Intelligent process parameter optimization method for rubber banburying

Abstract

The invention belongs to the technical field of rubber mixing processes, and particularly relates to an intelligent process parameter optimization method for rubber mixing. The method comprises the following steps: S1, acquiring production data by using an intelligent internal mixing AI management and control platform; s2, extracting training data from the production data, training the AI process learning model, and storing the training data in an intelligent internal mixing AI management and control platform; s3, the intelligent internal mixing AI management and control platform issues the trained AI process learning model to an AI intelligent terminal; s4, starting an AI process operation program, collecting process data of the internal mixing process in real time, inputting the collected data into the trained AI process learning model for optimizing process parameters, and outputting an optimized internal mixing AI process; s5, the upper auxiliary machine PLC terminal receives the internal mixing AI process and issues the process to the internal mixer; and S6, the internal mixer executes an internal mixing AI process. The system has the characteristics that the automation and intelligence of the internal mixing process can be realized, and the quality consistency and the production efficiency are improved.

Description

Technical Field

[0001] The invention belongs to the technical field of rubber mixing technology, and in particular relates to an intelligent process parameter optimization method for rubber mixing. Background Art

[0002] The mixing process of rubber is one of the key links in the production process of rubber products. It mainly refers to the uniform mixing of various compounding agents (such as fillers, vulcanizers, plasticizers, antioxidants, etc.) with raw rubber to achieve the required processing and use performance. The purpose of the mixing process is to allow the various ingredients in the formula to be fully dispersed in the rubber matrix to form a uniform mixture to ensure that the physical properties and chemical stability of the rubber products meet the requirements. The mixing process generally includes plasticizing, master mixing, and final mixing. The rubber that meets the technical requirements is used in the semi-finished product process.

[0003] 80% of the raw materials required for tire manufacturing are in the mixing process, about more than 100 kinds, and manufacturers frequently switch batches. At present, the rubber mixing process is formulated by the research institute, and the process technicians formulate the production process based on the production formula. The production equipment is executed according to the process steps and process parameters (such as time, temperature, energy, speed, and pressure) set by the process technicians. Generally, the rubber is discharged according to the set combination of mixing time, mixing temperature, and mixing energy. After many small batch tests and multiple process adjustments, a set of processes is determined until the production products fail the inspection. Due to factors such as raw material changes, weather changes, and performance requirements, process personnel need to frequently adjust process parameters, resulting in slow response of process technicians, and they cannot fully cope with the fluctuations of raw materials and product performance requirements, which greatly consumes manpower and financial resources and product quality control cannot be guaranteed.

[0004] Therefore, it is very important to design an intelligent process parameter optimization method for rubber mixing that can realize the automation and intelligence of the mixing process and improve quality consistency and production efficiency. Summary of the invention

[0005] The present invention aims to overcome the problems in the prior art that the traditional rubber mixing process requires manual modification, the process feedback is slow and not immediate, and the production conditions cannot be met. An intelligent process parameter optimization method for rubber mixing is provided, which can realize the automation and intelligence of the mixing process and improve the quality consistency and production efficiency.

[0006] In order to achieve the above-mentioned object of the invention, the present invention adopts the following technical solutions:

[0007] The intelligent process parameter optimization method for rubber mixing comprises the following steps:

[0008] S1, use the intelligent mixing AI control platform to obtain production data;

[0009] S2, extracting training data from the production data, training the AI ​​process learning model, and storing it in the intelligent mixing AI management and control platform;

[0010] S3, the intelligent mixing AI management and control platform sends the trained AI process learning model to the AI ​​intelligent terminal;

[0011] S4, start the AI ​​process operation program, and collect the mixing process data in real time, and input the collected data into the trained AI process learning model to optimize the process parameters, and output the optimized mixing AI process;

[0012] S5, the upper auxiliary machine PLC terminal receives the AI ​​process of internal mixing and sends it to the internal mixer;

[0013] S6, the internal mixer performs the internal mixing AI process.

[0014] Preferably, in step S1, the production data includes mixing process data, step sequence report, quality inspection data, raw material data, feeding data and weather information.

