Intelligent control system for waterproof composite material production

Abstract

The invention discloses a waterproof composite material production intelligent control system, and relates to the technical field of control systems, the system comprises a data acquisition layer, a standard template generation module, a data processing layer, a Transform-based model core layer and an output layer, the data acquisition layer acquires production data in real time; a data processing standard template is generated through a standard template generation module, standardization processing is conducted on real-time data through a data processing layer, production process modeling and production state analysis are conducted through a Transform model core layer, an analysis result is decoded into a control instruction through an output layer, and production line process parameters are adjusted. The production process is intelligently and optimally controlled by applying the artificial intelligence technology, the intelligent level, the operation efficiency and the product quality of the production line are remarkably improved, meanwhile, the production goals of energy conservation, efficiency improvement and intrinsic safety are achieved through refined management and control and whole-process safety protection, and intelligent and green development of the waterproof composite material industry is promoted.

Description

Technical Field

[0001] The invention relates to the technical field of control systems, in particular to an intelligent control system for the production of waterproof composite materials. Background Art

[0002] With the rapid development of the construction industry, the demand for waterproof materials continues to increase. In the traditional production process of waterproof composite materials, process parameter setting and quality control mainly rely on manual experience, which is difficult to adapt to the increasing production requirements. In order to improve the production efficiency and product quality of waterproof composite materials, it is urgent to develop an intelligent production control system.

[0003] Chinese invention patent CN116977116B discloses an intelligent monitoring and management system for waterproof material production based on an intelligent manufacturing factory. The system ensures that the quality of waterproof coating products meets national standards by real-time monitoring and data analysis of raw material quality, production process, environmental factors, etc. However, the system mainly focuses on monitoring and quality assessment of the production process, and lacks in-depth research on how to perform intelligent process optimization and control based on monitoring data.

[0004] Chinese invention patent CN114690730B discloses a method and system for automatically controlling process parameters in a composite material production process. The system is based on image processing technology and analyzes the degree of deformation of composite materials products to select the best combination of control parameters to achieve automatic control of the laying process. However, the system mainly optimizes local parameters of the composite material laying process and lacks systematic consideration of the entire production process, making it difficult to comprehensively improve production efficiency and product quality.

[0005] The above design has achieved intelligent control of the production process to a certain extent by introducing intelligent monitoring and image analysis technology, but it still has certain limitations, such as the lack of global modeling and analysis capabilities that integrate multi-source heterogeneous data, insufficient depth in exploring the process mechanism, and a relatively simple control method, making it difficult to adapt to complex and changing production scenarios. Summary of the invention

[0006] The purpose of the present invention is to overcome the deficiencies of the prior art and to provide an intelligent control system for the production of waterproof composite materials to solve the above-mentioned problems.

[0007] The present invention integrates multi-source sensor data, uses standard templates to achieve standardized processing of heterogeneous data, and innovatively introduces the Transformer deep learning model to perform end-to-end modeling of the entire production process, explore the inherent mechanism of the production process, and form an optimal control strategy. At the same time, the system also integrates a complete safety protection module, which builds a safety line through real-time monitoring and intelligent early warning of multi-dimensional parameters to ensure the smooth operation of the system. It can improve the intelligence level of the waterproof composite material production line, fully explore the value of production data, optimize key process parameters, reduce the loss of defective products, improve product quality, achieve the goals of energy saving, efficiency improvement and safe production, and accumulate expert experience at the same time.

