An artificial intelligence-based rubber tire vulcanization process transformation system and method

Abstract

This invention discloses an artificial intelligence-based system and method for modifying the vulcanization process of rubber tires, relating to the field of intelligent tire manufacturing control technology. The system includes a data acquisition layer, an edge perception layer, an artificial intelligence algorithm model layer, and a process execution control layer. The data acquisition layer collects vulcanization process data in real time through a sensor array. The edge perception layer preprocesses the collected data to generate feature vectors. Based on the preprocessed data, the artificial intelligence algorithm model layer performs collaborative calculations using a vulcanization equivalent crosslinking degree prediction model, a temperature-pressure coupling optimization model, an intelligent early warning model, and an adaptive curve model to generate dynamic control commands. The process execution control layer receives the control commands and drives the corresponding mechanisms to perform actions. This invention effectively solves the problems of under-vulcanization and over-vulcanization, improves the consistency of rubber crosslinking degree, and significantly reduces vulcanization energy consumption.

Description

Technical Field

[0001] This invention belongs to the field of intelligent tire manufacturing control technology, specifically relating to an artificial intelligence-based system and method for modifying the vulcanization process of rubber tires. Background Technology

[0002] Tire vulcanization is a core process in rubber tire manufacturing. It is the final key step that determines the cross-linking and curing of rubber, tire life, dynamic balance, wear resistance, aging resistance, and rolling resistance. Essentially, it involves heating and pressurizing rubber molecules to undergo a cross-linking reaction, thereby forming a three-dimensional network structure with specific mechanical properties. The quality of vulcanization directly determines the tire's strength, elasticity, wear resistance, and service life, thus affecting vehicle stability and driving safety.

[0003] New energy tires require low rolling resistance, high strength, load-bearing capacity, and heat aging resistance, placing extremely high demands on the uniformity of vulcanization crosslinking. Introducing artificial intelligence vulcanization technology can precisely control the crosslinking density, effectively avoiding increased rolling resistance and accelerated aging caused by over-vulcanization. This perfectly matches the process requirements of EV quiet tires and low rolling resistance green tires, and is currently the direction for technological upgrading for leading tire companies.

[0004] However, existing tire vulcanization processes still have common problems: (1) Process parameters (such as temperature, steam pressure, vulcanization time, constant internal pressure, etc.) are mostly fixed empirical values ​​and cannot be adjusted in real time according to the differences in the tire blank, ambient temperature and humidity and rubber fluctuations; (2) There are common problems of under-sulfurization and over-sulfurization. Under-sulfurization will lead to insufficient tire strength and delamination, while over-sulfurization will lead to rubber aging, hardening, increased rolling resistance, and high scrap rate. (3) Steam energy consumption is huge and waste heat is serious. This process is the most energy-consuming process in the whole plant; (4) When changing specifications, the debugging is slow and the parameters are adjusted by relying only on the experience of workers, which will lead to large quality fluctuations between batches. (5) Vulcanized capsules are consumed quickly and are prone to problems such as poor demolding, missing glue in the pattern, and many air bubbles; (6) There is a lack of online quality prediction methods. The quality inspection of vulcanized products can only be carried out after the finished products are off the production line. This remedial measure is obviously delayed.

[0005] Chinese invention patent application CN 121091699 A discloses a predictive optimization control method and system for tire vulcanization process based on process constraints, relating to the field of intelligent tire manufacturing control technology. This includes employing a hierarchical predictive collaborative optimization control framework to achieve multivariate collaborative rolling control of vulcanization kinetics and external temperature disturbances. In the upper offline stage, a neural network integrating vulcanization kinetics and process constraints is constructed to generate a baseline temperature trajectory that combines optimality and physical feasibility, providing a feasible initial solution for the lower tracking layer. Subsequently, in the lower online control stage, an event-triggered mechanism drives a sparse Bayesian learner to update the coupled model online. The feedforward compensator then quickly parses the control commands based on this model and directly adjusts the steam valve, thereby proactively offsetting coupling interference. This technical solution can effectively suppress temperature disturbances and improve system robustness. However, this technical solution still has significant limitations in optimizing the vulcanization process.

[0006] Therefore, there is an urgent need for a technical solution that can integrate process mechanisms and data-driven methods to achieve dynamic perception and intelligent control of the tire vulcanization process, in order to overcome the above problems and improve the consistency of vulcanization quality and production efficiency. Summary of the Invention

[0007] To address the shortcomings of the existing technologies described above, this invention provides an artificial intelligence-based rubber tire vulcanization process modification system. Through dynamic sensing and algorithm model collaborative calculation, it can precisely control the crosslinking density, avoid over-vulcanization which leads to increased rolling resistance and accelerated aging, perfectly match the process requirements of EV quiet tires and low rolling resistance green tires, fully conform to the actual production conditions of tire factories, and improve the adaptability and overall energy efficiency of the vulcanization process.

[0008] Another objective of this invention is to provide an artificial intelligence-based method for controlling the vulcanization process of rubber tires based on the aforementioned modified system.

[0009] To achieve the above objectives, the present invention adopts the following technical solution: an artificial intelligence-based rubber tire vulcanization process modification system, comprising: a data acquisition layer, an edge perception layer, an artificial intelligence algorithm model layer, and a process execution control layer; The data acquisition layer includes a sensor array arranged at key locations on the vulcanizing machine for real-time acquisition of multi-dimensional operating condition data during the vulcanizing process. The edge perception layer includes an edge computing device, which is used to receive the multi-dimensional working condition data and preprocess it to generate a corresponding feature vector sequence. The artificial intelligence algorithm model layer is deployed on the edge computing device or the cloud, and includes at least a sulfurization equivalent crosslinking degree prediction model, a temperature-pressure coupling optimization model, a sulfurization anomaly intelligent early warning model, and an energy consumption adaptive curve model. The artificial intelligence algorithm model layer performs collaborative operations based on the feature vector sequence and outputs dynamic process control commands. The process execution control layer is communicatively connected to the artificial intelligence algorithm model layer, and is used to receive the dynamic process control instructions and drive the thermal valve assembly and mechanical demolding mechanism of the vulcanizing machine to perform corresponding actions.

[0010] Furthermore, the multi-dimensional operating condition data includes at least: steam temperature, outer mold temperature, capsule internal pressure, inner liner water temperature, and vulcanization time; preferably, the multi-dimensional operating condition data also includes at least one of the following: hot plate temperature difference, ambient temperature and humidity, tire carcass weight deviation, Mooney viscosity of rubber compound, workshop steam pipeline pressure, waste heat recovery flow rate, and cumulative number of capsule uses.

[0011] Furthermore, the vulcanization equivalent crosslinking degree prediction model adopts a time-series neural network that integrates physical information. Based on the rubber crosslinking reaction kinetics, it integrates the vulcanization kinetic equation and the Arrhenius equation during model training. Taking the sequence characteristics of multi-dimensional working condition data as input, it outputs the real-time equivalent crosslinking degree and crosslinking change rate. When the real-time equivalent crosslinking degree reaches the set target crosslinking degree and tends to stabilize, it triggers the process execution control layer to terminate the vulcanization action. This can be achieved by generating a vulcanization termination command through the artificial intelligence algorithm model layer and sending it to the process execution control layer to drive the mechanical demolding mechanism to perform the vulcanization termination action.

[0012] Furthermore, the temperature-pressure coupled optimization model is based on the model predictive control framework and introduces a thermal coupling transfer function matrix. It takes the measured temperature of each temperature zone, the steam pressure fluctuation, the capsule internal pressure, and the optimal temperature set trajectory and pressure set trajectory output by the energy consumption adaptive curve model as inputs. Through rolling optimization calculation, it outputs the optimal heating rate, the optimal constant temperature set value, the internal pressure control curve, and the valve opening control quantity.

[0013] Furthermore, the intelligent early warning model for sulfurization anomalies calculates the under-sulfurization risk index and the over-sulfurization risk index in real time, and assesses the comprehensive risk level based on the maximum value of the two; when the comprehensive risk level reaches the set intervention threshold, it automatically triggers and outputs online correction action commands for temperature, pressure or time.

[0014] Furthermore, the energy consumption adaptive curve model takes the instantaneous steam flow rate, cumulative steam consumption and target degree of vulcanization as inputs, adopts a multi-objective optimization framework to output the optimal temperature setting trajectory and pressure setting trajectory, and dynamically adjusts the setting process parameters of the remaining vulcanization stage according to the feedback status of the real-time equivalent crosslinking degree.