[0015] Preferably, step S2 comprises the following steps:

[0016] S21, clean the production data, use the mean-shift algorithm to cluster the process parameters of different periods, and save all versions of process parameters in the intelligent mixing AI management and control platform according to the formula category;

[0017] S22, clean the production data, use the DQN algorithm for reinforcement learning, and save the AI ​​process learning model obtained based on the formula reinforcement learning in the intelligent mixing AI management and control platform.

[0018] Preferably, step S21 includes the following steps:

[0019] S211, performing outlier processing and feature extraction on the collected production data to process it into a required data format;

[0020] S212, for each recipe, applying the corresponding process parameters to the clustering process;

[0021] The clustering process is as follows:

[0022] Initialize the data points to the mean of each point and select the bandwidth parameter h; for the data point x, the corresponding Mean-Shift vector is expressed as:

[0023]

[0024] Where N(x) is the set of data points within the bandwidth parameter h, that is, the neighborhood of point x; x irepresents the i-th data point; parameter K represents the kernel function, specifically the Gaussian kernel function; the size and direction of the Mean-Shift vector depend on the data distribution in the neighborhood;

[0025] For each data point x, keep moving towards the high density direction, that is, execute: x←x+m(x), repeat several times until it converges to the local maximum density point; if the convergence point distance of adjacent data points is less than the set threshold, the corresponding data points are merged into a cluster; the final centroid and corresponding data point are the clustering result;

[0026] S213, save the characteristics and process parameters of multiple categories of each formula in the intelligent mixing AI management and control platform; the clustered process parameters are used to intelligently recommend processes based on changes in production conditions.

[0027] Preferably, step S22 includes the following steps:

[0028] S221, processing abnormal values ​​and missing values ​​in production data, specifically including missing value interpolation filling, data smoothing and data standardization;

[0029] S222, extracting features from the data processed in step S221, and filtering out effective feature data through statistical analysis and dimensionality reduction operations;

[0030] S223, dividing the valid feature data into a training set, a validation set, and a test set for subsequent model training and evaluation;

[0031] S224, using the DQN algorithm for reinforcement learning, the specific process is as follows:

[0032] Determine the state space of the mixing process, including process data, material feeding data and mixing quality data;

[0033] Define the action space of the mixing process, including speed, pressure, time, temperature, and energy;

[0034] Design reward functions based on the quality requirements of mixing, and design immediate rewards considering production efficiency and product quality;

[0035] Build a DQN network model, construct a simulation environment based on historical production data, use the DQN algorithm for reinforcement learning training, and gradually optimize the strategy of the DQN network model;

[0036] Save the trained DQN network model in the intelligent refining AI management and control platform;

[0037] Establish a feedback mechanism to collect DQN network model effect data in actual production, evaluate the performance of the DQN network model, and regularly retrain or fine-tune it based on new production data.

[0038] Preferably, step S4 comprises the following steps:

[0039] S41, during production, the AI ​​process operation program obtains the current production formula information and material feeding information from the MES interface, obtains the corresponding raw material inspection data from the WMS system interface, obtains weather data from the intelligent mixing AI management and control platform, and uses the cosine similarity algorithm to calculate the AI ​​process that best matches the current production environment;

[0040] S42, the AI ​​operation program on the AI ​​intelligent terminal writes the generated process parameters to the upper auxiliary machine PLC terminal, and notifies the upper auxiliary machine PLC terminal that the process parameters have been written;

[0041] S43, after the upper auxiliary machine PLC terminal receives the new process parameters written by the AI ​​process operation program, it replaces the currently executed process parameters;

[0042] S44, the internal mixer executes the AI ​​process to produce rubber, and the AI ​​process operation program captures real-time process data at a fixed frequency through the auxiliary machine PLC terminal;

[0043] S45, the AI ​​process operation program on the AI ​​intelligent terminal further optimizes the new process parameters according to the real-time process data, and generates a refining AI process, which is then re-sent to the upper auxiliary machine PLC terminal through the AI ​​operation program.

