achieved the goal of "monthly average equipment ineffective operation of less than 400 heavy boxes", while significantly improving the glass output rate and product qualification rate, greatly reducing unplanned downtime, reducing reliance on manual experience, and verifying the practicality and advanced nature of the present invention in actual production.

[0046] The above description is only a preferred embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any equivalent substitutions or modifications made by those skilled in the art within the scope of the technology disclosed in the present invention, based on the technical solution and inventive concept of the present invention, should be covered within the scope of protection of the present invention.

Claims

1. An intelligent control system for the cold end of float glass based on the Internet of Things and big data, characterized in that, It comprises a global perception layer, a data aggregation and edge processing layer, a cloud platform big data analysis and decision-making layer, and a human-computer interaction and execution layer. The global perception layer is deployed at key nodes in the cold end to collect glass strip quality information, operating parameters, and equipment status data. The data aggregation and edge processing layer receives data from the global perception layer, performs preprocessing and real-time local data processing, and uploads the data to the cloud platform big data analysis and decision-making layer. The cloud platform big data analysis and decision-making layer stores the data and generates decision instructions through intelligent models. The human-computer interaction and execution layer displays data and decision suggestions, and executes decision instructions to drive equipment operation.

2. The intelligent control system for the cold end of float glass based on the Internet of Things and big data as described in claim 1, characterized in that, The global perception layer includes a machine vision module, a laser speed / length measurement module, and equipment status monitoring sensors. The machine vision module is used to scan the glass strip online before cutting to detect and locate defects such as bubbles, scratches, and solder adhesion. The laser speed / length measurement module is used to accurately measure the running speed and length of the glass strip. The equipment status monitoring sensors include vibration sensors, temperature sensors, and current sensors, which are used to collect the operating parameters of the cutting wheel and the stacking robot.

3. The intelligent control system for the cold end of float glass based on the Internet of Things and big data as described in claim 1, characterized in that, The data aggregation and edge processing layer includes an IoT gateway and edge computing nodes; the IoT gateway filters, compresses, and performs protocol conversion on the massive data collected by the global perception layer; the edge computing nodes perform local processing on data with high real-time requirements, including analyzing images collected by the machine vision module, identifying defects and generating defect coordinate information, and sending it to the cutting controller to achieve defect avoidance.

4. The intelligent control system for the cold end of float glass based on the Internet of Things and big data as described in claim 1, characterized in that, The cloud platform's big data analysis and decision-making layer includes a data lake and a big data analysis engine; the data lake stores historical and real-time data uploaded from the edge processing layer, and the data types cover process parameters, equipment status data, quality data, and production order information; the big data analysis engine is used to build predictive maintenance models and adaptive optimization control models.

5. The intelligent control system for the cold end of float glass based on the Internet of Things and big data as described in claim 4, characterized in that, The predictive maintenance model is trained using machine learning algorithms such as time series analysis and anomaly detection, based on vibration spectrum and temperature rise trend data collected by equipment condition monitoring sensors. It is used to provide early warning of faults such as cutting wheel wear and bearing failure, and to generate maintenance suggestions. The adaptive optimization control model receives production orders, real-time quality data and equipment status information, and generates control commands through reinforcement learning or deep learning algorithms.

6. The intelligent control system for the cold end of float glass based on the Internet of Things and big data as described in claim 5, characterized in that, The control commands generated by the adaptive optimization control model include intelligent scheduling and cutting optimization commands, process parameter self-tuning commands, and collaborative scheduling commands; the intelligent scheduling and cutting optimization commands adjust the cutting scheme according to defect location information. The process parameter self-tuning command finely adjusts the cutting pressure and speed parameters according to changes in glass thickness and speed. The collaborative scheduling command sends stacking mode and path instructions to the stacking robot.

7. The intelligent control system for the cold end of float glass based on the Internet of Things and big data as described in claim 1, characterized in that, The human-machine interaction and execution layer includes an intelligent control platform and a PLC / actuator; the intelligent control platform provides a visual interface to display real-time data of the entire process, equipment health status, quality reports and optimization decision suggestions; the PLC / actuator receives optimization instructions from the cloud platform's big data analysis and decision layer or data aggregation and edge processing layer to drive the cutting machine and stacking robot to complete the actions.

8. A method for intelligent control of the cold end of float glass based on the system described in any one of claims 1-7, characterized in that, Includes the following steps: S1: Through the machine vision module, laser speed / length measurement module and equipment status monitoring sensor of the global perception layer, information on glass strip defects, running speed and length data and equipment operating parameters are collected respectively; S2: The IoT gateway of the data aggregation and edge processing layer filters, compresses and converts the data collected by S1. The edge computing node analyzes the glass strip defect information locally in real time and issues instructions. At the same time, all the processed data is uploaded to the data lake of the cloud platform big data analysis and decision-making layer. S3: The big data analysis engine of the cloud platform's big data analysis and decision-making layer calls data lake data to run predictive maintenance models and adaptive optimization control models; S4: The predictive maintenance model outputs equipment health status warnings and maintenance suggestions to the intelligent management and control platform; S5: The adaptive optimization control model integrates production orders, real-time quality data, and equipment status information to generate the optimal cutting scheme and equipment control parameter set; S6: Send the control parameter set generated in S5 to the PLC / actuator; S7: The PLC / actuator drives the equipment to complete the production operation and feeds the execution results back to the data lake to form a closed-loop optimization.

9. The intelligent control method for the cold end of float glass according to claim 8, characterized in that, The predictive maintenance model described in S3 determines the wear level of equipment components by analyzing vibration spectrum, identifies abnormal heating of equipment by tracking temperature rise trends, and optimizes the early warning accuracy by combining historical fault data; the adaptive optimization control model prioritizes avoiding defective areas when generating cutting schemes to ensure that the yield of high-quality glass is maximized.

10. The intelligent control method for the cold end of float glass according to claim 8, characterized in that, The execution results in S7 include actual equipment operating parameters, glass cutting quality inspection results, and stacking completion status; the feedback data is stored in the data lake and used to update the training data of the predictive maintenance model and the adaptive optimization control model, thereby improving the accuracy of model decision-making.

Images