Artificial intelligence-based optimization method and system for rubber melting and mixing

Abstract

The present invention relates to an artificial intelligence-based rubber melting and mixing optimization method and system, comprising the following steps: S1: arranging temperature, pressure, and viscosity sensors during the rubber melting process to collect equipment operating status and process parameters in real time; S2: storing the collected data in a database, using data cleaning technology to remove noise and outliers to ensure data quality; S3: constructing a rubber melting and mixing prediction model based on an improved BP neural network model to predict the properties of the mixed rubber; S4 constructing a comprehensive optimization objective function based on cost, energy consumption, and product characteristics, and based on the prediction results, using particle swarm optimization to find the optimal process parameter combination; S5: controlling the system based on the optimal process parameter combination in combination with fuzzy logic control. The present invention can not only improve production efficiency and product quality, but also help reduce costs and waste, and improve the overall flexibility and responsiveness of the production line.

Description

Technical Field

[0001] The present invention relates to the field of intelligent production, and in particular to an artificial intelligence-based rubber material melting and mixing optimization method and system. Background Art

[0002] Traditional methods of controlling the rubber melting process usually rely on the operator's experience and relatively simple parameter adjustments. Although this method has practical value in some cases, it often seems powerless when faced with increasingly complex nonlinear dynamic systems. The characteristic of nonlinear dynamic systems is that the relationship between input and output is not direct or linear, but changes with time, conditions and other external factors. In addition, under the modern manufacturing profile, multi-objective optimization requires considering multiple (sometimes conflicting) goals, such as reducing energy consumption, improving quality, shortening production time, etc., which increases the challenges of traditional methods. Summary of the invention

[0003] In order to solve the above problems, the purpose of the present invention is to provide an artificial intelligence-based rubber melting and mixing optimization method and system, which can not only improve production efficiency and product quality, but also help reduce costs and waste, and improve the overall flexibility and responsiveness of the production line.

[0004] To achieve the above object, the present invention adopts the following technical solutions:

[0005] An artificial intelligence-based optimization method for rubber melting and mixing comprises the following steps:

[0006] S1: Arrange temperature, pressure and viscosity sensors during the rubber melting process to collect equipment operation status and process parameters in real time;

[0007] S2: Store the collected data in a database and use data cleaning techniques to remove noise and outliers to ensure data quality;

[0008] S3: Building a rubber compound melting and mixing prediction model based on the improved BP neural network model to predict the properties of the mixed rubber compound;

[0009] S4: Combine cost, energy consumption and product characteristics to build a comprehensive optimization objective function, and based on the prediction results, use particle swarm optimization to find the optimal combination of process parameters;

[0010] S5: According to the optimal combination of process parameters, the system is controlled in combination with fuzzy logic control, as follows:

[0011] Receive data from sensors and calculate the error E and error change rate ΔE of each current control parameter based on the optimal process parameter combination;

[0012] E and ΔE are converted into corresponding fuzzy values ​​through membership functions respectively;

[0013] Use Mamdani inference engine to analyze all rules and calculate fuzzy output;

[0014] The fuzzy output is defuzzified into a specific control signal through the centroid method, and then the actuator is adjusted to change the behavior of the system;

[0015] Detect the output effect in real time, and further adjust the input error and error change rate based on real-time feedback to continuously iterate and optimize the control effect.

[0016] Furthermore, S1 is specifically as follows: the temperature sensor uses a thermistor sensor, which is installed at the junction of the smelting furnace and the conveying pipeline and at the inlet and outlet of the molten material; the pressure sensor uses a piezoresistive sensor, which is installed at the junction of the smelting furnace and the conveying pipeline and at the inlet and outlet of the molten material; the viscosity sensor uses a rotational viscometer, which is installed in the stirring area to measure the viscosity change of the material in the molten state online; and a distributed layout is adopted to connect the sensors through an industrial bus to ensure stable data transmission.

