Rubber extrusion intelligent formulation method and system

Abstract

The present application relates to rubber extrusion formula technology, discloses a kind of rubber extrusion intelligent formula method and system, its method includes: intelligent control management end uses mean-shift clustering algorithm to the recent production data clustering, calculates each specification recent optimal formula information;Intelligent control management end uses historical production data, and the influence factor in formula is regressed to different glue material and different production line state by random forest algorithm;When producing, intelligent control management end obtains the optimal formula information by PLC switching glue material specification from data platform and is automatically issued to machine table by specification clustering;After producing a period of time, intelligent terminal obtains glue material and production line information, is transmitted to random forest model, obtains standard correction and is issued to machine table.The rubber extrusion formula of the present application is high in efficiency, and stability and robustness are good.

Description

Technical Field

[0001] This invention relates to rubber extrusion formulation technology, and more particularly to a smart rubber extrusion formulation method and system. Background Technology

[0002] Semi-finished product extrusion is one of the four major steps in tire manufacturing. It is a process in which rubber compound is continuously extruded through an extruder to form semi-finished products of various shapes. After the rubber compound is fed into the barrel through the feeding port, the rotation of the screw causes the rubber compound to be subjected to a strong extrusion force between the screw and the barrel wall, forcing the rubber compound to move forward continuously and pass through the die, thereby obtaining semi-finished products of various shapes. The extrusion process generally includes rubber compound hot refining, extrusion, cooling, cutting, and receiving processes; except for hot refining, the other processes are all integrated into a single operation, with the total length of the production line ranging from 120 meters to 200 meters. Summary of the Invention

[0003] This invention addresses the problem that existing rubber extrusion processes are still primarily controlled manually or by expert systems. During production, parameters and formulations are recorded by operators in notebooks and adjusted based on experience, resulting in inconsistent effectiveness and a high dependence on operator experience. Automated programs based on operator experience generally cannot adapt to all specifications and exhibit poor robustness and stability. This invention provides an intelligent formulation method and system for rubber extrusion.

[0004] To solve the above-mentioned technical problems, the present invention provides the following technical solution:

[0005] A method for intelligent formulation of rubber extrusion includes an intelligent control and management terminal, an intelligent terminal, and a machine; the method includes:

[0006] The intelligent control and management terminal uses the mean-shift clustering algorithm to cluster recent production data and calculate the recent optimal formula information for each specification;

[0007] The intelligent control and management terminal uses historical production data to regress the influence factors of different rubber compounds and different production line states on the formula through the random forest algorithm.

[0008] During production, the intelligent control and management terminal switches the rubber specifications through the PLC, obtains the optimal formula information from the data platform by specification clustering, and automatically sends it to the machine.

[0009] After a period of production, the smart terminal obtains information about the rubber material and production line, transmits it to the random forest model, obtains standard corrections, and sends them to the machine.

[0010] As a preferred option, recent production data includes rubber specifications, line speed, upper extruder speed, middle extruder speed, lower extruder speed, front shoulder width, front total width, front weight, rear shoulder width, rear total width, and rear weight.

[0011] Preferably, historical production data includes rubber specifications, line speed, upper extruder speed, middle extruder speed, lower extruder speed, front shoulder width, front total width, front weight, rear shoulder width, rear total width, rear weight, upper extruder pressure, middle extruder pressure, lower extruder pressure, upper extruder temperature, middle extruder temperature, lower extruder temperature, upper extruder current, middle extruder current, lower extruder current, Mooney compound, compound hardness, compound specific gravity, and compound sulfur grade.

[0012] As a preferred method, the mean-shift clustering algorithm for clustering recent production data includes:

[0013] The intelligent control and management terminal collects recent production data and categorizes the acquired data.

[0014] Recent production data preprocessing is performed using a time-lag cross-correlation algorithm.

[0015] Initialize the processed recent production data by randomly selecting one data point as the starting center point x. 起始 ;

[0016] The bandwidth parameter h is estimated using the estimate_bandwidth function, thereby determining the updated center point position x. 更新， x 更新 =x 起始 +m(x);

[0017] m(x) is the drift vector.

[0018]

[0019] Among them, K h (u) is a Gaussian kernel function with bandwidth h.

[0020]

[0021] Kernel function satisfies

[0022] By obtaining the drift vector, the iteration stops when ||m(x)|| < ε or the number of iterations reaches the set maximum value T; otherwise, the above steps are returned to continue, thereby obtaining the optimal rubber formula based on recent production data.

[0023] As a preferred approach, the time-lag cross-correlation algorithm preprocesses recent production data, including...

