A fuzzy control-based intelligent temperature regulation method for rubber conveyor belt mixing

Abstract

The application discloses a kind of rubber conveyor belt mixed temperature intelligent regulation and control method based on fuzzy control, comprising: S1, environmental heat distribution and conveyor belt operating state data are collected;S2, construct space-time alignment environment prediction window, calculate expected heat loss gradient;S3, input Swin Transformer V2 network inversion real-time dynamic heat inertia and heat content change rate, extract real heat storage state characteristic vector;S4, input adaptive variable domain fuzzy controller, introduce reverse learning strategy to select optimal individual, calculate inertial mass and gravitational constant;S5, update population position vector to global optimization of expansion factor, adjust fuzzy domain and output refined control instruction;S6, drive key node execution unit collaborative work and real-time display.The present application realizes the deep fusion of multi-source data, greatly improves the temperature control precision and response speed, effectively reduces energy consumption and prolongs the service life of equipment.

Description

Technical Field

[0001] This invention relates to the field of industrial automation control technology, and in particular to a method for intelligent control of mixing temperature of rubber conveyor belts based on fuzzy control. Background Technology

[0002] Adaptive variable universe of discourse fuzzy control, due to its ability to handle nonlinear, strongly coupled, and uncertain systems, has been widely applied in recent years in fields such as industrial automatic control, energy management, and thermal process control, becoming an important development direction for achieving precise control in complex environments. However, in practical applications, the temperature control scenario of rubber conveyor belts faces many challenges, such as complex environmental thermal field distribution, dynamic changes in conveying status, and fuzzy interference across multiple temperature zones. The control accuracy of adaptive variable universe of discourse fuzzy control is still constrained by many factors.

[0003] Currently, most conveyor belt temperature control methods rely on single-mode characteristic inputs, making it difficult to fully utilize multi-source information such as environmental meteorological data, operating tension, and thermal conductivity coefficients. This results in a lack of comprehensiveness in modeling the heat storage state of the conveyor belt. Some systems only use control rules with fixed parameters, ignoring the combined influence of multi-dimensional factors such as thermal inertia, enthalpy changes, and the correlation of local hot spots, thus limiting the adaptive adjustment capability of the control strategy for time-varying lagging systems. At the same time, traditional control logic processes lack effective extraction of complex spatiotemporal characteristics, making it difficult to accurately establish the nonlinear correlation between the local and overall temperature fields of long-span conveyor belts, affecting the accuracy and timeliness of control decisions.

[0004] Furthermore, most existing intelligent optimization algorithms lack an effective balance of search dynamics during the optimization process and fail to flexibly adjust the relationship between global exploration and local development according to the iteration process. This results in the algorithms being prone to premature convergence or low computational efficiency when facing complex working conditions, making it difficult to adapt to the rapid dynamic changes when long-distance conveyor belts cross different temperature zones, which seriously affects the temperature balance effect and operational stability of the system in real-world scenarios.

[0005] Therefore, how to provide a method for intelligent control of mixing temperature of rubber conveyor belts based on fuzzy control is a problem that urgently needs to be solved by those skilled in the art. Summary of the Invention

[0006] One objective of this invention is to propose an intelligent temperature control method for rubber conveyor belts based on fuzzy control. This invention fully integrates key steps such as multi-source spatiotemporal feature construction, deep feature mining of the Swin Transformer V2 network, optimization of the universe of discourse using an improved gravity search algorithm, and control command execution. It constructs an intelligent control process with multi-source information alignment, nonlinear correlation extraction of the conveyor belt temperature field, adaptive variable universe of discourse fuzzy control, and multi-execution unit collaboration, achieving dynamic balance and precise control of conveyor belt temperature in complex temperature range environments. This invention optimizes the scaling factor using an improved gravity search algorithm with a back-learning strategy and utilizes the Swin Transformer V2 network to invert thermal inertia and enthalpy change rate. It possesses advantages such as deep state feature extraction, high accuracy in control parameter optimization, and adaptive adjustment using fuzzy rules, significantly improving the system's dynamic response speed and temperature control stability. This effectively solves the problems of single thermal state modeling, large lag, and the tendency of control strategies to fall into local optima in existing methods.