[0015] Furthermore, the system also includes: a digital twin unit for constructing a three-dimensional twin model of the vulcanizing machine and synchronously mapping real-time multi-dimensional operating condition data for visual monitoring; and a visual quality inspection feedback module deployed at the vulcanizing machine outlet, which includes image acquisition equipment and a defect detection algorithm model for identifying tire surface defects and associating the defect information with the vulcanizing process data of the corresponding batch to the artificial intelligence algorithm model layer to drive the adaptive iterative update of the model parameters.

[0016] Furthermore, the system also includes a full-specification tire vulcanization process knowledge base and a capsule life extension management module; the process knowledge base stores the association records of tire specifications, rubber batches and optimal process parameters, which are used to retrieve and automatically match the initial vulcanization parameters through feature vectors when changing tire specifications; the capsule life extension management module is used to record and calculate the capsule life loss of a single vulcanization cycle, and work with the artificial intelligence algorithm model layer to limit the pressurization rate and depressurization rate to reduce material fatigue loss.

[0017] Furthermore, the process execution control layer has a built-in fault-safe interlock mechanism: when the communication interruption time between the edge perception layer and the process execution control layer exceeds the safety threshold, or when the multi-dimensional operating condition data exceeds the safety preset range, the thermal valve assembly is triggered to reset to the safe closed state and an audible and visual alarm is activated.

[0018] An artificial intelligence-based method for controlling the vulcanization process of rubber tires, applied to the aforementioned modified system, includes the following steps: S1: Obtain the identification information of the tire blank to be vulcanized, and retrieve the corresponding rubber compound parameters and historical vulcanization benchmark data; S2: Based on the retrieved data, initialize the target degree of crosslinking, initial temperature / pressure trajectory, and basic vulcanization time window; S3: During the vulcanization process, vulcanization condition data is collected and preprocessed in real time according to a preset cycle; S4: Real-time calculation and output of the current equivalent crosslinking degree and the remaining required vulcanization equivalent through the vulcanization equivalent crosslinking degree prediction model; S5: By coordinating the temperature-pressure coupling optimization model and the energy consumption adaptive curve model, the optimal valve opening and internal pressure setpoints are output, and the heating temperature and capsule internal pressure of the vulcanizing machine are dynamically adjusted. S6: The risk level is calculated in real time through the intelligent early warning model for abnormal sulfurization. If the intervention conditions are met, online process correction is performed. S7: Determine whether the current equivalent crosslinking degree has reached the target crosslinking degree. If so, automatically terminate the heating and heat preservation process and enter the pressure relief, cooling and demolding stage. S8: During the finished product inspection stage, acquire appearance defect and uniformity inspection data, and feed the inspection data back to the artificial intelligence algorithm model layer for self-learning and parameter iteration.

[0019] Compared with the prior art, the present invention has the following beneficial effects: This invention leverages artificial intelligence to transform the intelligent vulcanization process. By combining the dynamics of rubber vulcanization with AI algorithms, it overcomes the bottlenecks of traditional fixed-parameter vulcanization, achieving precise control of equivalent vulcanization, adaptive adjustment of operating conditions, online quality early warning, intelligent energy consumption optimization, and a closed-loop data system throughout the entire process. It fundamentally addresses industry pain points such as under-vulcanization and over-vulcanization, large batch fluctuations, high energy consumption, and rapid material depletion in tires. Simultaneously, it meets the demands of intelligent manufacturing, green and low-carbon practices, and the mass production upgrade of high-end tires, making it a core upgrade module for the intelligent transformation of tire factories. This invention effectively replaces traditional fixed process curves, generating the optimal vulcanization process in real time using a model. It dynamically controls vulcanization time based on the target crosslinking degree, significantly reducing the incidence of under-vulcanization and over-vulcanization at the source, ensuring uniform and consistent rubber crosslinking. This solution can reduce vulcanization energy consumption by 20%-35%, significantly decrease steam consumption, significantly improve tire dynamic balance, uniformity, and durability, substantially reduce defect rates, and increase finished product qualification rates. It also improves the consistency of finished tire quality, extends bladder life, and supports intelligent matching of production change parameters, providing reliable technical support for the intelligent and low-carbon upgrades of tire manufacturing enterprises. Detailed Implementation

[0020] The following detailed descriptions are illustrative and intended to provide further explanation of this disclosure. Unless otherwise specified, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this disclosure pertains.

[0021] It should be noted that the terminology used herein is for the purpose of describing particular embodiments only and is not intended to limit the exemplary embodiments according to this disclosure. As used herein, the singular form is intended to include the plural form as well, unless the context clearly indicates otherwise. Furthermore, it should be understood that when the terms “comprising” and / or “including” are used in this specification, they indicate the presence of features, steps, operations, devices, components, and / or combinations thereof.

[0022] An artificial intelligence-based rubber tire vulcanization process modification system is proposed. The overall system architecture is divided into four layers: data acquisition layer, edge perception layer, artificial intelligence algorithm model layer, and process execution control layer. The layers interact with each other through industrial Ethernet, OPC•UA protocol, and shared memory mechanism.

[0023] The data acquisition layer consists of a sensor array arranged at key locations on the vulcanizing machine to collect multi-dimensional operating parameters during the vulcanization process in real time. Preferably, the sensor array includes: temperature sensors installed on the main steam pipeline and branch pipes of each temperature zone (insertion depth 1 / 3-1 / 2 of the pipe diameter) to monitor steam temperature; temperature sensors arranged in the temperature measuring holes of the molds (including the upper mold, middle mold, lower mold, and side mold) and uniformly covering each area of ​​the heating plate; pressure sensors connected to the bladder inflation pipeline for measurement to monitor the internal pressure of the bladder; temperature sensors installed on the hot water circulation pipeline to monitor the inner tank water temperature; multi-point temperature sensors for monitoring the temperature difference of the hot plates; sensors for monitoring ambient temperature and humidity; pipeline pressure sensors installed at the main pipeline and branch pipes; a counter for collecting vulcanization time; tire carcass weight deviation recorded by weighing the tire carcass before it enters the vulcanizing machine using a dynamic checkweigher; waste heat recovery flow rate measured by steam flow meters installed on the inlet and outlet pipes of the heat exchanger; and an identification unit for reading the tire carcass RFID information (tire carcass specifications, weight, rubber batch, and Mooney viscosity of the rubber). The aforementioned sensors collect data such as steam temperature, outer mold temperature, capsule internal pressure, inner liner water temperature, vulcanization time, hot plate temperature difference, ambient temperature and humidity, tire carcass weight deviation, Mooney viscosity of rubber compound, workshop steam pipeline pressure, and waste heat recovery flow rate, and connect them to the edge sensing layer.

[0024] The collected process data includes: steam temperature series T steam (t) = [T s1 (t),T s2 (t),…, T sm (t)] (m is the number of measuring points in the steam pipeline of each temperature zone), external model temperature sequence T mold (t) = [T m1 (t),T m2 (t),…, T mk (t) (k is the number of temperature measuring points on the upper, middle, and lower molds and the hot plate), capsule internal pressure P bladder (t), Inner tank water temperature T water (t), vulcanization time t, hot plate temperature difference ΔT plate (t), ambient temperature T amb (t) and ambient humidity RH amb (t), embryo weight deviation △W green (t), Mooney viscosity of rubber compound MV(t), workshop steam pipeline pressure P steam-net (t), Waste heat recovery flow rate Q recycle (t) Rubber batch identification ID batch The total number of times the capsules were used (N) bladder。

[0025] The edge sensing layer includes edge computing devices, field PLC controllers, IoT gateways, and internal preprocessing modules, enabling real-time data upload and low-latency control. It can operate locally even when the cloud is down. The edge computing devices preferably use industrial-grade embedded industrial PCs, communicating with the PLC and I / O modules of the data acquisition layer via Modbus•TCP, Profinet, or OPC•UA protocols to read data at preset intervals (e.g., 1s-5s, preferably 2s).

[0026] Edge computing devices connect upwards to the AI ​​algorithm model layer via Ethernet and downwards to the process execution control layer via a control bus. The data acquisition layer sequentially transmits the acquired data to the field PLC controller, which then directly transmits it to the edge computing devices. The IoT gateway enables remote data uploading and device network management. The edge computing devices upload the received data to the AI ​​algorithm model layer. Control commands generated by the AI ​​algorithm model layer are then sent down to the field PLC controller via the edge computing devices. Finally, the field PLC controller drives the process execution control layer to execute the commands.