[0024] Let the initial weight sequence be X = {x1, x2, ..., xn}n}

[0025] The subsequent weight sequence is Y = {y1, y2, ..., y} n}, where n is the length of the average length;

[0026] Time lag cross-correlation function R XY (τ) represents the correlation between time series Y and X at a time lag τ.

[0027]

[0028] Where τ represents the time lag; Let X and Y be the means of the time series. The mean of the time series XY; x t The values ​​of time series X at time points τ and t+τ, y t+τ These are the values ​​of time series X and Y at time points τ and t+τ, respectively; τ max =argmax τ |R XY (τ)|

[0029] Where, τ max The amount of weight that needs to be measured in advance is the total width of the back shoulders.

[0030] As a preferred method, the implementation of regression analysis using the random forest algorithm to determine the influencing factors of different rubber compounds and production line conditions on the formulation includes:

[0031] Historical production data acquisition: The intelligent control and management terminal collects historical production data.

[0032] Historical production data preprocessing is performed using a time-lag cross-correlation algorithm.

[0033] A random forest model was constructed using preprocessed historical production data.

[0034] The parameters of the random forest model are optimized, and the random forest model is trained using the optimized parameters.

[0035] Preferably, the parameters of the random forest model are optimized, and the model is trained using the optimized parameters. Tree-based Parsons optimization is then used to further optimize the random forest model. This can be achieved through methods such as:

[0036] Choose an initial hyperparameter x and compute two probabilistic models L(x) and G(x);

[0037] l(x) represents the probability that the objective function value is lower than a certain threshold γ given the hyperparameter x; L(x) = P(x|y<γ)

[0038] G(x) represents the probability (i.e., the parameter space of the "difference") that the objective function value is higher than a certain threshold γ given the hyperparameter x: G(x) = P(x|y≥γ)

[0039] p(y|x) is the conditional probability density function of the objective function value y given the hyperparameters x:

[0040] P(y|x)=L(x)·L(y<γ)+G(x)·G(y≥γ)

[0041] EI(x) is used to select hyperparameters x for evaluation and to record the objective function value. EI(x) is used to measure the expected improvement of the objective function value relative to the current best value γ after selecting a set of hyperparameters x.

[0042]

[0043] Where, μ g and σ g Φ(...) and φ(...) are the mean and standard deviation of the objective function value under G(x), respectively, and Φ(...) and φ(...) are the cumulative distribution function and probability density function of the standard normal distribution, respectively.

[0044] To address the aforementioned technical problems, this application also provides a smart rubber extrusion formulation system, comprising a smart control and management terminal, a smart terminal, and a machine; characterized in that it further includes,

[0045] The module for obtaining optimal formula information uses the mean-shift clustering algorithm to cluster recent production data and calculate the recent optimal formula information for each specification in the intelligent control and management terminal.

[0046] The Random Forest Regression Module uses historical production data in the intelligent control and management terminal to regress the influence factors of different rubber compounds and different production line states on the formulation through the Random Forest algorithm.

[0047] During production, the intelligent control and management terminal switches the rubber specifications through the PLC, obtains the optimal formula information from the data platform by specification clustering, and automatically sends it to the machine.

[0048] After a period of production, the smart terminal obtains information about the rubber material and production line, transmits it to the random forest model, obtains standard corrections, and sends them to the machine.

[0049] This invention, by adopting the above technical solutions, has significant technical effects:

[0050] This system consists of a formula model and a dynamic control algorithm, employing clustering and iterative methods for parameter adjustment. It captures 120,000 data points on production line speed, weight per meter, and single-line weight from a digital platform. The Meanshift clustering model is used for analysis to generate new production formula data, including weight per meter, extrusion speed, and rotation speed, which is then distributed to the on-site client for production and iterated and updated every 7 days. The (Random Forest) algorithm predicts extrusion results based on information such as rubber compound, specifications, and environment, corrects and recommends optimal production parameters in real time, and automatically sends them to the machine with one click. (XGBoost, MLP) algorithms dynamically adjust the extrusion weight per meter and production line speed in real time based on extrusion current, pressure, temperature, and weight per meter. Temperature and weight per meter measured at 10 meters of extrusion are used by the AI ​​model to predict the weight and width of the cooled tread after 200 meters.

[0051] The rubber extrusion formulation of this invention is highly efficient, stable, and robust; it shortens the response time by 95%, increases the weight qualification rate from 90.10% to 95.90%, and increases the dimensional qualification rate from 91.86% to 97.14%. Attached Figure Description

[0052] Figure 1 This is a schematic diagram of the rubber extrusion formulation structure of the present invention.