[0017] Furthermore, S2 is specifically:

[0018] According to historical data and process requirements, set the threshold range of each sensor and filter out the out-of-range values; verify the logical relationship between the data, including the correlation between temperature and pressure, identify unreasonable data combinations, and calculate the correlation coefficient r between temperature and pressure:

[0019] ;

[0020] in, and represent the i'th temperature and pressure data point respectively; and are the average values ​​of temperature and pressure data, respectively;

[0021] If the |r| value is significantly lower than the preset value, there is an anomaly; it is marked as an unreasonable data combination;

[0022] Use Python to write automated scripts combined with the Pandas library to perform data cleaning, configure the rule-based engine Apache NiFi, monitor and operate data cleaning in the pipeline; and use interpolation to fill missing data for missing values.

[0023] Furthermore, a rubber compound melting and mixing prediction model is constructed based on the improved BP neural network model, as follows:

[0024] The improved BP neural network model includes an input layer, a hidden layer and an output layer, wherein

[0025] Input layer: Input features include parameters during the rubber melting process, including temperature, pressure, and viscosity;

[0026] Hidden layer: Use a multi-layer hidden structure:

[0027] Hidden layer 1: 32 neurons, Leaky ReLU activation;

[0028] Second hidden layer: 16 neurons, Leaky ReLU activation;

[0029] For each layer of neurons, the forward propagation process is as follows:

[0030] Hidden layer calculation:

[0031] ;

[0032] Among them, l represents the layer number, w ji is the weight from the previous layer neuron i to the current neuron j, b j is bias; is the weighted input of the jth neuron in layer l; is the output of neuron i in layer l-1; n is the number of neurons;

[0033] The Leaky ReLU activation is as follows:

[0034] ;

[0035] in, is the preset parameter;

[0036] Output layer: Use linear activation function to predict the continuous properties of the rubber compound;

[0037] The mean square error (MSE) is used as the loss function, and the L2 regularization term is introduced into the loss function to control the weight size, thereby suppressing overfitting, and is used to measure the difference between the predicted value and the actual value. The derivative of the loss function with respect to each weight and bias is calculated through the back propagation algorithm to adjust these parameters, and the adaptive learning rate optimizer Adam is used in combination to improve the convergence speed and stability.

[0038] Furthermore, the mean square error (MSE) is used as the loss function, and the L2 regularization term is introduced into the loss function to control the weight size, thereby suppressing overfitting, and is used to measure the difference between the predicted value and the actual value, as follows:

[0039] Mean square error (MSE) is used to measure the deviation between the predicted value and the actual value:

[0040] ;

[0041] Where m is the number of samples, It is The predicted value of samples, It is The actual value of the samples

[0042] The L2 regularization term is used to control the weight size and reduce the risk of overfitting:

[0043] ;

[0044] Where λ is the regularization strength parameter, l is the layer number, j and i are the neuron indices in the current layer and the previous layer respectively; w ji is the weight from the previous layer neuron i to the current layer neuron j;

[0045] Combine the MSE and L2 regularization terms to construct the final loss function:

[0046] .

[0047] Further, back propagation:

[0048] For each output unit, calculate the gradient:

[0049] ;

[0050] Where L is the total number of hidden layers; represents partial derivative operation;

[0051] Using the error of the output layer, reversely calculate the error of the previous layer, and for each hidden layer, backpropagate the error:

[0052] ;

[0053] in, is the derivative of the activation function; ⊙ represents the Hadamard product, is the weight of the l+1th hidden layer; is the back propagation error of the lth layer;

[0054] Compute the gradients of weights and biases:

[0055] ;

[0056] .