[0008] The object of the present invention is achieved through the following technical solutions: an intelligent control system for waterproof composite material production, comprising: a data acquisition layer, a standard template generation module, a data processing layer, a Transformer-based model core layer and an output layer; The data acquisition layer is used to collect production line sensor data, which includes process parameters, equipment parameters, material parameters and quality parameters; The standard template generation module is used to analyze historical production data and generate data processing standard templates. The standard templates include parameter normalization templates, feature extraction templates, and sequence construction templates. The data processing layer is used to process real-time data using standard templates to generate feature Token sequences; The core layer of the Transformer-based model is used for sequence processing and analysis; The output layer is used to execute control commands and display results; The system's working steps include: S1: collect sensor data through the data acquisition layer; S2: Analyze historical production data through the standard template generation module to generate data processing standard templates, which include parameter normalization templates, feature extraction templates and sequence construction templates; S3: The data processing layer uses standard templates to process real-time data and generate feature token sequences; S4: Sequence processing and analysis through the core layer of the Transformer-based model, including: S41: Encoding the feature token sequence: mapping the feature vector to an embedding vector of fixed dimension, adding feature type embedding, generating position encoding and adding it to the feature embedding; S42: Processing sequence relations through multi-head attention mechanism: generating query vector (Q), key vector (K) and value vector (V), calculating attention score, normalizing attention weights through softmax, and weighted aggregation value vector; S43: Capture different feature relationships through a multi-head mechanism: compute h attention heads in parallel, each head focuses on a different feature combination, concatenates the multi-head outputs and performs linear transformation; S44: Processing feature interactions through feed-forward networks: nonlinear transformation of attention outputs, adding residual connections and layer normalization; S45: Generate feature representation required for decision making; S5: Execute control commands through the output layer and display the results, including: S51: Adjust the glue coating and sanding process parameters according to the Transformer output; S52: Generate real-time control instructions and warning information; S53: updating the process parameter optimization strategy; S54: Feedback optimization results for template updating.

[0009] The process parameters include glue coating temperature, pressure, and flow rate; the equipment parameters include motor operating status, power, and speed; the material parameters include hot melt adhesive temperature, viscosity, and glue supply; and the quality parameters include sand layer uniformity and adhesion.

[0010] The processing steps of the standard template generation module include: clustering analysis of historical data to identify characteristic patterns of different product specifications and material combinations; generating data preprocessing standard templates for each type of product based on the clustering results; establishing standardized conversion templates for different process parameter combinations; generating standard templates for time series feature extraction; classifying and storing verified templates, and establishing a template index library.

[0011] The steps of processing in the data processing layer using the standard template include: standardizing parameters based on the normalization template; applying the feature extraction template to perform feature engineering; constructing a feature sequence using the sequence construction template; and organizing the processed features into a Token sequence.

[0012] The system uses different feature extraction methods for different types of parameters: temperature parameters: mean, variance, rate of change; pressure parameters: peak value, valley value, fluctuation range; speed parameters: acceleration, stability index; quality parameters: consistency index, deviation value; equipment state parameters: state transition frequency and duration.

[0013] The system also includes a production line control module, which includes: Glue supply control unit, used to realize automatic adjustment of hot melt glue supply quantity, glue supply pressure and temperature control, and glue supply uniformity monitoring; Temperature control unit, used to realize independent temperature control of multiple sections of coating head, automatic temperature compensation, and heating power adjustment; Sand control unit, used to control the thickness and uniformity of the sand layer, automatically replenish the sand, and control the sand return system; Tension control unit, used to realize automatic adjustment of winding and unwinding tension, deviation correction control, and speed synchronization control; The specification switching unit is used to realize automatic identification of product specifications, automatic switching of process parameters, and automatic loading of production parameters.

[0014] The core layers of the Transformer-based model include: input embedding layer, which is used to convert tokens into vector representations; position encoding layer, which is used to add position information; multi-layer attention layer, which is used to process sequence relationships; feedforward network layer, which is used for feature transformation and interaction; normalization layer, which is used for data standardization processing.

[0015] The system also includes a security protection module, which includes: Temperature safety unit, used to realize over-temperature alarm, automatic temperature reduction control, and temperature emergency treatment; Pressure safety unit, used to realize abnormal pressure alarm, automatic pressure relief control, and pressure compensation adjustment; Glue supply safety unit, used to realize glue supply abnormality detection, automatic shutdown control, and glue supply safety interlock; Equipment protection unit, used to realize motor overload protection, equipment status monitoring, and abnormal operation alarm; Emergency processing unit, used to realize emergency shutdown control, safety interlock control, and fault diagnosis record; The security protection module communicates with the model core layer to achieve real-time monitoring and early warning control.