[0007] A method for intelligent control of mixing temperature of a rubber conveyor belt based on fuzzy control according to an embodiment of the present invention includes the following steps: S1. Synchronously collect environmental heat distribution data and rubber conveyor belt operating status data for the entire rubber conveyor belt route; S2. Construct an environment prediction window based on spatiotemporal alignment. Based on the current running speed and position coordinates of the rubber conveyor belt, map the environmental heat distribution data onto the future movement trajectory of the rubber conveyor belt, and calculate the expected heat loss gradient of the target area to be entered within a preset time period in the future. S3. Based on the operating status data of the rubber conveyor belt, construct a spatiotemporal feature map and input it into the Swin Transformer V2 network. Utilize hierarchical shift window and global self-attention mechanism to mine the nonlinear correlation of the temperature field, invert and calculate the real-time dynamic thermal inertia and enthalpy change rate, and fuse them with the temperature deviation to extract and output the feature vector of the real heat storage state. S4. Input the actual heat storage state feature vector and the expected heat loss gradient into the adaptive variable universe fuzzy controller based on the improved gravity search algorithm, and map them to the initial fuzzy universe. Call the Gaussian membership function to calculate the membership degree, initialize the population, introduce a reverse learning strategy to select the optimal individual, and calculate and output the inertial mass and gravitational constant. S5. Based on the gravitational constant and inertial mass, update the population position vector and perform global optimization on the scaling factor to obtain the global optimal position vector. Input and adjust the fuzzy universe of discourse, use fuzzy inference to decouple the mixed interference, and directly map the calculation results into a refined control instruction set. S6. Based on the refined control instruction set, drive the heating execution unit, cooling execution unit and frequency conversion drive unit distributed at key nodes of the rubber conveyor belt to work together and display the results in real time on the control console.

[0008] Optionally, S1 specifically includes: S11. Synchronously trigger the distributed fiber optic sensors and infrared thermal imagers deployed along the entire rubber conveyor belt to collect environmental heat distribution data along the entire rubber conveyor belt. The environmental heat distribution data includes the surface temperature field distribution and the meteorological environment temperature of different areas along the route. S12. Read the data from the tension sensor deployed on the rubber conveyor belt to obtain the load tension of the rubber conveyor belt, read the feedback signal from the frequency converter to obtain the real-time running speed of the rubber conveyor belt, read the preset physical property database of rubber material to obtain the real-time thermal conductivity coefficient of the rubber material, and combine them to generate the running status data of the rubber conveyor belt.

[0009] Optionally, S2 specifically includes: S21. Calculate the displacement of the rubber conveyor belt in a future preset time period. Calculate the target position coordinates of the rubber conveyor belt in the future preset time period by using the sum of the current real-time position coordinates of the rubber conveyor belt and the displacement. Based on the target position coordinates, retrieve and obtain the meteorological environmental temperature of different areas along the route corresponding to the target position coordinates from the environmental heat distribution data. S23. Calculate the difference between the meteorological ambient temperature at the target location coordinates and the current surface temperature field distribution. Divide the difference by the displacement to calculate the expected heat loss gradient of the target area that the rubber conveyor belt will enter in the future preset time period.