[0044] Preferably, step S41 includes the following steps:

[0045] S411, obtain the current production formula information and material feeding information from the data interface of the MES system, obtain the raw material inspection data and origin data from the data interface of the WMS system, and obtain the weather data from the intelligent mixing AI management and control platform;

[0046] S412, associating quality inspection data and raw material data through formula information and material input information to obtain current formula material composition and material inspection data;

[0047] S413, by processing the data obtained in step S411 and step S412, a feature vector A of the current feeding information is obtained. n ); the data processing process includes data formatting, outlier processing, missing value filling, feature extraction, numerical encoding and normalization of classification features;

[0048] Grab the clustering results of the current production formula of the intelligent mixing AI control platform, traverse each category of results, and extract the corresponding feature vector B=(b1, b2,…, b n ), calculate the cosine similarity:

[0049]

[0050] The class with the largest cosine similarity is taken as the process that best matches the current production environment.

[0051] Preferably, step S42 includes the following steps:

[0052] S421, the AI ​​intelligent terminal runs the AI ​​computing program to intelligently generate process parameters based on current production requirements, historical data, and raw material data;

[0053] S422, encoding the generated process parameters in a format that can be understood by the PLC;

[0054] S423, transmitting the encoded process parameters to the upper auxiliary machine PLC terminal via industrial Ethernet;

[0055] S424, after the transmission is completed, the AI ​​intelligent terminal notifies the auxiliary machine PLC terminal that the new process parameters have been written by updating the status of the corresponding monitoring points. After the PLC detects the status update, it sends the process parameters for execution.

[0056] Preferably, step S45 includes the following steps:

[0057] S451, after the AI ​​process operation program obtains the complete real-time process data, it performs analysis and processing and extracts the process curve characteristics;

[0058] S452, call the trained AI process learning model to further optimize the process parameters, generate the optimized AI mixing process, and then send it back to the upper auxiliary machine PLC terminal through the AI ​​operation program.

[0059] Preferably, step S6 comprises the following steps:

[0060] S61, conduct quality inspection on the produced rubber materials and synchronize the inspection results to the MES system;

[0061] S62, the intelligent mixing AI control platform collects the rubber inspection data and process data obtained from the mixing AI process production;

[0062] S63, adding the newly generated data in step S62 to the original production data, performing incremental learning, and repeating the process from step S1 to step S5.

[0063] Compared with the prior art, the present invention has the following beneficial effects: (1) the method of the present invention can improve product consistency; (2) the method of the present invention can improve production efficiency; (3) the method of the present invention uses data models to drive decisions and reduce interference from human factors; (4) the method of the present invention can reduce ineffective mixing time and reduce energy consumption; (5) the method of the present invention can reduce the burden on process technicians and allow more time and energy to be used to monitor and adjust the entire production process; (6) the method of the present invention increases the flexibility and diversity of the process and provides data support for further improving the process level. BRIEF DESCRIPTION OF THE DRAWINGS

[0064] Figure 1 A principle framework diagram of the intelligent process parameter optimization method for rubber mixing of the present invention;

[0065] Figure 2 A basic flow chart of the intelligent process parameter optimization method for rubber mixing of the present invention;

[0066] Figure 3 The present invention is a detailed flow chart of the intelligent process parameter optimization method for rubber mixing. DETAILED DESCRIPTION

[0067] In order to more clearly illustrate the embodiments of the present invention, the specific implementation methods of the present invention will be described below with reference to the accompanying drawings. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those of ordinary skill in the art, other accompanying drawings and other implementation methods can be obtained based on these accompanying drawings without creative work.

[0068] like Figure 1 As shown, in the method of the present invention, the process learning module cleans and processes the MES data, uses a clustering algorithm to generate process parameters for different periods and stores them in the AI ​​intelligent management and control platform. At the same time, a deep reinforcement learning model is trained. During production, the intelligent mixing software first uses a recommendation algorithm to find the most suitable process parameters based on the production information, and sends them to the mixer through the upper auxiliary machine PLC to obtain the mixing process data. Then, according to the mixing process data fed back by the upper auxiliary machine PLC, the trained reinforcement learning model is used to further optimize the process. Finally, the AI ​​process is sent to the mixer through the upper auxiliary machine PLC for execution.