[0016] The beneficial effects of the present invention are: 1. Improve the level of intelligent production: The system deploys multi-source heterogeneous sensors on the production line to collect process parameters, equipment parameters, material parameters and quality parameters in real time, and uses the Transformer deep learning model to perform end-to-end modeling and analysis on massive production data, explore process mechanisms, and optimize control strategies.

[0017] 2. Significantly improve product quality: The system makes full use of multi-parameter sensors and intelligent algorithms to achieve full-process monitoring of key factors affecting product quality. Through intelligent analysis of real-time data streams, the system can independently identify process anomalies, quality defects, etc., and eliminate process deviations in the bud in a timely manner through closed-loop feedback control, thereby ensuring the stability of product quality from the source.

[0018] 3. Quick response to market demand: The system has flexible production capabilities and can achieve rapid switching of production varieties and specifications. By connecting to the MES system, it can automatically sense order changes, call the process formula library, intelligently generate product switching plans, and accurately control key processes such as gluing and drying to ensure product quality consistency and quickly respond to customized market demands.

[0019] 4. Reduce losses from defective products: Traditional production models make it difficult to perceive process anomalies in real time, and problematic products are discovered late, resulting in a large number of defective products flowing into the next process, causing heavy losses. The system uses the Transformer model to perform correlation analysis on multi-source process data collected in real time, and accurately warns of quality defects through abnormal feature recognition, thereby minimizing the outflow of defective products and reducing quality losses.

[0020] 5. Energy saving and consumption reduction: The system applies data-driven intelligent decision-making technology to autonomously optimize energy consumption costs while meeting process quality requirements. It senses the operating status of equipment in real time, analyzes energy efficiency curves, intelligently schedules production tasks, smooths out peaks and fills valleys, actively reduces loads during energy shortages, and increases production loads during electricity price depressions, thereby achieving refined control of energy consumption and facilitating green manufacturing.

[0021] 6. Improve the overall efficiency of equipment: The system integrates comprehensive equipment status monitoring functions, applies vibration, oil, current and other sensors to collect equipment health data in real time. The Transformer model analyzes big data, learns the equipment degradation rules, predicts potential failures of key components in advance, guides preventive maintenance, reduces unplanned downtime, and improves the overall operation efficiency of the production line.

[0022] 7. Enhanced safety assurance: This system fully considers production safety risks and specially designs a full-chain safety protection plan. Through key monitoring and interlocking of dangerous processes and rapid response and isolation of safety incidents, a three-dimensional safety barrier is built. The integration of safety management and production control embeds the safety concept into the production decision-making logic, and safety KPI becomes a hard constraint for the optimization of the entire production process.

[0023] 8. Accumulation of expert experience: The standard template generation module of the system deconstructs complex production processes and extracts standardized data processing procedures. Experts can solidify their experience and knowledge into software algorithms through parameter configuration and logic editing, making production know-how reusable and inheritable. Data-driven knowledge accumulation avoids experience gaps caused by staff loss and provides a solid foundation for the sustainable development of the enterprise.

[0024] 9. Empowering lean management: The system's refined control at the production execution layer provides a digital foundation for the company's lean management. Through statistical analysis of production data, various KPI indicators can be quickly calculated, including OEE, FTT, defective rate, energy consumption, cost, etc., and presented intuitively through visualization methods, providing a basis for scientific decision-making by management. Production digital twins provide broad space for management innovation.

[0025] 10. Easy to promote and apply: This system adopts universal standard industrial communication protocols and interface specifications, has no special requirements for hardware equipment and data formats, has strong adaptability and compatibility, and the system architecture adopts modular design. The functional units can be flexibly combined and easy to customize according to actual scenarios. The control algorithm is implemented based on an open source deep learning framework. Through transfer learning, it can be quickly transplanted to other similar processes, thereby greatly improving the promotion and application value of the system. BRIEF DESCRIPTION OF THE DRAWINGS

[0026] Figure 1 It is a system interaction diagram of the present invention; Figure 2 It is a working step diagram of the present invention; Figure 3 It is a system architecture diagram of the present invention. DETAILED DESCRIPTION

[0027] The technical solution of the present invention will be clearly and completely described below in conjunction with the embodiments. Obviously, the described embodiments are only part of the embodiments of the present invention, rather than all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative work are within the scope of protection of the present invention.