[0010] Optionally, S3 specifically includes: S31. Align and stitch together the real-time running speed, load tension, and real-time thermal conductivity coefficient of the rubber conveyor belt in the rubber conveyor belt running status data according to the timestamp and the surface temperature field distribution of the rubber conveyor belt to construct a spatiotemporal feature map that integrates multi-source information. S32. Input the spatiotemporal feature map that integrates multi-source information into the PatchPartition layer of the Swin Transformer V2 network, divide the spatiotemporal feature map into several non-overlapping image blocks according to a preset size, flatten each image block into a one-dimensional vector, and linearly map the one-dimensional vector to a preset dimension to obtain the feature sequence. S33. Input the feature sequence into the Swing Transformer Block of the Swing Transformer V2 network, and use the hierarchical shift window mechanism to divide the feature sequence into multiple local windows according to the preset size. Calculate the query vector, key vector and value vector corresponding to all pixels in each local window. S34. Multiply the query vector by the transpose of the key vector and divide by the square root of the total number of pixels to obtain the attention score. Input the attention score into the Softmax function to normalize and obtain the weight coefficients, i.e., the multi-head self-attention weights. Multiply the weight coefficients by the value vector, sum all the product results within the local window, and update the feature vector within the local window. S35. Using a hierarchical shift window mechanism, perform a cyclic shift operation on the feature vectors within the updated local window, connect and merge the adjacent local windows after shifting to form an extended window, calculate the new query vector, key vector, and value vector on the merged feature vectors within the extended window, multiply the new query vector by the transpose of the new key vector and divide by the square root of the total number of pixels to obtain the extended attention score. S36. Input the extended attention score into the Softmax function to normalize and obtain the recalculated multi-head self-attention weights. Multiply the feature vector within the extended window with the recalculated multi-head self-attention weights to generate a feature vector that has extracted non-linear correlations. S37. Input the feature vector with extracted nonlinear correlation into the PatchMerging layer of the Swin Transformer V2 network, perform a concatenation operation between four adjacent feature blocks to form a concatenated feature block, multiply the concatenated feature block by a weight matrix of a preset size, perform dimensionality reduction processing, and construct high-level semantic features. S38. Input the high-level semantic features into the global self-attention layer of the Swin Transformer V2 network, calculate the correlation score matrix between any two high-level semantic features, multiply the high-level semantic features with the correlation score matrix to obtain the weighted global feature vector, and input it into the preset multilayer perceptron for nonlinear mapping to obtain the real-time dynamic thermal inertia and enthalpy change rate. S39. The real-time dynamic thermal inertia and enthalpy change rate are numerically added to obtain the thermal inertia characteristic value. The surface temperature field distribution of the rubber conveyor belt is read and the difference is calculated with the preset temperature threshold to obtain the current temperature deviation. The thermal inertia characteristic value and the current temperature deviation are spliced ​​in the channel dimension and input into the preset fully connected layer for linear transformation, and the real heat storage state feature vector is output.

[0011] Optionally, S4 specifically includes: S41. Construct an adaptive variable universe fuzzy controller based on an improved gravity search algorithm. Input the actual heat storage state feature vector of the rubber conveyor belt and the expected heat loss gradient as input data to the data input port of the adaptive variable universe fuzzy controller and map them to the preset initial fuzzy universe. S42. Call the Gaussian membership function calculation formula, take the difference between the input data and the center value of the Gaussian membership function as the numerator of the exponent term, take the square of the width of the Gaussian membership function as the denominator of the exponent term, calculate the membership degree of the input data on the fuzzy subset, initialize the scaling factor of the adaptive variable universe fuzzy controller, and preset the optimization range of the scaling factor. S43. Preset the population size parameter of the improved gravity search algorithm. Within the optimization range of the scaling factor, call the random number generation function to generate random numbers between zero and one. Add the lower bound of the optimization range to the random number multiplied by the difference between the upper and lower bounds of the optimization range to obtain the initial position vector of each individual. Generate an initial velocity vector with a value of zero. The position vector of each individual represents a set of scaling factor solutions of the adaptive variable universe fuzzy controller. Establish a one-to-one correspondence between the individual position and the scaling factor value of the adaptive variable universe fuzzy controller. S44. Introduce a reverse learning strategy to calculate the reverse position vector of each individual in the preset optimization space. The calculation formula is the upper bound value of the optimization space plus the lower bound value of the optimization space and minus the value of the initial position vector of the current individual. Calculate the fitness function value of the initial position vector and the reverse position vector of each individual. Compare the fitness function values ​​and retain the individuals with better fitness function values ​​to form the initial population. S45. Calculate the fitness value of each individual in the population, identify the individual with the smallest fitness function value in the current population as the current best individual, and identify the individual with the largest fitness function value as the current worst individual; S46. Calculate the difference between the best fitness value and the worst fitness value of the population at the current iteration number, calculate the difference between the fitness value and the worst fitness value of each individual, divide the difference between the fitness value and the worst fitness value by the difference between the best fitness value and the worst fitness value to obtain the intermediate ratio, and normalize the intermediate ratio to obtain the inertial mass of each individual. S47. Preset the initial value and decay rate of the gravitational constant, divide the current iteration number by the maximum iteration number to obtain the iteration ratio, calculate the natural constant by exponentiation of the iteration ratio, subtract the decay rate by the exponentiation result from the initial value, and calculate the gravitational constant under the current iteration number, so that the gravitational constant gradually decreases as the iteration number increases.