[0053] Figure 2 This is a flowchart of the present invention.

[0054] Figure 3 This is a flowchart of the process for obtaining the optimal formula information of this invention.

[0055] Figure 4 This is a flowchart of the training process for the random forest module of this invention. Detailed Implementation

[0056] The present invention will now be described in further detail with reference to the accompanying drawings and embodiments.

[0057] Example 1

[0058] A method for intelligent formulation of rubber extrusion includes an intelligent control and management terminal, an intelligent terminal, and a machine; the method includes:

[0059] The intelligent control and management terminal uses the mean-shift clustering algorithm to cluster recent production data and calculate the recent optimal formula information for each specification;

[0060] The intelligent control and management terminal uses historical production data to regress the influence factors of different rubber compounds and different production line states on the formula through the random forest algorithm.

[0061] During production, the intelligent control and management terminal switches the rubber specifications through the PLC, obtains the optimal formula information from the data platform by specification clustering, and automatically sends it to the machine.

[0062] After a period of production, the smart terminal obtains information about the rubber material and production line, transmits it to the random forest model, obtains standard corrections, and sends them to the machine.

[0063] Recent production data includes rubber specifications, line speed, upper extruder speed, middle extruder speed, lower extruder speed, front shoulder width, front total width, front weight, rear shoulder width, rear total width, and rear weight.

[0064] Historical production data includes rubber specifications, line speed, upper extruder speed, middle extruder speed, lower extruder speed, front shoulder width, front total width, front weight, rear shoulder width, rear total width, rear weight, upper extruder pressure, middle extruder pressure, lower extruder pressure, upper extruder temperature, middle extruder temperature, lower extruder temperature, upper extruder current, middle extruder current, lower extruder current, Mooney compound, compound hardness, compound specific gravity, and compound sulfur grade.

[0065] Table 1 lists rubber specification AH and its corresponding recent production data.

[0066] Table 1

[0067] Production line number Specification Update time Product type Linkage linear velocity Upper extruder speed medium extruder speed Lower extruder speed Standard shoulder width Overall width standard Weight Standard 1 A 16 / 3 / 2024 02：25：06 1 10.23 17.32 15.29 8.4 213.4 253.3 6.93 1 B 3 / 12 / 2023 05：3840 1 12.41 15.36 11.4 8.55 192.4 271.03 4.23 1 C 20 / 8 / 2024 09：19：19 1 10.36 17.1 15.61 8.17 198.54 297.71 6.7 1 D 11 / 9 / 2024 08：0023 1 11.97 18.8 15.43 12.09 198.73 275.63 5.74 1 E 5 / 11 / 2024 03：1854 1 13.08 14.97 8.92 8.96 189.96 268.05 4.24 1 F 26 / 4 / 2024 08：59：04 1 12.47 16.96 11.82 9.15 213.4 253.3 4.25 1 G 22 / 9 / 2024 10：5320 1 11.99 16.93 14.8 10.59 197.09 289.51 5.57 1 H 31 / 5 / 2024 08：19：14 1 10.52 16.4 14.34 9.84 198.42 283.39 5.76

[0068] The mean-shift clustering algorithm, which clusters recent production data, includes the following implementations:

[0069] The intelligent control and management terminal collects recent production data and categorizes the acquired data.

[0070] Recent production data preprocessing is performed using a time-lag cross-correlation algorithm.

[0071] Initialize the processed recent production data by randomly selecting one data point as the starting center point x. 起始 ;

[0072] The bandwidth parameter h is estimated using the estimate_bandwidth function, thereby determining the updated center point position x. 更新， x 更新 =x 起始 +m(x);

[0073] m(x) is the drift vector.

[0074]

[0075] Among them, K h (u) is a Gaussian kernel function with bandwidth h.

[0076]

[0077] Kernel function satisfies

[0078] By obtaining the drift vector, the iteration stops when ||m(x)|| < ε or the number of iterations reaches the set maximum value T; otherwise, the above steps are returned to continue, thereby obtaining the optimal rubber formula based on recent production data.

[0079] Time-lag cross-correlation algorithms are used to preprocess recent production data, including

[0080] Let the initial weight sequence be X = {x1, x2, ..., xn} n}

[0081] The subsequent weight sequence is Y = {y1, y2, ..., y} n}, where n is the length of the average length;

[0082] Time lag cross-correlation function R XY (τ) represents the correlation between time series Y and X at a time lag τ.