[0028] It is to be noted that the concepts of “left”, “right”, “up”, “down”, “front”, “back”, “inside” and “outside” in the following schemes are all relative directions and will not be listed one by one here.

[0029] Example 1 like Figures 1 to 3 As shown, this embodiment provides an intelligent control system for waterproof composite material production, which includes a data acquisition layer, a standard template generation module, a data processing layer, a Transformer-based model core layer and an output layer.

[0030] The data acquisition layer connects various sensors on the production line to collect production process data in real time. Among them, the process parameter sensor collects data such as glue coating temperature, pressure, flow, etc. The equipment parameter sensor collects data such as motor operating status, power, speed, etc. The material parameter sensor collects data such as hot melt adhesive temperature, viscosity, glue supply, etc. The quality parameter sensor collects data such as sand layer uniformity and adhesion.

[0031] The standard template generation module regularly analyzes and mines historical production data to generate a series of templates to guide real-time data processing. First, cluster analysis is performed on historical data to identify the characteristic patterns corresponding to different product specifications and material combinations. Then, based on the clustering results, standard templates for data preprocessing are generated, including parameter normalization templates, feature extraction templates, and sequence construction templates. The normalization template defines the standardization method for different parameters, the feature extraction template defines the calculation method for time series features, and the sequence template defines the construction rules for feature sequences. Finally, the verified templates are classified and stored, and an index library is established for subsequent calls.

[0032] The data processing layer is responsible for converting the collected real-time data into a standardized feature Token sequence. Specifically, it first performs dimensionless standardization on various parameters based on the normalization template. Then, it calculates the time series features according to the feature extraction template, including extracting the mean, variance, and rate of change for temperature parameters, extracting the peak and valley values ​​and fluctuation range for pressure parameters, extracting acceleration and stability indicators for speed parameters, extracting consistency indicators and deviation values ​​for quality parameters, and extracting state transition frequency and duration for equipment state parameters. Then, it uses the sequence construction template to generate a standard Token sequence representation of multi-source heterogeneous features.

[0033] The core layer of the Transformer-based model uses an attention mechanism and a feedforward neural network to perform end-to-end modeling and analysis of feature token sequences. First, the input embedding layer maps discrete tokens into continuous vector representations, and the position encoding layer incorporates sequence position information. Then, the multi-layer attention layer calculates the three matrices of Query, Key, and Value to mine the correlation between tokens and dynamically adjust feature weights. Next, the feedforward network layer further extracts high-order interaction features. Finally, the normalization layer performs normalization and outputs the state representation vector required for decision making.

[0034] The output layer generates optimization control instructions in real time based on the output of Transformer. On the one hand, it sends the temperature, pressure and other process parameter adjustment instructions to the glue coating and sand placing equipment for execution. On the other hand, it generates early warning prompts based on quality deviations, abnormal characteristics, etc., triggering the emergency disposal process. At the same time, it continuously tracks the changing trends of process parameters and quality data, dynamically updates the optimization strategy, and finally feeds back the optimization results to the standard template module to trigger the iterative update of the template. The system's workflow is as follows: The data acquisition layer collects process parameters, equipment parameters, material parameters and quality parameters in real time.

[0035] The standard template generation module analyzes historical production data offline to generate templates for parameter normalization, feature extraction, and sequence construction. The data processing layer uses standard templates to clean and transform real-time data and output standardized feature Token sequences.

[0036] The core layer of the Transformer-based model encodes the token sequence and extracts key features through the attention mechanism and feedforward network.

[0037] The output layer decodes the output vector of Transformer and generates real-time control instructions, warning information and optimization strategies.