[0012] Optionally, calculating the fitness function values ​​of each individual's initial position vector and reverse position vector specifically includes: The numerical values ​​of the initial position vector or reverse position vector corresponding to the individual are assigned to the scaling factor of the adaptive variable universe fuzzy controller. The current universe is calculated by multiplying the scaling factor by the preset boundary value of the initial fuzzy universe. The feature vector of the real heat storage state is mapped to the current universe. The Gaussian membership function calculation formula is called to calculate the function value of the input data at the center value of the Gaussian membership function as the membership degree. Using preset membership matching fuzzy rules for inference, the absolute value of the difference between the output value of the adaptive variable universe fuzzy controller and the preset target temperature value is taken to obtain the temperature control error. The temperature control error is multiplied by the preset energy consumption weight coefficient to obtain the energy consumption term. The value of the temperature control error is added to the value of the energy consumption term, and the sum is used as the fitness function value of the individual.

[0013] Optionally, S5 specifically includes: S51. Perform the difference operation on the corresponding dimension elements of the position vectors of the two individuals, add the squares of the difference results and take the square root to obtain the Euclidean distance between the two individuals. Multiply the gravitational constant by the product of the inertial masses of the two individuals and divide by the Euclidean distance to obtain the magnitude of the gravitational force between the two individuals. Multiply the magnitude of the gravitational force by the position of the second individual minus the difference of the position vector of the first individual and divide by the Euclidean distance to obtain the direction vector of the gravitational force of the second individual on the first individual. S52. In the dimensional space, the gravitational direction vectors of all other individuals in the population that the individual experiences are added together in the corresponding dimension. The random number generation function is called to generate a random number between -1 and 1 and then superimposed on the accumulation result to obtain the resultant force vector of the individual. S53. Divide the resultant force vector by the inertial mass of each individual to obtain the acceleration vector of each individual. Multiply the acceleration vector by a preset time step coefficient and add it to the current velocity vector of the individual to update the velocity vector of the individual. Multiply the updated velocity vector by the preset time step coefficient and add it to the current position vector of the individual to update the position vector of the individual. S54. Check whether the updated position vector of an individual exceeds the optimization range of the preset scaling factor. If the value of the position vector is greater than or less than the upper or lower bound of the optimization range, the position vector is forcibly assigned the upper or lower bound of the optimization range. The scaling factor of the adaptive variable universe fuzzy controller is updated using the updated position vector. It is determined whether the maximum number of iterations has been reached. If it has, the output is the global optimal position vector. S55. Input the global optimal position vector as the optimal scaling factor into the adaptive variable universe fuzzy controller. Multiply the optimal scaling factor by the boundary value of the initial fuzzy universe to obtain the transformed fuzzy universe. Remap the real heat storage state feature vector to the transformed fuzzy universe. For each preset membership degree matching fuzzy rule output set, read the recalculated membership degree as the weight coefficient. S56. Multiply the center value of the output set of each preset membership degree matching fuzzy rule by the corresponding weight coefficient to obtain a weighted product, and add them to obtain the sum of the numerators. Add the weight coefficients to obtain the sum of the denominators. Divide the sum of the numerators by the sum of the denominators to calculate the precise output control quantity. According to the magnitude and polarity of the output control quantity, map it to the heating power adjustment quantity, cooling intensity adjustment quantity and speed compensation command respectively, and combine them to generate a refined control command set.

[0014] Optionally, S6 specifically includes: parsing the refined control instruction set and obtaining the heating power adjustment amount, cooling intensity adjustment amount and speed compensation instruction; sending drive signals to the heating execution unit, cooling execution unit and frequency conversion drive unit respectively to execute the corresponding operation; synchronously collecting real-time feedback data and converting and encapsulating it with the target parameters; and calling the display interface to render the processed data to the console interface in real time.