[0027] The preprocessing module preprocesses the process data acquired by the data acquisition layer as follows: First, outliers are removed using a sliding window mid-range filter with 5-10 sampling points. Then, missing values ​​are filled using linear interpolation. Subsequently, Z-score standardization is used to normalize each parameter channel to a distribution with a mean of 0 and a standard deviation of 1 to eliminate dimensional differences. Finally, the batch ID of the rubber compound is assigned. batch Categorical variables are one-hot encoded to ultimately generate feature vector X. input (t). The eigenvector X input (t) includes: steam temperature sequence, outer mold temperature sequence, bladder internal pressure, inner liner water temperature, vulcanization time, hot plate temperature difference, ambient temperature and humidity, tire carcass weight deviation, Mooney viscosity of rubber compound, pipeline pressure, waste heat recovery flow rate, rubber batch identifier, and cumulative number of bladder uses. The edge perception layer uses shared memory (when the AI ​​model is deployed on the same device) or a message queue (when across devices) to process the processed feature vector X. input (t) is pushed to the artificial intelligence algorithm model layer in real time.

[0028] In other words, the preprocessing module uses sliding window mid-range filtering to remove outliers, linear interpolation to fill in missing values, and Z-score standardization to generate feature vector X. input (t). The processed feature vectors are pushed to the algorithm model layer in real time.

[0029] The artificial intelligence algorithm model layer can be deployed in an independent container within the same edge computing device. It includes four core models that are executed in parallel: a sulfurization equivalent crosslinking degree prediction model, a temperature-pressure coupling optimization model, a sulfurization anomaly intelligent early warning model, and an energy consumption adaptive curve model. Low-latency data exchange is achieved through shared memory.

[0030] The AI ​​algorithm model layer establishes a bidirectional data connection with the edge perception layer via the Ethernet TCP / IP protocol: on the one hand, the AI ​​algorithm model layer acquires pre-processed data from the edge perception layer in real time as input for model calculation; on the other hand, after completing the calculation, the AI ​​algorithm model layer sends the generated control commands to the edge perception layer, which then forwards them to the actuators. Furthermore, the AI ​​algorithm model layer is also connected to the vulcanization digital twin visualization system, pushing real-time process parameters and status data to the digital twin model for dynamic rendering of the 3D monitoring screen; simultaneously, the AI ​​algorithm model layer interfaces with the MES system to obtain production plans, order specifications, and rubber batch information to support adaptive production changeover and global scheduling optimization.

[0031] The artificial intelligence algorithm model layer includes the following four core models: Vulcanization Equivalent Crosslinking Degree Prediction Model: This model employs a temporal neural network (such as LSTM GRU, Temporal Convolutional Network (TCN), or Transformer) that integrates physical information. Based on the kinetics of rubber crosslinking reactions, it combines the vulcanization kinetic equation with the Arrhenius equation to calculate the equivalent vulcanization degree in real time. This model is used to estimate the crosslinking degree of tires during the vulcanization process. The input is a multi-dimensional operating condition data sequence, and the output is the real-time equivalent crosslinking degree α(t), the rate of change dα / dt, and the remaining required vulcanization equivalent Δα. When α(t) reaches the target crosslinking degree α... target When the process is continuous and stable, the model issues a termination signal for sulfurization.

[0032] The model's inputs include real-time external mold temperature sequence, bladder internal pressure sequence, steam temperature sequence, cooling water temperature, vulcanization time, Mooney viscosity of the rubber compound, tire thickness / weight, ambient temperature and humidity, and rubber batch ID. Outputs include real-time equivalent crosslinking degree α(t), rate of change dα / dt, remaining required vulcanization equivalent Δα, and the predicted remaining time required to reach the target crosslinking degree. This model serves as the core basis for determining vulcanization termination, replacing traditional fixed-time control with the calculated equivalent vulcanization degree to achieve precise vulcanization based on the crosslinking state. It only performs quantitative calculations and does not output any risk level or warning information. Its outputs α(t), dα / dt, and Δα are unidirectionally transmitted via shared memory to the vulcanization anomaly intelligent early warning model and process control module.

[0033] The vulcanization equivalent crosslinking degree prediction model collects historical vulcanization data of tires of different specifications, with a balanced proportion of normal tires, under-vulcanized tires, and over-vulcanized tires, using the pre-processed X... input Using (t) as input and the equivalent crosslinking degree measured in the laboratory (by a vulcanizer) as label, the model was trained for 200 epochs using the Adam optimizer (learning rate 0.001, batch size=64) with an early stopping patience value of 20. After training, the model achieved an RMSE ≤ 0.025 on the validation set, and R0 was 0.025. 2 ≥0.96.

[0034] The model has a sampling period Δt = 2 seconds. Real-time process data is input into the model via a sliding time window (60 sampling points in length, 1 sampling point in step). The model outputs the instantaneous equivalent crosslinking degree α(t) and its rate of change dα / dt at the current moment. The target value of the equivalent crosslinking degree α is... target The value is preset according to the tire specifications and rubber compound formulation, and ranges from 0.85 to 1.0.

[0035] The temperature-pressure coupling optimization model, based on the Model Predictive Control (MPC) framework, introduces an online-identified time-varying thermal coupling transfer function matrix. It outputs the optimal heating rate, isothermal setpoint, internal pressure control curve, and optimal steam valve opening, addressing the issues of uneven temperature and thermal interference across multiple temperature zones during vulcanization. This model characterizes the mutual heat transfer effects and pressure-temperature coupling effects between different temperature zones (tire crown, sidewall, and bead) of the vulcanizing machine. The model employs an online-identified time-varying coupling matrix; the specific identification method can utilize Laguerre polynomial expansion combined with sparse Bayesian learning (but is not limited to this) to adapt to the dynamic characteristic variations under different tire specifications and different rubber batches.

[0036] The inputs to the temperature-pressure coupled optimization model include: measured temperatures in each temperature zone, steam pressure fluctuations, hot plate temperature difference, bladder internal pressure, pipeline pressure, and coupled heat transfer coefficients obtained through online identification; the output is a set of optimal control parameters, specifically including: optimal heating rate, optimal isothermal setpoint, optimal internal pressure control curve, optimal steam valve opening, and thermal coupling compensation. These output parameters collectively constitute the temperature setpoint trajectory r. T (t) and pressure set trajectory r P = (t) is a complete parameterized representation, serving as a reference input for subsequent execution control layers. This model aims to address the issues of uneven temperature across multiple temperature zones, pressure fluctuations, and thermal interference during vulcanization, achieving high-precision tracking of the set trajectory.

[0037] Intelligent early warning model for sulfurization anomalies: Employing a hybrid architecture of classification and anomaly detection, it calculates the undersulfurization risk index R in real time. under( t) and the supersulfur risk index R over (t), and classify the comprehensive risk level L.v When a moderate to high risk is detected (e.g., a local temperature difference exceeding the limit for 30 seconds), the system outputs corrective suggestions or automatically triggers valve adjustment commands (such as automatically raising the temperature setpoint or opening the cooling water valve) to dynamically intervene before scrap is generated.

[0038] The intelligent early warning model for vulcanization anomalies is implemented using LightGBM, XGBoost, multilayer perceptron (MLP), or a temporal anomaly detection model. This model does not directly calculate the equivalent crosslinking degree. Instead, it uses the real-time equivalent crosslinking degree α(t), the remaining required vulcanization equivalent Δα, and the vulcanization rate trend output by the vulcanization equivalent crosslinking degree prediction model as core inputs. It also incorporates temperature deviation integrals, pressure fluctuation statistics, temperature uniformity in each region (tire crown / sidewall / bead) (used to detect localized insufficient vulcanization), and features from historical under-vulcanized tires, normally vulcanized tires, over-vulcanized tires, and tires failing durability tests extracted from the rubber vulcanization quality label database. If the measured temperature in any region deviates from the regional average temperature by more than ±5℃ for more than 30 seconds, it is determined that there is a risk of localized insufficient vulcanization in that region.

[0039] The model calculates the undersulfur risk index R in real time. under (t) and the supersulfur risk index R over (t), and based on the undersulfur risk index R under (t) and the supersulfur risk index R over (t) Calculate the comprehensive risk level, where the comprehensive risk level L is... v ∈{0,1,2,3} (0 = no risk, 1 = low risk, 2 = medium risk, 3 = high risk): L v =0 (no risk): max(Runder, Rover) < 0.2; L v =1 (mild risk): 0.2 ≤ max(Runder, Rover) < 0.4 L v =2 (moderate risk): 0.4 ≤ max(Runder, Rover) < 0.7; L v =3 (Severe Risk): max(Runder, Rover) ≥ 0.7.