[0069] like Figure 2 As shown, the present invention provides an intelligent process parameter optimization method for rubber mixing, comprising the following steps:

[0070] S1, use the intelligent mixing AI control platform to obtain production data;

[0071] S2, extracting training data from the production data, training the AI ​​process learning model, and storing it in the intelligent mixing AI management and control platform;

[0072] S3, the intelligent mixing AI management and control platform sends the trained AI process learning model to the AI ​​intelligent terminal;

[0073] S4, start the AI ​​process operation program, and collect the mixing process data in real time, and input the collected data into the trained AI process learning model to optimize the process parameters, and output the optimized mixing AI process;

[0074] S5, the upper auxiliary machine PLC terminal receives the AI ​​process of internal mixing and sends it to the internal mixer;

[0075] S6, the internal mixer performs the internal mixing AI process.

[0076] Detailed steps, such as Figure 2 As shown, the specific process is as follows:

[0077] S11, the data platform obtains production data through MES, WMS and other systems, such as mixing process data, step reports, quality inspection data, raw material data, feeding data, weather information, etc.

[0078] S21, clean the production data, use the mean-shift algorithm to cluster the process parameters of different periods, and save all versions of process parameters in the intelligent mixing AI management and control platform according to the formula category;

[0079] S22, clean the production data, use the DQN algorithm for reinforcement learning, and save the AI ​​process learning model obtained based on the formula reinforcement learning in the intelligent mixing AI management and control platform.

[0080] Wherein, step S21 specifically includes the following steps:

[0081] S211, performing outlier processing and feature extraction on the collected production data to process it into a required data format;

[0082] S212, for each recipe, applying the corresponding process parameters to the clustering process;

[0083] The clustering process is as follows:

[0084] Initialize the data points to the mean of each point and select the bandwidth parameter h; for the data point x, the corresponding Mean-Shift vector is expressed as:

[0085]

[0086] Where N(x) is the set of data points within the bandwidth parameter h, that is, the neighborhood of point x; x i represents the i-th data point; parameter K represents the kernel function, specifically the Gaussian kernel function; the size and direction of the Mean-Shift vector depend on the data distribution in the neighborhood;

[0087] For each data point x, keep moving towards the high density direction, that is, execute: x←x+m(x), repeat several times until it converges to the local maximum density point; if the convergence point distance of adjacent data points is less than the set threshold, the corresponding data points are merged into a cluster; the final centroid and corresponding data point are the clustering result;

[0088] S213, save the characteristics and process parameters of multiple categories of each formula in the intelligent mixing AI management and control platform; the clustered process parameters are used to intelligently recommend processes based on changes in production conditions.

[0089] Step S22 specifically includes the following steps:

[0090] S221, processing abnormal values ​​and missing values ​​in production data, specifically including missing value interpolation filling, data smoothing and data standardization;

[0091] S222, extracting features from the data processed in step S221, and filtering out effective feature data through statistical analysis and dimensionality reduction operations;

[0092] S223, dividing the valid feature data into a training set, a validation set, and a test set for subsequent model training and evaluation;

[0093] S224, using the DQN algorithm for reinforcement learning, the specific process is as follows:

[0094] Determine the state space of the mixing process, including process data, material feeding data and mixing quality data;

[0095] Define the action space of the mixing process, including speed, pressure, time, temperature, and energy;

[0096] Design reward functions based on the quality requirements of mixing, and design immediate rewards considering production efficiency and product quality;

[0097] Build a DQN network model, construct a simulation environment based on historical production data, use the DQN algorithm for reinforcement learning training, and gradually optimize the strategy of the DQN network model;

[0098] Save the trained DQN network model in the intelligent refining AI management and control platform;

[0099] Establish a feedback mechanism to collect DQN network model effect data in actual production, evaluate the performance of the DQN network model, and regularly retrain or fine-tune it based on new production data.

[0100] S31, after the training is completed, the trained process learning model (DQN network model) is sent to the corresponding AI smart terminal.