[0015] The beneficial effects of this invention are: This invention addresses the challenges of complex operating environments, strong nonlinear thermal state evolution, and large control lag in rubber conveyor belts by constructing a spatiotemporal feature map that integrates multi-source information and an environment prediction window based on spatiotemporal alignment. It employs a SwinTransformer V2 network, utilizing a hierarchical shift window mechanism and global self-attention characteristics to deeply mine the nonlinear correlation between the conveyor belt temperature field in local high-heat zones and the overall long-distance region. It inverses and calculates real-time dynamic thermal inertia and enthalpy change rate, extracting the true heat storage state feature vector. By inputting the true heat storage state feature vector and the expected heat loss gradient into an adaptive variable universe fuzzy controller, an improved gravity search algorithm with a back-learning strategy is introduced to globally optimize the scaling factor, calculate the gravitational constant and inertial mass, update the population position, dynamically adjust the fuzzy universe, and output a refined control command set including heating power, cooling intensity, and speed compensation. Furthermore, the control command set drives the heating, cooling, and frequency conversion drive units to work collaboratively, and real-time data is fed back to the console for display. Ultimately, it achieves dynamic balance and intelligent control of the mixed temperature when the rubber conveyor belt crosses different temperature zones, effectively improving the comprehensiveness of multi-source information fusion, the control accuracy under complex working conditions, and the system's anti-interference and adaptive capabilities. Attached Figure Description

[0016] The accompanying drawings are provided to further illustrate the invention and form part of the specification. They are used in conjunction with embodiments of the invention to explain the invention and do not constitute a limitation thereof. In the drawings: Figure 1 This is a flowchart of a method for intelligent control of mixing temperature of a rubber conveyor belt based on fuzzy control proposed in this invention; Figure 2 This is a flowchart of the multi-source spatiotemporal feature extraction and real heat storage state inversion based on the Swing Transformer V2 network proposed in this invention; Figure 3 The flowchart shows the optimization and fuzzy domain adaptive adjustment of the improved gravity search algorithm proposed in this invention, which incorporates a reverse learning strategy. Detailed Implementation

[0017] The present invention will now be described in further detail with reference to the accompanying drawings. These drawings are simplified schematic diagrams, illustrating only the basic structure of the invention, and therefore only show the components relevant to the invention.

[0018] refer to Figures 1-3 A method for intelligent control of mixing temperature in rubber conveyor belts based on fuzzy control includes the following steps: S1. Synchronously collect environmental heat distribution data and rubber conveyor belt operation status data along the entire rubber conveyor belt route. The environmental heat distribution data includes the surface temperature field distribution obtained by distributed fiber optic sensors or infrared thermal imagers and the meteorological environment temperature of different areas along the route. The rubber conveyor belt operation status data includes the real-time running speed, load tension, and real-time thermal conductivity coefficient of the rubber material. S2. Construct an environment prediction window based on spatiotemporal alignment. Based on the current running speed and position coordinates of the rubber conveyor belt, map the environmental heat distribution data onto the future movement trajectory of the rubber conveyor belt, and calculate the expected heat loss gradient of the target area that the rubber conveyor belt will enter in the future within a preset time period. S3. Based on the operating status data of the rubber conveyor belt, construct a spatiotemporal feature map and input it into the Swin Transformer V2 network. Utilize hierarchical shift window and global self-attention mechanism to mine the nonlinear correlation of the temperature field, invert and calculate the real-time dynamic thermal inertia and enthalpy change rate, and fuse them with the temperature deviation to extract and output the feature vector of the real heat storage state. S4. Input the actual heat storage state feature vector and the expected heat loss gradient into the adaptive variable universe fuzzy controller based on the improved gravity search algorithm, and map them to the initial fuzzy universe. Call the Gaussian membership function to calculate the membership degree, initialize the population, introduce a reverse learning strategy to select the optimal individual, and calculate and output the inertial mass and gravitational constant. S5. Based on the gravitational constant and inertial mass, update the population position vector and perform global optimization on the scaling factor to obtain the global optimal position vector. Input and adjust the fuzzy domain, use fuzzy inference to decouple the mixed interference, and directly map the calculation results into a refined control instruction set that includes heating power, cooling intensity and speed compensation. S6. Based on the refined control instruction set, the heating execution unit, cooling execution unit and frequency conversion drive unit distributed at the key nodes of the rubber conveyor belt are driven to work together and displayed in real time on the control console to achieve dynamic balance and intelligent control of the mixed temperature when crossing different temperature zones.