[0040] Based on the above risk levels, the warning level, predicted defect location (tire crown / sidewall / bead), and suggested corrective actions (adjusting upper and lower temperature limits, vulcanization time) are output. v When ≥2, output "Suggest adjusting process parameters", L v When the value is 3, the system outputs "emergency intervention" and triggers automatic protection, forcing the system to enter the adaptive process value adjustment stage.

[0041] Once an anomaly is detected (indicating under-sulfurization, over-sulfurization risk, or localized insufficient sulfidation), the system automatically adjusts the steam input, internal pressure, and holding time, correcting deviations online to bring the sulfidation process back on track without interrupting production and preventing the generation of defective products. When an alarm is triggered, the system automatically executes the following feedback actions: The system pushes early warning information to a large visual screen in real time, alerting operators via pop-ups and audible / visual alarms. A snapshot of the process data at the time of the warning (30 seconds before and 10 seconds after) is automatically saved to the anomaly database. For under-sulfur warnings, the temperature-pressure coupling optimization model automatically raises the temperature setpoint by 2-5°C, extends the vulcanization time by 5%-10% (within safety limits), or increases the steam valve opening. For over-sulfur warnings, the system automatically opens the cooling water valve in advance, reduces the steam valve opening, or proactively shortens the remaining pressure holding time to slow the crosslinking rate. Furthermore, the early warning model receives defect feedback from the AI ​​visual quality inspection module and employs an adaptive threshold adjustment strategy: if three consecutive tires in the same batch exhibit defects related to vulcanization, such as bubbles or missing rubber, the model automatically lowers the warning trigger threshold by 0.02 (not lower than 0.90) to improve sensitivity. If the subsequent 20 consecutive tires are defect-free, the threshold increases by 0.01 for every 10 tires until it returns to the original threshold (typically 0.96). Through the above mechanism, the model can provide early warnings online and intervene before waste is generated.

[0042] Energy consumption adaptive curve model: Employing a multi-objective optimization framework, the system reconstructs a customized vulcanization curve in real time based on steam flow rate, target vulcanization degree, and current operating conditions. The system dynamically adjusts the temperature / pressure setting trajectory based on actual crosslinking degree feedback and extends the recovery time as needed during the cooling phase, achieving intelligent standby hibernation and low-temperature high-efficiency vulcanization strategies, significantly reducing the total energy consumption of single-tire vulcanization.

[0043] The energy consumption adaptive curve model is used to minimize the total energy consumption of a single tire vulcanization cycle while ensuring that the equivalent crosslinking degree meets the standard. This model employs an offline pre-trained energy-optimized strategy network, combined with online dynamic adjustment, to achieve intelligent optimization of the vulcanization curve through artificial intelligence. Traditional three-stage vulcanization (heating → constant temperature → pressure holding) uses fixed curves, while the AI-modified model achieves a flexible dynamic vulcanization curve. This means that based on the individual differences of each tire blank (such as blank weight, rubber Mooney thickness, and component thickness deviations) and fluctuations in on-site operating conditions (such as steam pressure fluctuations, hot plate heat dissipation, and ambient temperature differences), the AI ​​reconstructs a dedicated vulcanization curve in real time, dynamically adjusting the heating rate, peak temperature, pressure holding time, and internal pressure decay curve.

[0044] Through the above technical solutions, the uniformity of vulcanization of tires of the same specification is greatly improved, and the cross-linking degree of multiple parts such as the tire shoulder, sidewall, and bead tends to be consistent, effectively solving the problem of uneven vulcanization inside and outside in traditional processes.

[0045] The inputs to the energy consumption adaptive curve model include: instantaneous steam flow rate, cumulative steam consumption, waste heat recovery flow rate, peak and off-peak steam times in the workshop, unit load, target vulcanization degree, current temperature and pressure, and operating condition characteristics (rubber activation energy, tire thickness, target crosslinking degree, pipeline pressure, ambient temperature, Mooney viscosity, etc.); the output is the optimal temperature setting trajectory r. T (t) and pressure set trajectory r P (t) serves as the reference input for the temperature-pressure coupled optimization model. The output includes the minimum energy consumption vulcanization curve, optimal steam supply strategy, intelligent waste heat allocation scheme, standby / dormancy temperature setting, and energy saving rate prediction. During vulcanization, the model dynamically adjusts the set process correction values ​​for the remaining stages based on feedback from the actual degree of crosslinking, and links with the waste heat recovery control module to extend the recovery time as needed during the cooling stage. This achieves the lowest energy consumption in the vulcanization process while ensuring vulcanization quality.

[0046] In the actual vulcanization process, the energy consumption adaptive curve model receives real-time feedback on the equivalent degree of crosslinking and dynamically adjusts the temperature / pressure setting curves for the remaining stages. The adjustment strategy is as follows: If the actual degree of crosslinking α(t) is higher than the predicted trajectory, and the predicted final degree of crosslinking will exceed the upper limit of the target value (0.98), then the temperature setpoint of the subsequent stage will be appropriately reduced (by 2-5℃) or the heat preservation time will be shortened; if the actual degree of crosslinking is lower than the predicted trajectory, and the predicted final degree of crosslinking will be lower than the lower limit of the target value (0.93), then the temperature setpoint of the subsequent stage will be appropriately increased (by 2-5℃) or the heat preservation time will be extended; at the same time, the cumulative energy consumption value will be monitored. If the current energy consumption is close to 80% of the historical average and the vulcanization progress is less than 60%, then the system will automatically switch to "energy-saving mode" to reduce the rate of temperature rise.

[0047] Furthermore, the model deeply integrates the following strategies: Based on pipeline steam pressure fluctuations, the model optimizes steam injection volume and employs pressure feedforward control. When the pipeline pressure exceeds a set value (e.g., >0.8MPa), it automatically reduces the steam valve opening, utilizing pipeline residual pressure to compensate for steam supply and prevent excessive steam supply. When the pressure is low, it increases the valve opening in advance and extends the heating time to prevent insufficient steam supply. The model is linked with the waste heat recovery control module, using artificial intelligence to allocate waste heat recovery based on the peak and off-peak heating periods of the vulcanizing unit: during peak periods of concentrated vulcanization heating, recovered waste heat is prioritized for preheating new tire blanks or auxiliary insulation; during off-peak periods, excess waste heat is stored or used for workshop heating. Specifically, this is achieved through real-time data acquisition of Q... recycle (t) and the load of the machine group, during the cooling stage after vulcanization, according to the waste heat recovery flow rate Q recycle(t) Automatically adjust the cooling rate. When the recovery flow rate is large (exceeding 120% of the demand), appropriately extend the cooling time to recover more waste heat, achieving on-demand recovery and further improving the energy utilization efficiency of the entire line. The low-temperature high-efficiency vulcanization strategy is reflected in the model reducing ineffective high-temperature losses while meeting the crosslinking requirements: while meeting the equivalent crosslinking degree target (α) target Under the premise of ≥0.93, by optimizing the heating rate and peak temperature, excessively high temperatures are avoided for extended periods (e.g., the upper limit of the constant temperature stage is reduced from 185℃ to 178℃), while the low-temperature crosslinking time is appropriately extended, resulting in a typical peak temperature reduction of 5-8℃. Intelligent sleep mode during idle standby is implemented by the model in real-time monitoring of the vulcanizing machine group's occupancy status: when a vulcanizing machine is idle for more than 10 minutes (no tire blanks enter), the model automatically issues a standby command, reducing the hot plate temperature to 120±5℃ (normal production temperature is 175-185℃), closing the steam valve to 5% opening (maintaining only positive pressure), and reducing the circulating fan speed to 30%. When a new tire blank is detected entering, the model rapidly heats up to the operating temperature within 30 seconds based on the pre-learned preheating curve to reduce hot plate idle consumption. Through the above comprehensive energy-saving strategies, in actual production line verification, steam consumption per unit of the vulcanizing process decreased by more than 25%, electricity costs decreased simultaneously, and carbon emissions were correspondingly reduced, becoming a core highlight of the dual-carbon technology upgrade. Through the above design, the model achieves the lowest energy consumption in the vulcanization process while ensuring vulcanization quality, and fully supports all the requirements of AI-driven energy-saving vulcanization.