[0101] S41, during production, the AI ​​process operation program obtains the current production formula information and material feeding information from the MES interface, obtains the corresponding raw material inspection data from the WMS system interface, obtains weather data from the intelligent mixing AI management and control platform, and uses the cosine similarity algorithm to calculate the AI ​​process that best matches the current production environment;

[0102] S42, the AI ​​operation program on the AI ​​intelligent terminal writes the generated process parameters to the upper auxiliary machine PLC terminal, and notifies the upper auxiliary machine PLC terminal that the process parameters have been written;

[0103] S43, after the upper auxiliary machine PLC terminal receives the new process parameters written by the AI ​​process operation program, it replaces the currently executed process parameters;

[0104] S44, the internal mixer executes the AI ​​process to produce rubber, and the AI ​​process operation program captures real-time process data at a fixed frequency through the auxiliary machine PLC terminal;

[0105] S45, the AI ​​process operation program on the AI ​​intelligent terminal further optimizes the new process parameters according to the real-time process data, and generates a refining AI process, which is then re-sent to the upper auxiliary machine PLC terminal through the AI ​​operation program.

[0106] Wherein, step S41 specifically includes the following steps:

[0107] S411, obtain the current production formula information and material feeding information from the data interface of the MES system, obtain the raw material inspection data and origin data from the data interface of the WMS system, and obtain the weather data from the intelligent mixing AI management and control platform;

[0108] S412, associating quality inspection data and raw material data through formula information and material input information to obtain current formula material composition and material inspection data;

[0109] S413, by processing the data obtained in step S411 and step S412, a feature vector A of the current feeding information is obtained. n ); the data processing process includes data formatting, outlier processing, missing value filling, feature extraction, numerical encoding of classification features, normalization, etc. commonly used in the prior art;

[0110] Grab the clustering results of the current production formula of the intelligent mixing AI control platform, traverse each category of results, and extract the corresponding feature vector B=(b1, b2,…, b n ), calculate the cosine similarity:

[0111]

[0112] The class with the largest cosine similarity is taken as the process that best matches the current production environment.

[0113] Step S42 includes the following steps:

[0114] S421, the AI ​​intelligent terminal runs the AI ​​computing program to intelligently generate process parameters based on current production requirements, historical data, and raw material data;

[0115] S422, encoding the generated process parameters in a format that can be understood by the PLC;

[0116] S423, transmitting the encoded process parameters to the upper auxiliary machine PLC terminal via industrial Ethernet;

[0117] S424, after the transmission is completed, the AI ​​intelligent terminal notifies the auxiliary machine PLC terminal that the new process parameters have been written by updating the status of the corresponding monitoring points. After the PLC detects the status update, it sends the process parameters for execution.

[0118] Preferably, step S45 includes the following steps:

[0119] S451, after the AI ​​process operation program obtains the complete real-time process data, it performs analysis and processing.

[0120] And extract the process curve characteristics;

[0121] S452, call the trained AI process learning model to further optimize the process parameters, generate the optimized AI mixing process, and then send it back to the upper auxiliary machine PLC terminal through the AI ​​operation program.

[0122] S5, the upper auxiliary machine PLC terminal receives the AI ​​process of internal mixing and sends it to the internal mixer.

[0123] S61, conduct quality inspection on the produced rubber materials and synchronize the inspection results to the MES system;

[0124] S62, the intelligent mixing AI control platform collects the rubber inspection data and process data obtained from the mixing AI process production;

[0125] S63, adding the newly generated data in step S62 to the original production data, performing incremental learning, and repeating the process from step S1 to step S5.

[0126] At this point, the smart process is complete.

[0127] The system of the present invention is composed of an AI process learning model and an AI process operation program, and the model training is carried out in the form of "clustering algorithm" + "recommendation algorithm" + "reinforcement learning". Nearly 3 million MES production data are synchronized through the AI ​​intelligent internal mixing control platform, and after data extraction using data mining technology, the process parameters of different periods are clustered using the mean-shift algorithm. At the same time, the AI ​​reinforcement learning model is trained using the reinforcement learning DQN algorithm. After the model is generated, the AI ​​control platform will automatically send it to the corresponding AI intelligent terminal. During the initial operation, the AI ​​process operation program will recommend a set of process parameters based on the feeding information and weather information, and send it to the internal mixer through the upper auxiliary machine PLC for execution. At the same time, the process data of the MES system and the upper auxiliary machine PLC are collected in real time for processing, and the processed features are input into the AI ​​process learning model to achieve real-time optimization of the production process. The AI ​​control platform will extract the data of the rubber after quality inspection to evaluate the control effect of the model process. If the control effect does not meet expectations, the newly generated data will be incrementally trained on the basis of the original model and re-sent to the corresponding AI intelligent terminal. After the system was put into operation, the final Mooney qualification rate increased from 91.75% to 98.55%, and the carbon black dispersion qualification rate increased from 78.85% to 96.5%. The mixing time was reduced by 2.6 seconds, which improved the quality of the rubber compound while reducing energy consumption and improving production efficiency.