[0019] This invention significantly improves the accuracy of temperature control and system response speed for long-distance conveyor belts. By constructing a spatiotemporally aligned environmental prediction window, it achieves dynamic mapping between environmental heat distribution and motion trajectory, accurately predicting future heat loss gradients. Utilizing the Swing Transformer V2 network to mine nonlinear correlations in the temperature field, it accurately inverts dynamic thermal inertia and enthalpy change rate, effectively capturing the time-varying characteristics of thermal inertia. An adaptive variable universe fuzzy controller based on an improved gravity search algorithm is introduced. Through a back-learning strategy and global optimization, it adjusts the fuzzy universe and scaling factor in real time, decoupling mixed disturbances. This method ensures the coordinated operation of heating, cooling, and frequency conversion drive, achieving dynamic temperature balance across different temperature zones. Under complex operating conditions, it effectively eliminates temperature overshoot and control lag, significantly reduces energy consumption fluctuations, and greatly improves the safety, stability, and intelligent control level of conveyor belt operation.

[0020] In this embodiment, S1 specifically includes: S11. Synchronously trigger the distributed fiber optic sensors and infrared thermal imagers deployed along the entire rubber conveyor belt to collect environmental heat distribution data along the entire rubber conveyor belt. The environmental heat distribution data includes the surface temperature field distribution and the meteorological environment temperature of different areas along the route. S12. Read the data from the tension sensor deployed on the rubber conveyor belt to obtain the load tension of the rubber conveyor belt, read the feedback signal from the frequency converter to obtain the real-time running speed of the rubber conveyor belt, read the preset physical property database of rubber material to obtain the real-time thermal conductivity coefficient of the rubber material, and combine them to generate the running status data of the rubber conveyor belt.

[0021] In this embodiment, S2 specifically includes: S21. Calculate the displacement of the rubber conveyor belt in a future preset time period. Calculate the target position coordinates of the rubber conveyor belt in the future preset time period by using the sum of the current real-time position coordinates of the rubber conveyor belt and the displacement. Based on the target position coordinates, retrieve and obtain the meteorological environmental temperature of different areas along the route corresponding to the target position coordinates from the environmental heat distribution data. S23. Calculate the difference between the meteorological ambient temperature at the target location coordinates and the current surface temperature field distribution. Divide the difference by the displacement to calculate the expected heat loss gradient of the target area that the rubber conveyor belt will enter in the future preset time period.