[0057] Furthermore, S4 is specifically:

[0058] Combining cost, energy consumption and product characteristics for weighted comprehensive optimization, the objective function J is expressed as:

[0059] ;

[0060] Among them, w 1 ,w 2 ,w 3 is the weight of each target, Cost(x), Energy(x), Quality(x) are different target values ​​evaluated by BP model;

[0061] A particle swarm is randomly generated, each particle represents a candidate solution for a process parameter combination, and the speed and position of each particle are initialized;

[0062] Update the particle's velocity and position to explore the global optimal solution;

[0063] ;

[0064] ;

[0065] Where w is the inertia factor, c 1 ,c 2 is the learning factor, r 1 , r 2 is a random number; is the velocity of particle a at the t+1th iteration; The position of particle a at iteration t; is the best historical position of particle a at the tth iteration; g(t) is the position of the global optimal solution;

[0066] The target value of each particle under the process parameter combination is evaluated through the trained BP neural network; the objective function value J is calculated to evaluate the fitness.

[0067] If the current state of the particle can improve the adaptability, the personal optimum is updated; if the global optimum can be replaced by the current particle, the global optimum is updated;

[0068] The speed, position update and fitness evaluation process are repeated until the maximum number of iterations is reached or the stopping condition is met, and the optimal process parameter combination is selected as the final optimization result.

[0069] Furthermore, the Mamdani inference engine is used to analyze all the rules and calculate the fuzzy output, as follows:

[0070] Preset fuzzy sets and membership functions. The fuzzy sets include negative large, negative medium, zero, positive medium, and positive large. Fuzzification is performed on the input error E and the error change rate ΔE.

[0071] For each fuzzy rule, calculate its activation , that is, select the minimum membership of all antecedents;

[0072] ;

[0073] in, and are the membership function values ​​of error E and error change rate ΔE respectively;

[0074] For each rule output, perform the composite calculation:

[0075] The influence of each rule on the output is represented by a fuzzy set, which is obtained by intercepting The fuzzy output set is used to represent the contribution of the rule to the output. Each rule constructs a truncated fuzzy set. is the cut-off height;

[0076] Combine the output fuzzy sets of all rules and use the maximization operation to complete the combination:

[0077]

[0078] in, represents the fuzzy set output by the g-th rule; is the aggregated fuzzy set after combination, that is, the final comprehensive output.

[0079] A rubber melting and mixing optimization system based on artificial intelligence includes a processor, a memory and a computer program stored in the memory. When the processor executes the computer program, it specifically executes the steps in the rubber melting and mixing optimization method based on artificial intelligence as described above.

[0080] The present invention has the following beneficial effects:

[0081] 1. The present invention can not only improve production efficiency and product quality, but also help reduce costs and waste, and improve the overall flexibility and responsiveness of the production line;

[0082] 2. The present invention uses an improved BP neural network to predict the properties of rubber materials, which can better capture complex nonlinear relationships, improve prediction accuracy and adaptability, and based on multi-objective functions, adopts a particle swarm optimization algorithm to simultaneously consider multiple process objectives and intelligently search for the optimal process parameter combination;

[0083] 3. The present invention combines fuzzy logic control to handle complex system uncertainties. It can flexibly combine system responses in different states through a rule base to achieve intelligent and efficient control. It also uses Mamdani reasoning and the centroid method to effectively convert fuzzy control quantities into precise control signals, ensuring that the actuator can be controlled accurately and flexibly, providing fine-grained response adjustment for the system. BRIEF DESCRIPTION OF THE DRAWINGS

[0084] Figure 1 The figure is a flow chart of the method of the present invention. DETAILED DESCRIPTION

[0085] The present invention is further described in detail below with reference to the accompanying drawings and specific embodiments:

[0086] refer to Figure 1 In this embodiment, a method for optimizing the melting and mixing of rubber materials based on artificial intelligence is provided, comprising the following steps:

[0087] S1: Arrange temperature, pressure and viscosity sensors during the rubber melting process to collect equipment operation status and process parameters in real time;

[0088] S2: Store the collected data in a database and use data cleaning techniques to remove noise and outliers to ensure data quality;

[0089] S3: Building a rubber compound melting and mixing prediction model based on the improved BP neural network model to predict the properties of the mixed rubber compound;

[0090] S4: Combine cost, energy consumption and product characteristics to build a comprehensive optimization objective function, and based on the prediction results, use particle swarm optimization to find the optimal combination of process parameters;