[0128] The present invention mainly solves the drawbacks of the current rubber mixing process from the perspective of intelligence, automation and adaptability. The present invention focuses on utilizing the MES manufacturing execution system, WMS warehouse management system and PLC control system, combined with clustering algorithms, recommendation algorithms and deep reinforcement learning algorithms, to automatically generate the optimal mixing process according to changes in production conditions, and to send it to the mixer for execution through the upper auxiliary machine PLC, and can optimize the process in real time during the production execution process.

[0129] The above description is only a detailed description of the preferred embodiments and principles of the present invention. For ordinary technicians in this field, according to the ideas provided by the present invention, there will be changes in the specific implementation methods, and these changes should also be regarded as the protection scope of the present invention.

Claims

1. An intelligent process parameter optimization method for rubber mixing, characterized in that: The steps include: S1, use the intelligent mixing AI control platform to obtain production data; S2, extracting training data from the production data, training the AI ​​process learning model, and storing it in the intelligent mixing AI management and control platform; S3, the intelligent mixing AI management and control platform sends the trained AI process learning model to the AI ​​intelligent terminal; S4, start the AI ​​process operation program, and collect the mixing process data in real time, and input the collected data into the trained AI process learning model to optimize the process parameters, and output the optimized mixing AI process; S5, the upper auxiliary machine PLC terminal receives the AI ​​process of internal mixing and sends it to the internal mixer; S6, the internal mixer performs the internal mixing AI process.

2. The intelligent process parameter optimization method for rubber mixing according to claim 1, characterized in that: In step S1, the production data includes mixing process data, step sequence report, quality inspection data, raw material data, feeding data and weather information.

3. The intelligent process parameter optimization method for rubber mixing according to claim 2, characterized in that: Step S2 includes the following steps: S21, clean the production data, use the mean-shift algorithm to cluster the process parameters of different periods, and save all versions of process parameters in the intelligent mixing AI management and control platform according to the formula category; S22, clean the production data, use the DQN algorithm for reinforcement learning, and save the AI ​​process learning model obtained based on the formula reinforcement learning in the intelligent mixing AI management and control platform.

4. The intelligent process parameter optimization method for rubber mixing according to claim 3 is characterized in that: Step S21 includes the following steps: S211, performing outlier processing and feature extraction on the collected production data to process it into a required data format; S212, for each recipe, applying the corresponding process parameters to the clustering process; The clustering process is as follows: Initialize the data points to the mean of each point and select the bandwidth parameter h; for data point z, the corresponding Mean-Shift vector m(x) is expressed as: Where N(x) is the set of data points within the bandwidth parameter h, that is, the neighborhood of point x; x i represents the i-th data point; parameter K represents the kernel function, specifically the Gaussian kernel function; the size and direction of the Mean-Shift vector depend on the data distribution in the neighborhood; For each data point x, keep moving towards the high density direction, that is, execute: x←x+m(x), repeat several times until it converges to the local maximum density point; if the convergence point distance of adjacent data points is less than the set threshold, the corresponding data points are merged into a cluster; the final centroid and corresponding data point are the clustering result; S213, save the characteristics and process parameters of multiple categories of each formula in the intelligent mixing AI management and control platform; the clustered process parameters are used to intelligently recommend processes based on changes in production conditions.