[0038] The process parameter adjustment instructions are sent to the execution equipment, the optimization model is updated synchronously, and the feedback results are used to iterate the template.

[0039] Digitalization of the entire process: Comprehensive collection of production factors through sensors, transforming experience-driven into data-driven, Adaptive optimization: Continuously analyze production data, automatically generate and dynamically update optimization strategies, so that the control system can adapt to changes in products and processes, Timely detection of anomalies: Fully explore the associations between multi-source heterogeneous data, and timely warn and deal with various hidden anomalies, Strong generalization ability: End-to-end modeling based on attention mechanism and feedforward network can flexibly adapt to different production scenarios and has stronger generalization ability, Expert experience precipitation: Through the standard template mechanism, expert experience is solidified into reusable data processing rules to achieve experience precipitation and inheritance.

[0040] Example 2 like Figures 1 to 3 As shown, based on Example 1, this embodiment further provides an intelligent control system for the production of waterproof composite materials, which also includes a production line control module.

[0041] The production line control module consists of multiple control units, which are responsible for each link of the production process. The glue supply control unit is connected to the hot melt glue supply system. By adjusting the speed of the glue pump and the opening of the glue valve, the glue supply amount can be automatically adjusted. At the same time, the glue supply unit integrates temperature sensors and pressure sensors to monitor the hot melt glue temperature and glue supply pressure in real time. The heating device and pump pressure are controlled by PID algorithm to ensure the stability of glue supply temperature and pressure. The uniformity of glue supply is monitored by multi-point flow meters, and the feedback signal is introduced into the closed-loop control of glue supply to ensure the consistency of glue distribution during the coating process. The temperature control unit performs zone control on the coating head. The coating head is divided into multiple control sections. Each section is equipped with an independent temperature control system, which is composed of precision temperature sensors, heaters, radiators, and PID controllers. The temperature of each section can be independently set and adjusted according to process requirements. At the same time, the temperature control unit also has an automatic compensation function. When the actual temperature deviates from the target temperature, the controller will automatically calculate the compensation amount and correct the temperature curve by adjusting the heating power to ensure uniform coating temperature.

[0042] The sand placing control unit is responsible for laying the sand layer. The unit integrates a laser rangefinder to measure the thickness distribution of the sand layer in real time and feed back the data to the controller. The controller realizes dynamic control of the thickness of the sand layer by adjusting the rotation speed and vibration frequency of the sand cloth roller. When the local sand layer is uneven, the system automatically identifies the weak area and replenishes sand to ensure the uniformity of the final sand layer thickness. At the same time, the sand placing unit is also equipped with a sand quantity monitoring device and an automatic sand replenishing mechanism, which can track the sand bin inventory in real time. When the sand quantity is insufficient, automatic sand replenishment is triggered, and the sand is automatically transported from the sand reservoir to the sand bin to achieve continuity of sand supply.

[0043] The tension control unit is used to adjust the winding and unwinding tension of the substrate. Multiple tension sensors are arranged at the unwinding, gluing and drying stations to collect substrate tension data in real time. The controller adjusts the speed and torque of the unwinding / winding roller through the servo motor based on the feedback information to achieve dynamic balance of tension. At the same time, the unit also integrates the correction control function. The edge position sensor is used to detect the lateral displacement of the substrate. The correction actuator automatically adjusts the angle of the unwinding roller according to the offset and corrects the substrate to the set trajectory. The operating speed of the gluing machine and the dryer is synchronized through the frequency conversion speed regulation mechanism to ensure the stability of coating thickness control.

[0044] The specification switching unit is used to cope with the rapid switching of production varieties. The unit is connected to the MES system and automatically downloads product BOM information. According to the BOM requirements, the system automatically identifies the product specifications and retrieves the corresponding process parameters from the formula database. The production line's glue flow, temperature curve, drying time and other parameters are automatically switched to the new specifications. The entire switching process does not require human intervention. The automatic loading of production parameters greatly improves production flexibility and shortens variety switching time.