[0091] S5: According to the optimal combination of process parameters, the system is controlled in combination with fuzzy logic control, as follows:

[0092] Receive data from sensors and calculate the error E and error change rate ΔE of each current control parameter based on the optimal process parameter combination;

[0093] E and ΔE are converted into corresponding fuzzy values ​​through membership functions respectively;

[0094] Use Mamdani inference engine to analyze all rules and calculate fuzzy output;

[0095] The fuzzy output is defuzzified into a specific control signal through the centroid method, and then the actuator is adjusted to change the behavior of the system;

[0096] Detect the output effect in real time, and further adjust the input error and error change rate based on real-time feedback to continuously iterate and optimize the control effect.

[0097] In this embodiment, S1 is specifically: the temperature sensor adopts a thermistor sensor, which is installed at the junction of the smelting furnace and the conveying pipeline and at the inlet and outlet of the molten material; the pressure sensor adopts a piezoresistive sensor, which is installed at the junction of the smelting furnace and the conveying pipeline and at the inlet and outlet of the molten material; the viscosity sensor adopts a rotational viscometer, which is installed in the stirring area to measure the viscosity change of the material in the molten state online; and a distributed layout is adopted to connect the sensors through an industrial bus (CAN, Modbus) to ensure stable data transmission.

[0098] In this embodiment, S2 is specifically:

[0099] According to historical data and process requirements, set the threshold range of each sensor and filter out the out-of-range values; verify the logical relationship between the data, including the correlation between temperature and pressure, identify unreasonable data combinations, and calculate the correlation coefficient r between temperature and pressure:

[0100] ;

[0101] in, and represent the i'th temperature and pressure data point respectively; and are the average values ​​of temperature and pressure data, respectively;

[0102] If the |r| value is significantly lower than the preset value, there is an anomaly; it is marked as an unreasonable data combination;

[0103] Use Python to write automated scripts combined with the Pandas library to perform data cleaning, configure the rule-based engine Apache NiFi, monitor and operate data cleaning in the pipeline; and use interpolation to fill missing data for missing values.

[0104] In this embodiment, a rubber material melting and mixing prediction model is constructed based on an improved BP neural network model, as follows:

[0105] The improved BP neural network model includes an input layer, a hidden layer and an output layer, wherein

[0106] Input layer: Input features include parameters during the rubber melting process, including temperature, pressure, and viscosity;

[0107] Hidden layer: Use a multi-layer hidden structure:

[0108] Hidden layer 1: 32 neurons, Leaky ReLU activation;

[0109] Second hidden layer: 16 neurons, Leaky ReLU activation;

[0110] For each layer of neurons, the forward propagation process is as follows:

[0111] Hidden layer calculation:

[0112] ;

[0113] Among them, l represents the layer number, w ji is the weight from the previous layer neuron i to the current neuron j, b j is bias; is the weighted input of the jth neuron in layer l; is the output of neuron i in layer l-1; n is the number of neurons;

[0114] The Leaky ReLU activation is as follows:

[0115] ;

[0116] in, is the preset parameter;

[0117] Output layer: Use linear activation function to predict the continuous properties of the rubber compound;

[0118] The mean square error (MSE) is used as the loss function, and the L2 regularization term is introduced into the loss function to control the weight size, thereby suppressing overfitting, and is used to measure the difference between the predicted value and the actual value. The derivative of the loss function with respect to each weight and bias is calculated through the back propagation algorithm to adjust these parameters, and the adaptive learning rate optimizer Adam is used in combination to improve the convergence speed and stability.