[0048] The artificial intelligence algorithm model layer simultaneously receives defect information from the visual quality inspection feedback module, dynamically adjusting the early warning threshold and knowledge base update strategy. The outputs of the four models (α(t), u...) total (k), warning level, and set trajectory) are transmitted back to the edge perception layer via shared memory, and then forwarded by the edge perception layer to the process execution control layer and visualization module. The above four algorithm models run in parallel in the edge computing box for data interaction. The data flow is as follows: the full-dimensional process data collected by the high-precision sensor array is preprocessed to obtain X. input (t), which is fed into four models simultaneously as input.

[0049] During the prediction and optimization phase, the equivalent crosslinking degree prediction model for sulfurization outputs the equivalent crosslinking degree α(t) and its rate of change dα / dt at the current moment in real time; the temperature-pressure coupled optimization model receives α(t) relative to the target value α. target The deviation is calculated using rolling optimization to determine the final control quantity u. total (k); The energy consumption adaptive curve model outputs the optimal set trajectory r based on the current working conditions (including rubber activation energy, tire thickness, pipeline pressure, ambient temperature, etc.). T (t) and r P(t), which serves as the reference input for the coupled optimization model; the intelligent early warning model for sulfurization anomalies simultaneously receives α(t) and dα / dt, and outputs the early warning level and corresponding process adjustment suggestions.

[0050] In the feedback loop, the suggested actions output by the early warning model automatically correct the constraint parameters of the coupled optimization model (such as upper and lower temperature limits, vulcanization time, etc.), while the defect information fed back by the visual quality inspection module is used to dynamically adjust the early warning threshold and update the process knowledge base. Finally, the control quantity u calculated by the coupled optimization model... total (k) The control is sent to the steam valve actuator to complete the closed-loop control. The entire process is executed in a cycle of Δt=2 s until the current tire vulcanization process ends.

[0051] The process execution control layer receives instructions from the algorithm model layer and directly drives the thermal valve components (such as steam valves, drain valves, and internal pressure regulating valves), heating modules, and demolding mechanisms of the vulcanizing machine. This layer incorporates a fault-safe interlock mechanism: when communication is interrupted for more than a set time (e.g., 5 seconds) or sensor data is abnormal, the valves automatically reset to a safe closed state, triggering alarms and equipment shutdown protection. The process execution control layer acts as a bridge between the artificial intelligence algorithm model layer and the physical equipment of the vulcanizing machine. It consists of a field control cabinet (including a PLC or remote I / O module), signal conversion units, and various actuators. Instructions issued by the artificial intelligence algorithm model layer directly control the steam valves, drain valves, internal pressure regulating valves, heating modules, and demolding mechanisms of the vulcanizing machine, achieving unattended automatic vulcanization. The steam valve is installed on the main steam inlet pipeline of the vulcanizing machine. It is suitable for high-temperature and high-pressure steam environments and is equipped with an electric actuator to receive a 4-20mA current signal to control the steam flow. The drain valve is a two-position two-way or two-position three-way solenoid valve (normally closed type), installed at the bottom of the steam chamber of the vulcanizing machine and in the drain pipeline inside the capsule. The valve body is connected by threads or flanges and is suitable for the discharge of condensate and high-temperature liquids. The artificial intelligence algorithm model layer determines the opening and closing timing of the drain valve based on the vulcanization stage (heating, constant temperature, cooling) and pressure status. The control command is sent to the PLC via Ethernet. The PLC connects or disconnects the power supply to the solenoid valve coil through the digital output module. The valve's on / off status is fed back to the PLC's digital input module through limit switches or proximity switches for status confirmation and fault diagnosis. The internal pressure regulating valve is a pneumatic diaphragm regulating valve, installed in the capsule pressurization pipeline (nitrogen or steam medium), and equipped with an electrical converter to receive a 3-15psi pneumatic signal to control the capsule's internal pressure. Depending on the heating method of the vulcanizing machine's hot plate, the heating module can use electric heating tubes or steam-assisted heating coils, embedded inside the vulcanizing machine's hot plate, with power adjustment via actuators. The artificial intelligence algorithm model layer calculates the required heating power percentage based on temperature deviation and the compensation amount of the temperature-pressure coupling optimization model. The PLC sends control signals to the actuators via the PWM output module, and the actuators adjust the average power of the heating tubes using a rapid on / off method. The demolding mechanism is typically driven by a hydraulic cylinder, installed on the vulcanizing machine base or upper crossbeam, used to eject the tire or lift the upper mold after vulcanization. After the artificial intelligence algorithm model layer determines that vulcanization is complete and safety conditions are met (internal pressure ≤ 0.1MPa, bladder temperature ≤ 80℃, mold opened to a safe position), it triggers the extension coil of the demolding solenoid valve via the PLC digital output module. The cylinder pushes the ejector plate to remove the tire. Once in position, the limit switch sends a feedback signal to the PLC, shutting off the solenoid valve power. During demolding, the PLC continuously monitors the oil pressure and stroke; if jamming or timeout occurs, it automatically stops and alarms.

[0052] To ensure system safety, the execution control layer has a built-in fault-safe interlock mechanism: when the communication between the edge computing box and the execution control layer is interrupted for more than 5 seconds, or when the sensor data exceeds a reasonable range (such as temperature > 200℃ or pressure > 2.5MPa), all valves will automatically reset to the safe closed state (steam valves will be fully closed, and cooling water valves will be opened as needed), and an audible and visual alarm and a shutdown request for the MES system will be triggered to automatically perform safety protection.

[0053] The control commands are issued using a dual-redundant method: the primary path uses the Ethernet TCP / IP protocol, and the backup path uses a 4-20mA hard-wired analog signal for backup. If the primary path fails to respond to three consecutive commands, it automatically switches to the backup path and notifies maintenance personnel.

[0054] This system also provides a visual quality inspection feedback module. The visual quality inspection module uses industrial cameras and defect detection models (such as the YOLO series) to identify defects such as bubbles and missing glue. It feeds back the defect features to the perception layer via the MQTT protocol, driving the model to dynamically adjust the warning threshold and control weight matrix.

[0055] The visual quality inspection feedback module is deployed at the vulcanizing machine outlet and includes an industrial camera, an image preprocessing unit, and a YOLOv8-based defect detection model. This module interfaces with finished product AI appearance inspection and dynamic balance inspection data, receiving data including bubbles, missing rubber, tread defects, deformation, and poor uniformity, and feeding this data back to the AI ​​model. After each tire vulcanization is completed, the camera captures an image of the tire surface. After preprocessing (grayscale conversion, histogram equalization, and size normalization), the image is fed into the defect detection model, outputting the defect type, location, area, and confidence score (0-1). If the confidence score of any defect exceeds the threshold (0.6), the module pushes the defect information (including tire batch ID, defect type, and location) to the edge perception layer via the MQTT protocol. The edge perception layer then processes the tire's process data throughout the vulcanization cycle. input (t) is associated with defect information to extract feature snapshots (30s before to 10s after) before and after the defect occurs.

[0056] The associated data is fed into the artificial intelligence algorithm model layer, driving the model's continuous self-learning iteration. The temperature-pressure coupling optimization model adjusts the weight matrix of the corresponding temperature zone according to the defect type (for example, if the defect is "local lack of rubber" and is located on the tire shoulder, the tracking weight of the tire shoulder temperature measurement point in MPC is increased by 20%). If the defect is of the "deformation" or "poor uniformity" type, the model adjusts the constraint weights of the pressure curve and cooling sequence. The under-curing / over-curing warning model dynamically adjusts the warning threshold according to the defect statistics, using the adaptive threshold adjustment strategy described above. The energy consumption adaptive curve model marks the defect batch as a negative sample when vulcanizing tires of the same specification for the next time, reducing the recommendation probability of similar trajectories during policy network inference.

[0057] Through the aforementioned mechanism, this module forms a complete closed-loop self-optimizing system encompassing production, inspection, feedback, AI optimization, and reprocessing. Appearance inspection data and dynamic balance data from each batch of vulcanized tires are automatically fed back to the AI ​​model, driving continuous iteration of model parameters. As the number of batches increases, the warning threshold, weight matrix, and recommended trajectory continuously approach their optimal values, making the system more accurate. The entire optimization process is completed before the next tire begins vulcanization (typically taking ≤30 seconds), thus achieving iterative optimization from "passive alarm" to "active elimination." A comparison of key indicators before and after the modification shows that the overall tire finished product qualification rate has increased to 99.5%-99.8%.