[0045] When the system is working, the data acquisition layer collects production data in real time through the sensor network and transmits it to the data processing layer. The processing layer cleans and extracts features from the original data to generate standardized time series feature Tokens. The core layer based on Transformer performs sequence modeling and decision analysis on the feature Tokens. The output layer decodes the decision results into control instructions and sends them to the production line control module. The control unit accurately controls each actuator according to the optimal control strategy, and collects process parameters in real time through sensors to form a closed-loop feedback and continuously iterate and optimize the control model.

[0046] The control granularity of the production line is refined to each process unit. The process parameters are accurately adjusted through an independent control system to ensure the stability and consistency of the entire production process. The system perceives the production status in real time through sensors. The Transformer model autonomously learns the mapping process mechanism, adaptively compensates for interference factors, and outputs the optimal control strategy to improve control accuracy and efficiency. The control units are coordinated and optimized, and unified scheduling is achieved through real-time data interaction, dynamic update of control strategies, and multi-objective optimization to achieve intelligent collaborative control of the entire line production. The specification switching unit supports flexible manufacturing, realizes automatic switching and rapid loading of product BOM and recipe parameters, minimizes downtime, and improves production efficiency. While production is being executed, each unit uses sensors to collect process parameters in real time, forming a closed-loop data flow, providing data support for production optimization, and realizing the digitalization of process improvement and quality control.

[0047] Example 3 like Figures 1 to 3 As shown, based on Examples 1 and 2, this example further provides an intelligent control system for the production of waterproof composite materials, which also includes a safety protection module.

[0048] The safety protection module consists of multiple safety units, which are responsible for temperature safety, pressure safety, glue supply safety, equipment protection and emergency handling in the production process. Each unit collects production data in real time through sensors and communicates with the core layer of the Transformer-based model in real time to achieve all-round safety monitoring and early warning control of the production process.

[0049] The temperature safety unit is deployed at the gluing station and the drying station. It collects equipment and material temperature data in real time through high-precision thermocouples. When the actual temperature exceeds the safety threshold, the unit immediately triggers an audible and visual alarm and sends an over-temperature signal to the control system. The control system responds to the alarm and executes the preset cooling strategy to quickly bring the temperature back to a safe range by adjusting the heating power or increasing the coolant flow. In an emergency, the temperature safety unit can also directly intervene in the temperature control loop to forcibly cut off the heating power supply to avoid major accidents caused by temperature out of control.

[0050] The pressure safety unit is connected to the glue coating pump station and the hot melt adhesive pipeline, and monitors the pipeline pressure in real time through the pressure transmitter. When the pressure in the pipe fluctuates or exceeds the safety upper limit, the unit issues an abnormal alarm and sends a pressure message to the control system. The control system determines the abnormal pressure level and performs pressure relief or pump stop operations according to the plan to control the pipeline pressure within the safe working range. When the glue supply pressure is insufficient, the pressure safety unit will also link the glue supply control unit to increase the pump speed or adjust the glue supply valve to compensate the pressure to the optimal working point while ensuring the coating quality.

[0051] The glue supply safety unit is connected to the glue storage tank and the glue supply pipeline, and senses the glue inventory and flow status in real time through the liquid level meter and flow meter. When it is found that the glue remaining amount is insufficient or the pipeline is blocked, the unit immediately alarms the control system and automatically cuts off the power supply of the glue pump to avoid damage to the equipment due to dry pump operation. In the event of an emergency shutdown, the glue supply safety unit will also synchronously close the feed valve to prevent the glue in the pipe from flowing back. The entire interlocking process is reliably executed by the PLC programmable controller to ensure the inherent safety of the glue supply system.

[0052] The equipment protection unit mainly monitors the operating status of key transmission components, monitors the motor current through current sensors, and determines whether overload or stall occurs. At the same time, vibration sensors collect the vibration spectrum of bearings and gearboxes in real time. The expert system conducts intelligent analysis of the spectrum characteristics to detect hidden dangers such as mechanical wear or poor lubrication as early as possible. When the monitoring indicators exceed the standard, the equipment protection unit promptly alerts the central control system and reduces the mechanical load or shuts down for maintenance according to the strategy to minimize unplanned equipment shutdowns.