[0119] In this embodiment, the mean square error MSE is used as the loss function, and the L2 regularization term is introduced into the loss function to control the weight size, thereby suppressing overfitting, and is used to measure the difference between the predicted value and the actual value, as follows:

[0120] Mean square error (MSE) is used to measure the deviation between the predicted value and the actual value:

[0121] ;

[0122] Where m is the number of samples, It is The predicted value of samples, It is The actual value of samples;

[0123] The L2 regularization term is used to control the weight size and reduce the risk of overfitting:

[0124] ;

[0125] Where λ is the regularization strength parameter, l is the layer number, j and i are the neuron indices in the current layer and the previous layer respectively; w ji is the weight from the previous layer neuron i to the current layer neuron j;

[0126] Combine the MSE and L2 regularization terms to construct the final loss function:

[0127] .

[0128] In this example, back propagation:

[0129] For each output unit, calculate the gradient:

[0130] ;

[0131] Where L is the total number of hidden layers; represents partial derivative operation;

[0132] Using the error of the output layer, reversely calculate the error of the previous layer, and for each hidden layer, backpropagate the error:

[0133] ;

[0134] in, is the derivative of the activation function; ⊙ represents the Hadamard product, is the weight of the l+1th hidden layer; is the back propagation error of the lth layer;

[0135] Compute the gradients of weights and biases:

[0136] ;

[0137] .

[0138] In this embodiment, S4 is specifically:

[0139] Combining cost, energy consumption and product characteristics for weighted comprehensive optimization, the objective function J is expressed as:

[0140] ;

[0141] Among them, w 1 ,w 2 ,w 3 is the weight of each target, Cost(x), Energy(x), Quality(x) are different target values ​​evaluated by BP model;

[0142] A particle swarm is randomly generated, each particle represents a candidate solution for a process parameter combination, and the speed and position of each particle are initialized;

[0143] Update the particle's velocity and position to explore the global optimal solution;

[0144] ;

[0145] ;

[0146] Where w is the inertia factor, c 1 ,c 2 is the learning factor, r 1 , r 2 is a random number; is the velocity of particle a at the t+1th iteration; The position of particle a at iteration t; is the best historical position of particle a at the tth iteration; g(t) is the position of the global optimal solution;

[0147] The target value of each particle under the process parameter combination is evaluated through the trained BP neural network; the objective function value J is calculated to evaluate the fitness.

[0148] If the current state of the particle can improve the adaptability, the personal optimum is updated; if the global optimum can be replaced by the current particle, the global optimum is updated;

[0149] The speed, position update and fitness evaluation process are repeated until the maximum number of iterations is reached or the stopping condition is met, and the optimal process parameter combination is selected as the final optimization result.

[0150] In this embodiment, the Mamdani inference engine is used to analyze all rules and calculate the fuzzy output, as follows:

[0151] Preset fuzzy sets and membership functions. The fuzzy sets include negative large, negative medium, zero, positive medium, and positive large. Fuzzification is performed on the input error E and the error change rate ΔE.

[0152] For each fuzzy rule, calculate its activation , that is, select the minimum membership of all antecedents:

[0153] ;

[0154] in, and are the membership function values ​​of error E and error change rate ΔE respectively;

[0155] For each rule output, perform the composite calculation:

[0156] The influence of each rule on the output is represented by a fuzzy set, which is obtained by intercepting The fuzzy output set is used to represent the contribution of the rule to the output. Each rule constructs a truncated fuzzy set. is the cut-off height;

[0157] Combine the output fuzzy sets of all rules and use the maximization operation to complete the combination:

[0158]

[0159] in, represents the fuzzy set output by the g-th rule; is the aggregated fuzzy set after combination, that is, the final comprehensive output.

[0160] It will be appreciated by those skilled in the art that embodiments of the present invention may be provided as methods, systems, or computer program products. Therefore, the present invention may take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware. Furthermore, the present invention may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0161] The present invention is described with reference to flowcharts and / or block diagrams of methods, devices (systems), and computer program products according to embodiments of the present invention. It should be understood that each process and / or block in the flowchart and / or block diagram, as well as the combination of processes and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0162] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0163] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process in the computer or other programmable device. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.