[0058] This system also provides a comprehensive tire vulcanization process knowledge base and a capsule life extension management module: The knowledge base stores the optimal process curves and quality scores for each tire specification / batch. During production changeovers, it automatically matches the optimal parameters by querying vectors, shortening the debugging time. The life extension management module records the temperature, pressure peak values, and change rates for each cycle. Combined with the algorithm layer, it limits the pressure increase rate (preferably not exceeding 0.05 MPa / s) and pressure decrease rate (preferably not exceeding 0.1 MPa / s) to actively mitigate pressure shocks and delay capsule aging.

[0059] The AI-based rubber tire vulcanization process modification system also includes a built-in knowledge base for full-specification tire vulcanization processes and a capsule life extension management module. The full-specification tire vulcanization process knowledge base is built using a graph database (Neo4j), stored in an SSD on an edge computing node, and periodically synchronized to the cloud. The knowledge base covers various tire specifications, including semi-steel, all-steel, light truck, snow tires, and low rolling resistance new energy tires. After each batch of vulcanization, if the finished tire's quality score is ≥90 points, the process parameters (discrete representation of temperature curve, discretized representation of pressure curve, vulcanization time, energy consumption index), operating condition characteristics (rubber batch, specification, Mooney viscosity, ambient temperature and humidity), and quality score of this batch are recorded as a new record. After feature extraction, this record is written into the knowledge base as a key-value pair using tire specification, rubber compound number, and optimal process curve hash value. Simultaneously, the weights of the "parameter-quality association" edges in the knowledge graph are updated.

[0060] When a tire specification change is required, the operator only needs to select the target tire specification in the system (e.g., switching from "semi-steel" to "low rolling resistance new energy tire"). The energy consumption adaptive curve model in the artificial intelligence algorithm model layer encodes the current operating condition characteristics into a query vector. Using an approximate nearest neighbor algorithm, it retrieves the top K historical records (K=3) with the highest similarity in the knowledge base and automatically matches the optimal vulcanization parameter package, including temperature setting trajectory, pressure setting trajectory, vulcanization time, and energy consumption optimization strategy. If the highest similarity is below the threshold (0.7), a knowledge base expansion request is triggered, which is then manually entered after confirmation by an engineer. Every month or after accumulating 200 new records, a full re-indexing of the knowledge base is triggered, updating the association weight matrix and eliminating old records with a quality score below 85, maintaining the timeliness and effectiveness of the knowledge base. The entire process requires no manual vulcanization trials or reliance on experienced technicians for parameter adjustments; intelligent matching of production change parameters can be completed with a single click. Compared to the traditional manual production change method (which requires trial vulcanization and multiple parameter adjustments), the production change time is reduced by more than 70%.

[0061] The main causes of capsule aging, breakage, and bulging can be categorized into three types: excessively high temperatures accelerate the thermo-oxidative aging of rubber materials; excessive internal pressure impact causes instantaneous stress concentration in the capsule; and excessively rapid pressurization / depressurization rates lead to fatigue damage in the capsule materials. To address this, this module proposes an AI-driven joint optimization strategy for capsule lifespan. By monitoring key operating parameters in real time and utilizing AI algorithms to proactively optimize the pressurization curve, depressurization rate, and cooling sequence, this strategy mitigates pressure impacts, reduces thermal aging, and extends capsule lifespan.

[0062] The system records the following impact characteristics in real time during each vulcanization cycle: peak temperature T of the capsule inner wall. peak (Unit: °C), Peak intracapsule pressure P peak (Unit: MPa), Average rate v during the boost phase pressurize (Unit: MPa / s), Average rate v during the pressure reduction phase depressurize (Unit: MPa / s), and the temperature change rate dT / dt during the cooling phase. cool (Unit: ℃ / min). A single-cycle lifespan loss function for the capsule is constructed, which is a weighted combination of various impact characteristics (including exponential and quadratic terms). Empirical weighting coefficients are calibrated based on field data. When the cumulative loss reaches a set threshold, the system predicts that the capsule has reached the end of its lifespan and issues a replacement reminder in advance.

[0063] In terms of proactive optimization control, the artificial intelligence algorithm model layer (especially the energy consumption adaptive curve model and the temperature-pressure coupled optimization model) jointly executes the following optimization actions to slow down capsule aging at its source: (1) Optimize the pressure rise curve: Change the traditional step pressure rise to a multi-segment gradual pressure rise curve. Artificial intelligence dynamically sets the segmented pressure set points of the pressure rise stage according to the current capsule temperature, rubber viscosity and tire blank weight deviation. The pressure rise rate is limited to no more than 0.05 MPa / s, so that the capsule gradually expands and fits the tire blank, avoiding instantaneous large pressure impact.

[0064] (2) Optimize the depressurization rate: During the depressurization stage after vulcanization, artificial intelligence calculates the maximum allowable depressurization rate (related to the material fatigue strength) in real time based on the temperature of the inner wall of the capsule. An exponential decay depressurization curve is adopted (initially rapidly reduced to 60% of the working pressure, and later slowly reduced to atmospheric pressure) to avoid rapid depressurization causing capsule bulging or delamination. The depressurization rate is limited to ≤0.1 MPa / s.

[0065] (3) Optimize the cooling sequence: During the cooling stage after vulcanization, artificial intelligence dynamically adjusts the opening of the cooling water valve and the cooling time to control the temperature change rate of the capsule within the range of 3-8℃ / min; when the capsule temperature drops below 80℃, the forced cooling is automatically stopped and natural cooling is switched to avoid thermal stress concentration; if the local temperature difference of the capsule is detected to exceed 30℃, the cooling is paused and the heat exchange cycle is started.

[0066] After the above strategy was deployed on an actual tire vulcanization production line, a six-month follow-up statistical analysis was conducted on tires of the same specifications and with the same capsule material. The on-site results showed that the capsule lifespan was increased by 15%-25%, and the cost of consumables was reduced.

[0067] Through the aforementioned AI-driven joint optimization strategy, this device has achieved a paradigm shift from "passively withstanding shocks" to "actively and steadily regulating" capsule life extension, significantly reducing production costs and equipment maintenance frequency.

[0068] This system also provides a digital twin visualization unit. The digital twin unit builds a vulcanizing machine model in a 3D engine, mapping real-time temperature / pressure field cloud maps and equivalent vulcanizing progress, supporting global remote monitoring and intelligent scheduling optimization of production status.

[0069] The vulcanization digital twin visualization system is deployed on the central control screen in the central monitoring room and multiple workstation terminals in the workshop. It is used to realize global real-time monitoring, remote operation and intelligent decision-making of the vulcanization machine group, and realize unattended operation of the workshop in the dark.

[0070] The hardware of the vulcanization digital twin visualization system includes: an industrial central control screen, workstation touch terminals, vulcanizing machine body, temperature sensors, pressure transmitters, steam flow meters, edge computing gateways, and inspection robots; the software includes a 3D twin rendering engine, data parsing services, cloud map interpolation algorithms, equivalent vulcanization calculation modules, fault early warning and judgment modules, energy consumption statistics modules, and batch traceability databases.

[0071] The aforementioned hardware and software are connected and integrated through an "edge-cloud" collaborative architecture. Temperature sensors, pressure transmitters, and steam flow meters are installed at key measuring points on each vulcanizing machine (such as hot plates, inside the capsule, and steam pipelines), and are connected to the edge computing gateway via analog signals (4-20mA or 0-10V) or RS485 / Modbus bus. The edge computing gateway is also interconnected with the vulcanizing machine controller (PLC) via industrial Ethernet to obtain equipment status, fault codes, and process settings. The inspection robot communicates with the edge computing gateway via WiFi / 5G to upload on-site video, thermal imaging, and equipment inspection data. The industrial central control screen and workstation touch terminals are connected to the edge computing gateway and digital twin server via the workshop LAN (TCP / IP) to achieve human-machine interaction. The software modules can be centrally deployed on a high-performance edge server (or local server) or run in a containerized manner in the extension unit of the edge computing gateway. The data parsing service subscribes to raw process data (sampling period 2s) from the gateway's real-time database, and distributes it to the cloud map interpolation algorithm and the equivalent vulcanization calculation module after format parsing. The cloud map interpolation algorithm uses inverse distance weighting or Kriging to expand the discrete measurement point temperature / pressure values ​​into a full-field cloud map. The equivalent vulcanization calculation module outputs the crosslinking progress in real time based on the Arrhenius equation. The fault early warning judgment module receives the risk level from the early warning model, the energy consumption statistics module accumulates steam and electricity consumption, and the batch traceability database associates and stores the serial number of each tire with all process data and quality inspection results.