[0053] The emergency handling unit prepares for various emergencies and provides rapid handling solutions at the system level. When an emergency such as fire or power outage occurs at the production site, the unit immediately activates the emergency plan, cuts off the power supply to all equipment, closes the valves, and notifies relevant personnel via text messages and voice calls. The emergency handling unit also integrates safety interlock control logic, and connects the safety switches of workstations such as gluing, drying, and traction in series in the emergency stop circuit. Any switch action will trigger an emergency stop of the entire line to prevent the accident from expanding. At the same time, the emergency handling unit also has fault diagnosis and record storage functions. The expert system analyzes historical data, summarizes the laws of accidents, forms prevention and control strategies, and guides production prevention.

[0054] The safety protection module communicates with the core layer of the Transformer-based model in real time. Each safety unit uploads the collected temperature, pressure, liquid level, vibration and other data to the core layer in real time and integrates them into the feature Token sequence. Transformer uses attention mechanism and timing analysis to quickly identify the changing trends of various process parameters and predict dangerous conditions in advance. When the safety KPI approaches the critical value, an early warning is triggered and the control instructions output by the model are called to guide timely on-site disposal. At the same time, the model can also correlate and analyze the abnormal data of each safety unit, infer the fault propagation path, and provide a complete causal logic chain for emergency disposal.

[0055] The safety protection module is connected to the production line, and through multi-parameter sensors, it can realize active perception and real-time monitoring of dangerous factors, prevent safety accidents in advance, and coordinate various safety units to break through the limitations of single-machine protection, improve the inherent safety level from the system level, deeply integrate safety strategies and production control, and the safety protection data stream is converged to the Transformer core in real time, integrating process status parameters to build a global safety portrait, and assisting the management and control layer to make the best decision. The emergency handling unit provides a standard plan library and integrates multi-level interlocking logic. It can quickly isolate the source of danger, block the spread of accidents, and enhance the system's rapid recovery capabilities in response to emergencies. By recording the entire process of safety incidents and handling processes, it accumulates experience data for iterative optimization of safety strategies and early warning models, realizes a closed loop of safety management, and continuously improves the inherent safety level.

[0056] The above is only a preferred embodiment of the present invention. It should be understood that the present invention is not limited to the form disclosed herein and should not be regarded as excluding other embodiments. It can be used in various other combinations, modifications and environments, and can be modified within the scope of the concept described herein through the above teachings or the technology or knowledge of the relevant field. The changes and modifications made by those skilled in the art do not depart from the spirit and scope of the present invention, and should all be within the scope of protection of the claims attached to the present invention.

Claims

1. An intelligent control system for the production of waterproof composite materials, characterized in that: include: Data collection layer, standard template generation module, data processing layer, Transformer-based model core layer and output layer; The data acquisition layer is used to collect production line sensor data, and the sensor data includes process parameters, equipment parameters, material parameters and quality parameters; The standard template generation module is used to analyze historical production data and generate a data processing standard template, wherein the standard template includes a parameter normalization template, a feature extraction template and a sequence construction template; The data processing layer is used to process the real-time data using the standard template to generate a feature Token sequence; The core layer of the Transformer-based model is used for sequence processing and analysis; The output layer is used to execute control commands and display results; The working steps of the system include: S1: Collect the sensor data through the data collection layer; S2: Analyze historical production data through the standard template generation module to generate a data processing standard template, wherein the standard template includes a parameter normalization template, a feature extraction template and a sequence construction template; S3: Processing the real-time data using the standard template through the data processing layer to generate a feature Token sequence; S4: performing sequence processing and analysis through the core layer of the Transformer-based model, specifically including: S41: encoding the feature token sequence: mapping the feature vector to an embedding vector of fixed dimension, adding feature type embedding, generating position encoding and adding it to the feature embedding; S42: Processing sequence relations through multi-head attention mechanism: generating query vector (Q), key vector (K) and value vector (V), calculating attention score, normalizing attention weights through softmax, and weighted aggregation value vector; S43: Capture different feature relationships through a multi-head mechanism: compute h attention heads in parallel, each head focuses on a different feature combination, concatenates the multi-head outputs and performs linear transformation; S44: Processing feature interactions through feed-forward networks: nonlinear transformation of attention outputs, adding residual connections and layer normalization; S45: Generate feature representation required for decision making; S5: Execute the control command through the output layer and display the result, specifically including: S51: Adjust the glue coating and sanding process parameters according to the Transformer output; S52: Generate real-time control instructions and warning information; S53: updating the process parameter optimization strategy; S54: Feedback optimization results for template updating.