[0164] The above is only a preferred embodiment of the present invention, and does not limit the present invention in other forms. Any technician familiar with the profession may use the above disclosed technical content to change or modify it into an equivalent embodiment with equivalent changes. However, any simple modification, equivalent change and modification made to the above embodiment according to the technical essence of the present invention without departing from the technical solution of the present invention still belongs to the protection scope of the technical solution of the present invention.

Claims

1. The optimization method of rubber melting and mixing based on artificial intelligence is characterized by: The following steps are involved: S1: Arrange temperature, pressure and viscosity sensors during the rubber melting process to collect equipment operation status and process parameters in real time; S2: Store the collected data in a database and use data cleaning techniques to remove noise and outliers to ensure data quality; S3: Building a rubber compound melting and mixing prediction model based on the improved BP neural network model to predict the properties of the mixed rubber compound; S4: Combine cost, energy consumption and product characteristics to build a comprehensive optimization objective function, and based on the prediction results, use particle swarm optimization to find the optimal combination of process parameters; S5: According to the optimal combination of process parameters, the system is controlled in combination with fuzzy logic control, as follows: Receive data from sensors and calculate the error E and error change rate ΔE of each current control parameter based on the optimal process parameter combination; E and ΔE are converted into corresponding fuzzy values ​​through membership functions respectively; Use Mamdani inference engine to analyze all rules and calculate fuzzy output; The fuzzy output is defuzzified into a specific control signal through the centroid method, and then the actuator is adjusted to change the behavior of the system; Detect output effects in real time, and further adjust input error and error change rate based on real-time feedback, continuously iterating and optimizing control effects; The S4 is specifically: Combining cost, energy consumption and product characteristics for weighted comprehensive optimization, the objective function J is expressed as: ; Among them, w1, w2, w3 are the weights of each target, Cost(x), Energy(x), Quality(x) are different target values ​​evaluated by BP model; A particle swarm is randomly generated, each particle represents a candidate solution for a process parameter combination, and the speed and position of each particle are initialized; Update the particle's velocity and position to explore the global optimal solution; ； ； Among them, w is the inertia factor, c1, c2 are learning factors, r1, r2 are random numbers; is the velocity of particle a at the t+1th iteration; The position of particle a at iteration t; is the best historical position of particle a at the tth iteration; g(t) is the position of the global optimal solution; The target value of each particle under the process parameter combination is evaluated through the trained BP neural network; the objective function value J is calculated to evaluate the fitness; If the current state of the particle can improve the adaptability, the personal optimum is updated; if the global optimum can be replaced by the current particle, the global optimum is updated; Repeat the speed, position update and fitness evaluation process until the maximum number of iterations is reached or the stop condition is met, and select the optimal process parameter combination as the final optimization result; All rules are analyzed using the Mamdani inference engine and the fuzzy output is calculated as follows: Preset fuzzy sets and membership functions. The fuzzy sets include negative large, negative medium, zero, positive medium, and positive large. Fuzzification is performed on the input error E and the error change rate ΔE. For each fuzzy rule, calculate its activation , that is, select the minimum membership of all antecedents: ； in, and are the membership function values ​​of error E and error change rate ΔE respectively; For each rule output, perform the composite calculation: The influence of each rule on the output is represented by a fuzzy set, which is obtained by intercepting The fuzzy output set is used to represent the contribution of the rule to the output. Each rule constructs a truncated fuzzy set. is the cut-off height; Combine the output fuzzy sets of all rules and use the maximization operation to complete the combination: in, represents the fuzzy set output by the g-th rule; is the aggregated fuzzy set after combination, that is, the final comprehensive output.

2. The artificial intelligence-based optimization method for rubber melting and mixing according to claim 1 is characterized in that: Specifically, S1 is as follows: the temperature sensor is a thermistor sensor, which is installed at the junction of the smelting furnace and the conveying pipeline and at the inlet and outlet of the molten material; the pressure sensor is a piezoresistive sensor, which is installed at the junction of the smelting furnace and the conveying pipeline and at the inlet and outlet of the molten material; the viscosity sensor is a rotational viscometer, which is installed in the stirring area to measure the viscosity change of the material in the molten state online; and a distributed layout is adopted to connect the sensors through an industrial bus to ensure stable data transmission.