[0072] During actual operation, the data parsing service acquires the measured temperature, pressure, and flow data of each vulcanizing machine every 2 seconds, triggering the cloud map interpolation algorithm to generate temperature and pressure field matrices, which are then dynamically updated in the Unity3D model by the rendering engine. Simultaneously, the equivalent vulcanization calculation module pushes the current crosslinking degree to the progress bar control. All results (including equipment operating status, fault warnings, energy consumption statistics, and quality reports) are bound to the UI interface of the central control screen. Operators can click on any vulcanizing machine model to view detailed curves and alarm records. When a new batch begins, the batch traceability database automatically generates a production record and continuously writes data throughout the vulcanization cycle, linking it to the visual quality inspection results upon completion. Inspection robots periodically photograph equipment instruments and capsule appearance according to a set route, uploading anomalies (such as steam leakage or capsule bulging) through image recognition algorithms and overlaying them onto the warning list of the digital twin interface. Operators can directly drag the viewpoint, zoom the model, or issue production change parameters with a single click on the workstation touch terminal. The interface calls the knowledge base to retrieve the optimal curve, which is then forwarded to the vulcanizing machine PLC via the edge computing gateway. The entire system ensures low-latency two-way communication through MQTT and OPC UA protocols, thereby enabling remote monitoring, unmanned inspection, and intelligent decision-making in a dark workshop.

[0073] The vulcanization digital twin visualization system constructs a digital twin interface for the vulcanizing machine cluster. A 1:1 vulcanizing machine model is built in Unity3D. Real-time data from the core layer of the artificial intelligence algorithm is subscribed to via MQTT, dynamically updating the real-time temperature and pressure field cloud maps (color mapping) for each vulcanizing machine, and displaying a real-time equivalent vulcanization progress bar in the 3D scene. Furthermore, the interface integrates equipment operating status (normal, standby, fault), fault warning information (including warning level and suggested actions), energy consumption statistics (instantaneous steam flow, cumulative steam consumption, electricity consumption), quality reports (pass rate, defect distribution), and batch traceability query functions. The central control screen supports global monitoring, allowing remote viewing of detailed process parameters and historical curves for any vulcanizing machine.

[0074] During production changeover, operators can select the new tire specification (e.g., "235 / 45R18 EV low rolling resistance") on the digital twin interface. The system then calls upon an energy consumption adaptive curve model for offline simulation, displaying the estimated steam consumption and vulcanization time. The parameter package is then sent out with a single click. Compared to traditional manual production changeover methods, this reduces changeover time by more than 70%. All process and quality inspection data for each tire are linked to an encrypted database via its serial number, supporting traceability at any time.

[0075] Through the aforementioned digital twin and global monitoring system, the central control room can achieve unattended operation of the entire plant's vulcanizing machine group: the inspection robot automatically completes on-site equipment inspections, and the system automatically pushes alarm information and attempts to remotely reset when abnormalities occur; operators can complete production changeovers, parameter adjustments, and fault handling without entering the workshop, truly realizing intelligent production management in the dark workshop.

[0076] This invention discloses an AI-based rubber tire vulcanization process modification system that utilizes existing hydraulic vulcanizing machines. It fully considers the actual conditions of existing tire vulcanization production lines, supporting gradual, low-impact modifications without interrupting production. It eliminates the need to replace the entire vulcanizing machine; only sensors, gateways, and edge control computers are added to the existing machine, preserving the original mechanical structure and hydraulic system to the greatest extent possible, thus reducing modification costs and downtime. The AI ​​algorithm platform is deployed in a containerized manner on the edge control computer, interfacing with the existing vulcanizing machine's PLC system for compatibility with older equipment. Data interfaces with the factory's MES system, WMS system, and rubber mixing / molding processes are established, enabling seamless integration of rubber batch information, tire specifications, production plans, and quality feedback throughout the entire process, providing the AI ​​model with complete operational characteristics. The implementation path follows a distributed approach: "single-unit pilot → small-batch verification → full-line rollout." First, a single vulcanizing machine is selected for modification and trial operation for 1-2 months to verify model accuracy and energy-saving effects. Then, this is expanded to a vulcanizing unit (4-6 units) for batch verification. Finally, based on data feedback, the model is rolled out to the entire workshop, ensuring that the modification process does not affect normal production. Initially, historical vulcanizing data is used for offline training of the initial model. Later, an online self-learning mode is activated, with each batch of finished product testing data fed back into the model to continuously update the knowledge base and model parameters, achieving a more accurate long-term evolution.

[0077] Through the above-mentioned modification process, traditional vulcanizing machines can be upgraded to vulcanizing machines equipped with artificial intelligence dynamic adaptive vulcanizing systems without interrupting production or replacing the main equipment. This allows for precise control of crosslinking density, avoiding over-vulcanization that leads to increased rolling resistance and accelerated aging. It perfectly matches the process requirements of EV quiet tires and low rolling resistance green tires, and fully aligns with the actual production conditions of tire factories.

[0078] The AI-based rubber tire vulcanization process control method relies on the aforementioned system to implement the complete AI vulcanization process flow, with the specific steps as follows: S1: The tire blank to be vulcanized enters the loading station before entering the vulcanizing machine. Each tire blank has an embedded identification tag in the molding process. The identification tag embedded in the tire blank is read by an RFID reader to obtain key information such as tire specifications, rubber batch, Mooney viscosity, tire blank weight and thickness. Based on the batch ID, the vulcanization characteristic parameters of the corresponding rubber and the historical vulcanization benchmark data of tires of the same specification are retrieved from the MES system or local knowledge base.

[0079] S2: Based on the retrieved data, the energy consumption adaptive curve model retrieves the initial temperature / pressure setting trajectory of the tire specification from the knowledge base; the vulcanization equivalent crosslinking degree prediction model estimates the basic vulcanization time window based on the Mooney viscosity of the rubber compound, the tire blank thickness, and the ambient temperature; the temperature-pressure coupling optimization model initializes the MPC controller state. After completing the above calculations, the vulcanizing machine controller (PLC) receives the positioning command, closes the mold, and pre-presses to the set position.

[0080] S3: During the vulcanization process, high-precision temperature sensors, pressure transmitters, and steam flow meters installed on the vulcanizing machine's hot plate, inside the capsule, and on the steam pipeline continuously collect multi-dimensional operating condition data such as steam temperature, outer mold temperature, capsule internal pressure, inner liner water temperature, hot plate temperature difference, and ambient temperature and humidity with a sampling period of Δt=2 seconds. Outliers are removed using sliding window mid-value filtering, missing values ​​are filled using linear interpolation, and feature vector sequences are generated through Z-score standardization to complete data preprocessing.

[0081] S4: Real-time calculation is performed through the sulfur equivalent crosslinking degree prediction model. The preprocessed feature vector is received in a sliding time window (window length of 60 sampling points). The instantaneous equivalent crosslinking degree α(t) and the remaining required sulfur equivalent Δα at the current moment are output in real time. It can correct the crosslinking rate deviation caused by temperature fluctuation or pressure change in real time.

[0082] S5: Coordinated calculation using a temperature-pressure coupled optimization model and an energy consumption adaptive curve model: the former uses the equivalent crosslinking degree α(t) relative to the target value α. target Driven by the deviation (typically 0.93-0.98, depending on the tire specifications), the Model Predictive Control (MPC) framework is used to continuously output the optimal valve opening and internal pressure settings. The latter dynamically adjusts the heating temperature of the vulcanizing machine and the internal pressure of the bladder based on the current energy consumption progress and crosslinking progress, so that the actual temperature and pressure accurately track the set trajectory.

[0083] S6: The intelligent early warning model for sulfurization anomalies calculates the risk level in real time. If the intervention conditions are met (e.g., moderate to high risk), the system will display an early warning on the large visualization screen and automatically execute online temperature or pressure correction actions: If there is a trend of undersulfurization, the temperature-pressure coupled optimization model will raise the temperature setpoint by 2-5℃ or extend the pressure holding time; if there is a risk of oversulfurization, the system will open the cooling water valve in advance or reduce the steam valve opening; when the risk is determined to be severe, the system will trigger emergency intervention and alarm, automatically notifying on-site inspection personnel. A snapshot of the process data at all early warning moments (from the first 30 seconds to the last 10 seconds) is automatically saved to the anomaly database for subsequent analysis.