2. The intelligent control system for waterproof composite material production according to claim 1, characterized in that: The process parameters include the glue coating temperature, pressure, and flow rate; the equipment parameters include the motor operating status, power, and speed; the material parameters include the hot melt adhesive temperature, viscosity, and glue supply; and the quality parameters include the sand layer uniformity and adhesion.

3. The intelligent control system for waterproof composite material production according to claim 1, characterized in that: The processing steps of the standard template generation module include: performing cluster analysis on historical data to identify characteristic patterns of different product specifications and material combinations; generating data preprocessing standard templates for each type of product based on the clustering results; establishing standardized conversion templates for different process parameter combinations; generating standard templates for time series feature extraction; classifying and storing verified templates, and establishing a template index library.

4. The intelligent control system for waterproof composite material production according to claim 1, characterized in that: The steps of the data processing layer using the standard template for processing include: performing parameter standardization based on the normalization template; applying the feature extraction template to perform feature engineering; using the sequence construction template to construct a feature sequence; and organizing the processed features into a Token sequence.

5. The intelligent control system for waterproof composite material production according to claim 1, characterized in that: The system adopts different feature extraction methods for different types of parameters: temperature parameters: mean, variance, rate of change; pressure parameters: peak value, valley value, fluctuation range; speed parameters: acceleration, stability index; quality parameters: consistency index, deviation value; equipment state parameters: state transition frequency and duration.

6. The intelligent control system for waterproof composite material production according to claim 1, characterized in that: The system also includes a production line control module, which includes: Glue supply control unit, used to realize automatic adjustment of hot melt glue supply quantity, glue supply pressure and temperature control, and glue supply uniformity monitoring; Temperature control unit, used to realize independent temperature control of multiple sections of coating head, automatic temperature compensation, and heating power adjustment; Sand control unit, used to control the thickness and uniformity of the sand layer, automatically replenish the sand, and control the sand return system; Tension control unit, used to realize automatic adjustment of winding and unwinding tension, deviation correction control, and speed synchronization control; The specification switching unit is used to realize automatic identification of product specifications, automatic switching of process parameters, and automatic loading of production parameters.

7. The intelligent control system for waterproof composite material production according to claim 1, characterized in that: The core layers of the Transformer-based model include: an input embedding layer for converting tokens into vector representations; a position encoding layer for adding position information; a multi-layer attention layer for processing sequence relationships; a feedforward network layer for feature transformation and interaction; and a normalization layer for data standardization processing.

8. The intelligent control system for waterproof composite material production according to claim 1, characterized in that: The system further includes a security protection module, which includes: Temperature safety unit, used to realize over-temperature alarm, automatic temperature reduction control, and temperature emergency treatment; Pressure safety unit, used to realize abnormal pressure alarm, automatic pressure relief control, and pressure compensation adjustment; Glue supply safety unit, used to realize glue supply abnormality detection, automatic shutdown control, and glue supply safety interlock; Equipment protection unit, used to realize motor overload protection, equipment status monitoring, and abnormal operation alarm; Emergency processing unit, used to realize emergency shutdown control, safety interlock control, and fault diagnosis record; The security protection module communicates with the model core layer to achieve real-time monitoring and early warning control.

Images