[0022] In this embodiment, S3 specifically includes: S31. Align and stitch together the real-time running speed, load tension, and real-time thermal conductivity coefficient of the rubber conveyor belt in the rubber conveyor belt running status data according to the timestamp and the surface temperature field distribution of the rubber conveyor belt to construct a spatiotemporal feature map that integrates multi-source information. S32. Input the spatiotemporal feature map that integrates multi-source information into the PatchPartition layer of the Swin Transformer V2 network, divide the spatiotemporal feature map into several non-overlapping image blocks according to a preset size, flatten each image block into a one-dimensional vector, and linearly map the one-dimensional vector to a preset dimension to obtain a feature sequence containing position encoding. S33. Input the feature sequence into the Swing Transformer Block of the Swing Transformer V2 network, and use the hierarchical shift window mechanism to divide the feature sequence into multiple local windows according to the preset size. Calculate the query vector, key vector and value vector corresponding to all pixels in each local window. S34. Multiply the query vector by the transpose of the key vector and divide by the square root of the total number of pixels to obtain the attention score. Input the attention score into the Softmax function to normalize and obtain the weight coefficients, i.e., the multi-head self-attention weights. Multiply the weight coefficients by the value vector, sum all the product results within the local window, and update the feature vector within the local window. S35. Using a hierarchical shift window mechanism, perform a cyclic shift operation on the feature vectors within the updated local window, connect and merge the adjacent local windows after shifting to form an extended window, calculate the new query vector, key vector, and value vector on the merged feature vectors within the extended window, multiply the new query vector by the transpose of the new key vector and divide by the square root of the total number of pixels to obtain the extended attention score. S36. Input the extended attention score into the Softmax function to normalize and obtain the recalculated multi-head self-attention weights. Multiply the feature vector in the extended window with the recalculated multi-head self-attention weights to generate a feature vector that extracts the nonlinear correlation between the temperature field of the conveyor belt in the local high-heat zone and the overall long-distance region. S37. Input the feature vector with extracted nonlinear correlation into the PatchMerging layer of the Swin Transformer V2 network, perform a concatenation operation between four adjacent feature blocks to form a concatenated feature block, multiply the concatenated feature block by a weight matrix of a preset size, perform dimensionality reduction processing, and construct a high-level semantic feature containing global context information. S38. Input the high-level semantic features into the global self-attention layer of the Swin Transformer V2 network, calculate the correlation score matrix between any two high-level semantic features, multiply the high-level semantic features with the correlation score matrix to obtain the weighted global feature vector, and input it into the preset multilayer perceptron for nonlinear mapping to obtain the real-time dynamic thermal inertia and enthalpy change rate containing the global receptive field. S39. The real-time dynamic thermal inertia and enthalpy change rate are numerically added to obtain the thermal inertia characteristic value. The surface temperature field distribution of the rubber conveyor belt is read and the difference is calculated with the preset temperature threshold to obtain the current temperature deviation. The thermal inertia characteristic value and the current temperature deviation are spliced ​​in the channel dimension and input into the preset fully connected layer for linear transformation, and the real heat storage state feature vector is output.

[0023] This implementation introduces the Swin Transformer V2 network as an innovative technology, which has significant differences and advantages compared to traditional CNNs and standard Transformer models. Traditional CNNs are limited by their local receptive field, making it difficult to capture global thermal correlations between long-distance areas of the conveyor belt; while the standard Transformer has global modeling capabilities, its computational complexity increases quadratically with image size, facing huge resource consumption and inference latency bottlenecks when processing high-resolution thermal images. This invention, through the hierarchical shifting window mechanism of Swin Transformer V2, achieves refined local extraction of the temperature field and cross-window information interaction while maintaining linear computational complexity. In particular, the cyclic shifting and expanding window strategy used in S34 to S36 effectively establishes a nonlinear connection between local high-heat areas and the overall background, greatly enriching the completeness of feature representation. This mechanism can accurately invert dynamic thermal inertia and enthalpy change rate under complex thermal disturbances, significantly reducing computational costs, improving inference speed, and significantly enhancing the extraction accuracy and model robustness of the conveyor belt's heat storage state.

[0024] In this embodiment, S4 specifically includes: S41. Construct an adaptive variable universe fuzzy controller based on an improved gravity search algorithm. Input the actual heat storage state feature vector of the rubber conveyor belt and the expected heat loss gradient as input data to the data input port of the adaptive variable universe fuzzy controller and map them to the preset initial fuzzy universe. S42. Call the Gaussian membership function calculation formula, take the difference between the input data and the center value of the Gaussian membership function as the numerator of the exponent term, take the square of the width of the Gaussian membership function as the denominator of the exponent term, calculate the membership degree of the input data on the fuzzy subset, initialize the scaling factor of the adaptive variable universe fuzzy controller, and preset the optimization range of the scaling factor. S43. Preset the population size parameter of the improved gravity search algorithm. Within the optimization range of the scaling factor, call the random number generation function to generate random numbers between zero and one. Add the lower bound of the optimization range to the random number multiplied by the difference between the upper and lower bounds of the optimization range to obtain the initial position vector of each individual. Generate an initial velocity vector with a value of zero. The position vector of each individual represents a set of scaling factor solutions of the adaptive variable universe fuzzy controller. Establish a one-to-one correspondence between the individual position and the scaling factor value of the adaptive variable universe fuzzy controller. S44. Introduce a reverse learning strategy to calculate the reverse position vector of each individual in the preset optimization space. The calculation formula is the upper bound value of the optimization space plus the lower bound value of the optimization space and minus the value of the initial position vector of the current individual. Calculate the fitness function value of each individual's initial position vector and reverse position vector. Compare the fitness function values ​​and retain the individuals with better fitness function values ​​to form the initial population, so as to complete the initialization of the population and enhance the global search capability. S45. Calculate the fitness value of each individual in the population, identify the individual with the smallest fitness function value in the current population as the current best individual, and identify the individual with the largest fitness function value as the current worst individual; S46. Calculate the difference between the best and worst fitness values ​​of the population at the current iteration number. Calculate the difference between the fitness value and the worst fitness value of each individual. Divide the difference between the fitness value and the worst fitness value by the difference between the best and worst fitness values ​​to obtain the intermediate ratio. Normalize the intermediate ratio to obtain the inertial mass of each individual, so that individuals with better fitness values ​​have greater inertial mass. S47. Preset the initial value and decay rate of the gravitational constant, divide the current iteration number by the maximum iteration number to obtain the iteration ratio, calculate the natural constant by exponentiation of the iteration ratio, subtract the decay rate by the exponentiation result from the initial value, and calculate the gravitational constant under the current iteration number, so that the gravitational constant gradually decreases as the iteration number increases.