[0084] S7: Determine whether the current equivalent crosslinking degree has reached the target crosslinking degree and tends to be stable (5 consecutive sampling points remain stable, fluctuation ≤0.005). If so, the artificial intelligence determines that vulcanization is complete, automatically terminates the heating and heat preservation process, the controller closes the steam valve, opens the drain valve and exhaust valve, and enters the pressure relief and cooling stage. After the pressure drops to the safe pressure (≤0.1MPa) and cools to the set temperature (capsule inner wall temperature ≤80℃), the automatic demolding and material discharge stage is entered. The vulcanizing machine automatically opens the mold, and the ejection mechanism pushes the vulcanized tire to the unloading roller.

[0085] S8: During the finished product inspection stage, appearance defects (such as bubbles, missing glue, etc.) and dynamic balance inspection data are obtained through the appearance inspection system and dynamic balance inspection equipment. The inspection results are then fed back to the algorithm model layer after being associated with the batch process data of the corresponding serial number to drive the iterative optimization of model parameters.

[0086] The above complete process realizes fully automated and intelligent operation from tire blank feeding to finished product inspection feedback. It can dynamically adapt to fluctuations in working conditions without human intervention, and significantly reduce energy consumption and defect rate while ensuring vulcanization quality. Comparison of key indicators before and after the renovation

[0087] This system was deployed on the rubber tire vulcanizing machine group of a tire factory. Key performance indicators were compared before and after the system was modified. After three months of continuous operation, the statistical results are as follows: Under-sulfurization / over-sulfurization failure rate: decreased by more than 85%; Overall tire product qualification rate: increased to 99.5%-99.8%; Steam energy consumption in the vulcanization process: reduced by 22%-35%; Production cycle time for vulcanized single tires: shortened by 8%-15%; Capsule lifespan: Increased by 15%-25%; Dependence on manual parameter adjustment has been largely eliminated, enabling unmanned and intelligent vulcanization. Tire durability performance: Improved by 5%-10%, with more stable low rolling resistance performance; Batch quality fluctuations are extremely poor: shrinkage of 60%.

[0088] The above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Any modifications, substitutions or equivalent changes made by those skilled in the art without departing from the spirit and scope of the present invention should fall within the protection scope of the claims of the present invention.

Claims

1. A rubber tire vulcanization process modification system based on artificial intelligence, characterized in that, include: Data acquisition layer, edge perception layer, artificial intelligence algorithm model layer, and process execution control layer; The data acquisition layer includes a sensor array arranged in the vulcanizing machine for real-time acquisition of multi-dimensional operating condition data during the vulcanizing process. The edge perception layer includes an edge computing device, which is used to receive the multi-dimensional working condition data and preprocess it to generate a corresponding feature vector sequence. The artificial intelligence algorithm model layer is deployed on the edge computing device or the cloud, and includes at least a sulfurization equivalent crosslinking degree prediction model, a temperature-pressure coupling optimization model, a sulfurization anomaly intelligent early warning model, and an energy consumption adaptive curve model. The artificial intelligence algorithm model layer performs collaborative operations based on the feature vector sequence and outputs dynamic process control commands. The process execution control layer is communicatively connected to the artificial intelligence algorithm model layer, and is used to receive the dynamic process control instructions and drive the thermal valve assembly and mechanical demolding mechanism of the vulcanizing machine to perform corresponding actions.

2. The artificial intelligence-based rubber tire vulcanization process modification system according to claim 1, characterized in that, The multi-dimensional operating condition data includes at least: steam temperature, outer mold temperature, capsule internal pressure, inner liner water temperature and vulcanization time; and also includes at least one of the following: hot plate temperature difference, ambient temperature and humidity, tire carcass weight deviation, rubber Mooney viscosity, workshop steam network pressure, waste heat recovery flow rate and cumulative capsule usage times.

3. The artificial intelligence-based rubber tire vulcanization process modification system according to claim 1, characterized in that, The vulcanization equivalent crosslinking degree prediction model adopts a time-series neural network that integrates physical information, and integrates the vulcanization kinetic equation and the Arrhenius equation during model training; the vulcanization equivalent crosslinking degree prediction model takes the sequence characteristics of multi-dimensional working condition data as input and outputs the real-time equivalent crosslinking degree and crosslinking change rate; when the real-time equivalent crosslinking degree reaches the set target crosslinking degree and tends to stabilize, the process execution control layer is triggered to terminate the vulcanization action.

4. The artificial intelligence-based rubber tire vulcanization process modification system according to claim 1, characterized in that, The temperature-pressure coupled optimization model is based on the model predictive control framework and introduces a thermal coupling transfer function matrix. It takes the measured temperature of each temperature zone, the steam pressure fluctuation and the internal pressure of the capsule, as well as the optimal temperature setpoint trajectory and pressure setpoint trajectory output by the energy consumption adaptive curve model as input. Through rolling optimization calculation, it outputs the optimal heating rate, the optimal constant temperature setpoint, the internal pressure control curve and the valve opening control quantity.

5. The artificial intelligence-based rubber tire vulcanization process modification system according to claim 1, characterized in that, The intelligent early warning model for abnormal sulfidation calculates the undersulfurization risk index and the oversulfurization risk index in real time, and assesses the comprehensive risk level based on the maximum value of the two. When the overall risk level reaches the set intervention threshold, it automatically triggers and outputs online corrective action commands for temperature, pressure, or time.

6. The artificial intelligence-based rubber tire vulcanization process modification system according to claim 1, characterized in that, The energy consumption adaptive curve model takes instantaneous steam flow rate, cumulative steam consumption and target degree of vulcanization as inputs, and uses a multi-objective optimization framework to output the optimal temperature setting trajectory and pressure setting trajectory. Based on the feedback status of real-time equivalent crosslinking degree, it dynamically adjusts the setting process parameters of the remaining vulcanization stage.

7. The artificial intelligence-based rubber tire vulcanization process modification system according to claim 1, characterized in that, The system also includes: a digital twin unit, used to construct a three-dimensional twin model of the vulcanizing machine and synchronously map real-time multi-dimensional working condition data for visual monitoring; and a visual quality inspection feedback module, deployed at the vulcanizing machine outlet, which includes image acquisition equipment and a defect detection algorithm model, used to identify tire surface defects and correlate the defect information with the vulcanizing process data of the corresponding batch to the artificial intelligence algorithm model layer to drive the adaptive iterative update of the model parameters.

8. The artificial intelligence-based rubber tire vulcanization process modification system according to claim 7, characterized in that, The system also includes a full-specification tire vulcanization process knowledge base and a capsule life extension management module. The process knowledge base stores the association records of tire specifications, rubber batches and optimal process parameters, which are used to retrieve and automatically match the initial vulcanization parameters through feature vectors when changing tire specifications. The capsule life extension management module is used to record and calculate the capsule life loss of a single vulcanization cycle, and works with the artificial intelligence algorithm model layer to limit the pressurization rate and depressurization rate to reduce material fatigue loss.

9. The artificial intelligence-based rubber tire vulcanization process modification system according to claim 1, characterized in that, The process execution control layer has a built-in fault-safe interlock mechanism: when the communication interruption between the edge perception layer and the process execution control layer exceeds the safety threshold, or when the multi-dimensional operating condition data exceeds the safety preset range, the thermal valve assembly is triggered to reset to the safe closed state and an audible and visual alarm is activated.

10. A method for controlling the vulcanization process of rubber tires based on artificial intelligence, applied to the artificial intelligence-based rubber tire vulcanization process modification system as described in any one of claims 1 to 9, characterized in that, Includes the following steps: S1: Obtain the identification information of the tire blank to be vulcanized, and retrieve the corresponding rubber compound parameters and historical vulcanization benchmark data; S2: Based on the retrieved data, initialize the target degree of crosslinking, initial temperature / pressure trajectory, and basic vulcanization time window; S3: During the vulcanization process, vulcanization condition data is collected and preprocessed in real time according to a preset cycle; S4: Real-time calculation and output of the current equivalent crosslinking degree and the remaining required vulcanization equivalent through the vulcanization equivalent crosslinking degree prediction model; S5: By coordinating the temperature-pressure coupling optimization model and the energy consumption adaptive curve model, the optimal valve opening and internal pressure setpoints are output, and the heating temperature and capsule internal pressure of the vulcanizing machine are dynamically adjusted. S6: The risk level is calculated in real time through the intelligent early warning model for abnormal sulfurization. If the intervention conditions are met, online process correction is performed. S7: Determine whether the current equivalent crosslinking degree has reached the target crosslinking degree. If so, automatically terminate the heating and heat preservation process and enter the pressure relief, cooling and demolding stage. S8: During the finished product inspection stage, acquire appearance defect and uniformity inspection data, and feed the inspection data back to the artificial intelligence algorithm model layer for self-learning and parameter iteration.