[0083]

[0084] Where τ represents the time lag; Let X and Y be the means of the time series. The mean of the time series XY; x t The values ​​of time series X at time points τ and t+τ, y t+τ These are the values ​​of time series X and Y at time points τ and t+τ, respectively; τ max =argmax τ |R XY (τ)|

[0085] Where, τ max The amount of weight that needs to be measured in advance is the total width of the back shoulders.

[0086] The implementation of the random forest algorithm to regress the influence factors of different rubber compounds and different production line conditions on the formulation includes:

[0087] Historical production data acquisition: The intelligent control and management terminal collects historical production data.

[0088] Historical production data preprocessing is performed using a time-lag cross-correlation algorithm.

[0089] A random forest model was constructed using preprocessed historical production data.

[0090] The parameters of the random forest model are optimized, and the random forest model is trained using the optimized parameters.

[0091] The parameters of the random forest model are optimized, and the model is trained using the optimized parameters. Tree-based Parsons optimization is used to further optimize the random forest model. This can be achieved through methods including:

[0092] Choose an initial hyperparameter x and compute two probabilistic models L(x) and G(x);

[0093] l(x) represents the probability that the objective function value is lower than a certain threshold γ given the hyperparameter x; L(x) = P(x|y<γ)

[0094] G(x) represents the probability (i.e., the parameter space of the "difference") that the objective function value is higher than a certain threshold γ given the hyperparameter x: G(x) = P(x|y≥γ)

[0095] p(y|x) is the conditional probability density function of the objective function value y given the hyperparameters x:

[0096] P(y|x)=L(x)·L(y<γ)+G(x)·G(y≥γ)

[0097] EI(x) is used to select hyperparameters x for evaluation and to record the objective function value. EI(x) is used to measure the expected improvement of the objective function value relative to the current best value γ after selecting a set of hyperparameters x.

[0098]

[0099] Where, μ g and σ g Φ(...) and φ(...) are the mean and standard deviation of the objective function value under G(x), respectively, and Φ(...) and φ(...) are the cumulative distribution function and probability density function of the standard normal distribution, respectively.

[0100] Example 2

[0101] Based on Example 1, this example is a smart rubber extrusion formulation system, including a smart control and management terminal, a smart terminal, and a machine; it also includes,

[0102] The module for obtaining optimal formula information uses the mean-shift clustering algorithm to cluster recent production data and calculate the recent optimal formula information for each specification in the intelligent control and management terminal.

[0103] The Random Forest Regression Module uses historical production data in the intelligent control and management terminal to regress the influence factors of different rubber compounds and different production line states on the formulation through the Random Forest algorithm.

[0104] During production, the intelligent control and management terminal switches the rubber specifications through the PLC, obtains the optimal formula information from the data platform by specification clustering, and automatically sends it to the machine.

[0105] After a period of production, the smart terminal obtains information about the rubber material and production line, transmits it to the random forest model, obtains standard corrections, and sends them to the machine.

[0106] Figure 1 In this process, data from the MES, 3D cameras, and PLCs is transmitted to the process formula model for learning, then stored on the control platform. The control platform then issues commands to the semi-finished product AI intelligent terminal, which in turn issues commands to the production line. The AI ​​predictive control model, learned from the 3D camera and PLC data, is deployed to the semi-finished product AI intelligent terminal, controlling the production line and achieving automated control.

Claims

1. A method for intelligent formulation of rubber extrusion, comprising an intelligent control management terminal, an intelligent terminal, and a machine; The methods include: The intelligent control and management terminal uses the mean-shift clustering algorithm to cluster recent production data and calculate the recent optimal formula information for each specification; The intelligent control and management terminal uses historical production data to regress the influence factors of different rubber compounds and different production line states on the formula through the random forest algorithm. During production, the intelligent control and management terminal switches the rubber specifications through the PLC, obtains the optimal formula information from the data platform by specification clustering, and automatically sends it to the machine. After a period of production, the smart terminal obtains information about the rubber material and production line, transmits it to the random forest model, obtains standard corrections, and sends them to the machine. The parameters of the random forest model are optimized, and the model is trained using the optimized parameters. Tree-based Parsons optimization is used to further optimize the random forest model. This can be achieved through methods including: Calculate two probabilistic models by selecting an initial hyperparameter x. and ; l(x) is the probability that the objective function value is lower than a certain threshold γ given the hyperparameter x; Given hyperparameter x, the probability that the objective function value is higher than a certain threshold γ is: p(y|x) is the conditional probability density function of the objective function value y given the hyperparameters x: use We select the hyperparameter x for evaluation and record the objective function value. Used to measure the expected improvement of the objective function value relative to the current best value γ after selecting a set of hyperparameters x; in, and They are The mean and standard deviation of the objective function values ​​under the given conditions. and These are the cumulative distribution function and probability density function of the standard normal distribution, respectively.