5. The intelligent process parameter optimization method for rubber mixing according to claim 3, characterized in that: Step S22 includes the following steps: S221, processing abnormal values ​​and missing values ​​in production data, specifically including missing value interpolation filling, data smoothing and data standardization; S222, extracting features from the data processed in step S221, and filtering out effective feature data through statistical analysis and dimensionality reduction operations; S223, dividing the valid feature data into a training set, a validation set, and a test set for subsequent model training and evaluation; S224, using the DQN algorithm for reinforcement learning, the specific process is as follows: Determine the state space of the mixing process, including process data, material feeding data and mixing quality data; Define the action space of the mixing process, including speed, pressure, time, temperature, and energy; Design reward functions based on the quality requirements of mixing, and design immediate rewards considering production efficiency and product quality; Build a DQN network model, construct a simulation environment based on historical production data, use the DQN algorithm for reinforcement learning training, and gradually optimize the strategy of the DQN network model; Save the trained DQN network model in the intelligent refining AI management and control platform; Establish a feedback mechanism to collect DQN network model effect data in actual production, evaluate the performance of the DQN network model, and regularly retrain or fine-tune it based on new production data.

6. The intelligent process parameter optimization method for rubber mixing according to claim 1, characterized in that: Step S4 includes the following steps: S41, during production, the AI ​​process operation program obtains the current production formula information and material feeding information from the MES interface, obtains the corresponding raw material inspection data from the WMS system interface, obtains weather data from the intelligent mixing AI management and control platform, and uses the cosine similarity algorithm to calculate the AI ​​process that best matches the current production environment; S42, the AI ​​operation program on the AI ​​intelligent terminal writes the generated process parameters to the upper auxiliary machine PLC terminal, and notifies the upper auxiliary machine PLC terminal that the process parameters have been written; S43, after receiving the new process parameters written by the AI ​​process operation program, the upper auxiliary machine PLC terminal replaces the currently executed process parameters; S44, the internal mixer executes the AI ​​process to produce rubber, and the AI ​​process operation program captures real-time process data at a fixed frequency through the auxiliary machine PLC terminal; S45, the AI ​​process operation program on the AI ​​intelligent terminal further optimizes the new process parameters according to the real-time process data, and generates a refining AI process, which is then re-sent to the upper auxiliary machine PLC terminal through the AI ​​operation program.

7. The intelligent process parameter optimization method for rubber mixing according to claim 6, characterized in that: Step S41 includes the following steps: S411, obtain the current production formula information and material feeding information from the data interface of the MES system, obtain the raw material inspection data and origin data from the data interface of the WMS system, and obtain the weather data from the intelligent mixing AI management and control platform; S412, associating quality inspection data and raw material data through formula information and material input information to obtain current formula material composition and material inspection data; S413, by processing the data obtained in step S411 and step S412, a feature vector A of the current feeding information is obtained. n ); the data processing process includes data formatting, outlier processing, missing value filling, feature extraction, numerical encoding and normalization of classification features; Grab the clustering results of the current production formula of the intelligent mixing AI control platform, traverse each category of results, and extract the corresponding feature vector B=(b1, b2,…, b n ), calculate the cosine similarity: The class with the largest cosine similarity is taken as the process that best matches the current production environment.

8. The intelligent process parameter optimization method for rubber mixing according to claim 7, characterized in that: Step S42 includes the following steps: S421, the AI ​​intelligent terminal runs the AI ​​computing program to intelligently generate process parameters based on current production requirements, historical data, and raw material data; S422, encoding the generated process parameters in a format that can be understood by the PLC; S423, transmitting the encoded process parameters to the upper auxiliary machine PLC terminal via industrial Ethernet; S424, after the transmission is completed, the AI ​​intelligent terminal notifies the auxiliary machine PLC terminal that the new process parameters have been written by updating the status of the corresponding monitoring points. After the PLC detects the status update, it sends the process parameters for execution.

9. The intelligent process parameter optimization method for rubber mixing according to claim 6, characterized in that: Step S45 includes the following steps: S451, after the AI ​​process operation program obtains the complete real-time process data, it performs analysis and processing and extracts the process curve characteristics; S452, call the trained AI process learning model to further optimize the process parameters, generate the optimized AI mixing process, and then send it back to the upper auxiliary machine PLC terminal through the AI ​​operation program.

10. The intelligent process parameter optimization method for rubber mixing according to claim 1, characterized in that: Step S6 includes the following steps: S61, conduct quality inspection on the produced rubber materials and synchronize the inspection results to the MES system; S62, the intelligent mixing AI control platform collects the rubber inspection data and process data obtained from the mixing AI process production; S63, adding the newly generated data in step S62 to the original production data, performing incremental learning, and repeating the process from step S1 to step S5.

Images