3. The artificial intelligence-based optimization method for rubber melting and mixing according to claim 1 is characterized in that: The S2 is specifically: According to historical data and process requirements, set the threshold range of each sensor and filter out the out-of-range values; verify the logical relationship between the data, including the correlation between temperature and pressure, identify unreasonable data combinations, and calculate the correlation coefficient r between temperature and pressure: ； in, and represent the i'th temperature and pressure data point respectively; and are the average values ​​of temperature and pressure data, respectively; If the |r| value is significantly lower than the preset value, there is an anomaly; it is marked as an unreasonable data combination; Use Python to write automated scripts combined with the Pandas library to perform data cleaning, configure the rule-based engine Apache NiFi, monitor and operate data cleaning in the pipeline; and use interpolation to fill missing values.

4. The artificial intelligence-based optimization method for rubber melting and mixing according to claim 1 is characterized in that: The rubber material melting and mixing prediction model is constructed based on the improved BP neural network model, as follows: The improved BP neural network model includes an input layer, a hidden layer and an output layer, wherein Input layer: Input features include parameters during the rubber melting process, including temperature, pressure, and viscosity; Hidden layer: Use a multi-layer hidden structure: Hidden layer 1: 32 neurons, Leaky ReLU activation; Second hidden layer: 16 neurons, Leaky ReLU activation; For each layer of neurons, the forward propagation process is as follows: Hidden layer calculation: ; Among them, l represents the layer number, w ji is the weight from the previous layer neuron i to the current neuron j, b j is bias; is the weighted input of the jth neuron in layer l; is the output of neuron i in layer l-1; n is the number of neurons; The Leaky ReLU activation is as follows: ； in, is the preset parameter; Output layer: Use linear activation function to predict the continuous properties of the rubber compound; The mean square error (MSE) is used as the loss function, and the L2 regularization term is introduced into the loss function to control the weight size, thereby suppressing overfitting, and is used to measure the difference between the predicted value and the actual value. The derivative of the loss function with respect to each weight and bias is calculated through the back propagation algorithm to adjust these parameters, and the adaptive learning rate optimizer Adam is used in combination to improve the convergence speed and stability.

5. The artificial intelligence-based optimization method for rubber melting and mixing according to claim 4 is characterized in that: The mean square error MSE is used as the loss function, and the L2 regularization term is introduced into the loss function to control the weight size, thereby suppressing overfitting, and is used to measure the difference between the predicted value and the actual value, as follows: Mean square error (MSE) is used to measure the deviation between the predicted value and the actual value: ； Where m is the number of samples, It is The predicted value of samples, It is The actual value of samples; The L2 regularization term is used to control the weight size and reduce the risk of overfitting: ； Where λ is the regularization strength parameter, l is the layer number, j and i are the neuron indices in the current layer and the previous layer respectively; w ji is the weight from the previous layer neuron i to the current layer neuron j; Combine the MSE and L2 regularization terms to construct the final loss function: 。 6. The artificial intelligence-based optimization method for rubber melting and mixing according to claim 5 is characterized in that: The back propagation: For each output unit, calculate the gradient: ； Where L is the total number of hidden layers; represents partial derivative operation; Using the error of the output layer, reversely calculate the error of the previous layer, and for each hidden layer, backpropagate the error: ； in, is the derivative of the activation function; ⊙ represents the Hadamard product; is the weight of the l+1th hidden layer; is the back propagation error of the lth layer; Compute the gradients of weights and biases: ； 。 7. An artificial intelligence-based rubber melting and mixing optimization system, characterized in that: It comprises a processor, a memory and a computer program stored in the memory. When the processor executes the computer program, it specifically performs the steps in the artificial intelligence-based rubber melting and mixing optimization method as described in any one of claims 1 to 6.

Images