2. The intelligent formulation method for rubber extrusion according to claim 1, characterized in that, Recent production data includes rubber specifications, line speed, upper extruder speed, middle extruder speed, lower extruder speed, front shoulder width, front total width, front weight, rear shoulder width, rear total width, and rear weight.

3. The intelligent formulation method for rubber extrusion according to claim 1, characterized in that, Historical production data includes rubber specifications, line speed, upper extruder speed, middle extruder speed, lower extruder speed, front shoulder width, front total width, front weight, rear shoulder width, rear total width, rear weight, upper extruder pressure, middle extruder pressure, lower extruder pressure, upper extruder temperature, middle extruder temperature, lower extruder temperature, upper extruder current, middle extruder current, lower extruder current, Mooney compound, compound hardness, compound specific gravity, and compound sulfur grade.

4. The intelligent formulation method for rubber extrusion according to claim 1, characterized in that, The mean-shift clustering algorithm, which clusters recent production data, includes the following implementations: The intelligent control and management terminal collects recent production data and categorizes the acquired data. Recent production data preprocessing is performed using a time-lag cross-correlation algorithm. Initialize the processed recent production data by randomly selecting one data point as the starting center point x. 起始 ; The bandwidth parameter h is estimated using the estimate_bandwidth function, thereby determining the updated center point position x. 更新， x 更新 =x 起始 +m(x); m(x) is the drift vector. ； in, Let h be the bandwidth of the Gaussian kernel function. ； Kernel function satisfies ; By obtaining the drift vector, when If the number of iterations reaches the set maximum value T, stop the iteration; otherwise, return to the above steps to continue, thereby obtaining the optimal rubber formula based on recent production data.

5. The intelligent formulation method for rubber extrusion according to claim 1, characterized in that, Time-lag cross-correlation algorithms are used to preprocess recent production data, including The previous weight sequence is set as ; The subsequent weight sequence is ,in, For length; Time lag cross-correlation function Representing time series and Time lag The correlation below, The time lag is the time delay. Time series For time series y For time series x at time points and The value, For the time series y at time points and The value; ; The amount of weight that needs to be measured in advance is the total width of the back shoulders.

6. The intelligent formulation method for rubber extrusion according to claim 1, characterized in that, The implementation of the random forest algorithm to regress the influence factors of different rubber compounds and different production line conditions on the formulation includes: Historical production data acquisition: The intelligent control and management terminal collects historical production data. Historical production data preprocessing is performed using a time-lag cross-correlation algorithm. A random forest model was constructed using preprocessed historical production data. The parameters of the random forest model are optimized, and the random forest model is trained using the optimized parameters.

7. A rubber extrusion intelligent formulation system, comprising an intelligent control and management terminal, an intelligent terminal, and a machine; characterized in that: It also includes, The module for obtaining optimal formula information uses the mean-shift clustering algorithm to cluster recent production data and calculate the recent optimal formula information for each specification in the intelligent control and management terminal. The Random Forest Regression Module uses historical production data in the intelligent control and management terminal to regress the influence factors of different rubber compounds and different production line states on the formulation through the Random Forest algorithm. During production, the intelligent control and management terminal switches the rubber specifications through the PLC, obtains the optimal formula information from the data platform by specification clustering, and automatically sends it to the machine. After a period of production, the intelligent terminal receives information about the rubber compound and production line, transmits it to the random forest model, obtains standard corrections, and sends them to the machines. The parameters of the random forest model are optimized, and the optimized parameters are used to train the random forest model. Tree-based Parsons optimization is then used to further optimize the random forest model. The optimization methods include: Calculate two probabilistic models by selecting an initial hyperparameter x. and ; l(x) is the probability that the objective function value is lower than a certain threshold γ given the hyperparameter x; Given hyperparameter x, the probability that the objective function value is higher than a certain threshold γ is: p(y|x) is the conditional probability density function of the objective function value y given the hyperparameters x: use We select the hyperparameter x for evaluation and record the objective function value. Used to measure the expected improvement of the objective function value relative to the current best value γ after selecting a set of hyperparameters x; in, and They are The mean and standard deviation of the objective function values ​​under the given conditions. and These are the cumulative distribution function and probability density function of the standard normal distribution, respectively.

Images