[0025] This invention optimizes the adaptive variable universe fuzzy controller by introducing an improved gravity search algorithm, achieving global dynamic optimization of the scaling factor and adaptive adjustment of the control rules. The actual heat storage state feature vector and the expected heat loss gradient are mapped to the initial fuzzy universe. The membership degree is accurately calculated using a Gaussian membership function. A high-quality initial population is constructed through a back-learning strategy, and the inertial mass and gravitational constant are calculated. The algorithm normalizes the inertial mass using fitness differences, giving superior individuals a greater gravitational guidance direction, and the constant decays with iteration to balance the search capability. This invention effectively avoids traditional methods getting trapped in local optima, adjusts the fuzzy universe in real time under complex thermal disturbance environments, and significantly improves the system's response speed and control accuracy.

[0026] The improved gravity search algorithm of this invention is similar to the original gravity search algorithm in that both retain the core physical mechanism of the gravity search algorithm, namely, calculating the magnitude of gravity based on the mass of individuals in the population, using the law of universal gravitation to guide individuals to move in the search space, evaluating the quality of an individual's position by calculating the fitness value, and approximating the global optimum based on the position update formula.

[0027] The difference lies in that this invention breaks through the limitation of the original algorithm, which relies solely on random initialization distributions for searching, by introducing a reverse learning strategy and constructing a high-starting-point initialization system. Building upon the original algorithm's direct generation of random position vectors, this invention incorporates a reverse learning operation in step S44. This strategy calculates the corresponding reverse position based on the current individual's position in the optimization space by subtracting the position value from the sum of the upper and lower bounds. Next, the fitness function values ​​of the original and reverse individuals are compared, and the individuals with better performance are retained to form the initial population. Furthermore, in step S46, a normalization mechanism is introduced to the difference between the best and worst fitness values, accurately calculating the inertial mass, thus giving individuals with better fitness a larger gravitational weight, rather than a simple linear mapping.

[0028] The benefits of the improvements lie in the fact that, through the reverse learning strategy and dynamic inertial mass calculation, the improved gravity search algorithm can effectively avoid premature convergence caused by uneven initial population distribution, significantly enhance the algorithm's ability to escape local optima, and achieve a balance between global search range and local accuracy. This design significantly improves the algorithm's depth of exploration in the complex nonlinear space of the scaling factor, enabling it to locate the optimal parameter combination of the fuzzy controller more quickly. The optimized mass weighting mechanism accelerates the convergence speed while ensuring the robustness of the control strategy, enhancing the system's adaptive adjustment capability and response accuracy under complex thermal environments.

[0029] In this embodiment, calculating the fitness function values ​​of each individual's initial position vector and reverse position vector specifically includes: The numerical values ​​of the initial position vector or reverse position vector corresponding to the individual are assigned to the scaling factor of the adaptive variable universe fuzzy controller. The current universe is calculated by multiplying the scaling factor by the preset boundary value of the initial fuzzy universe. The feature vector of the real heat storage state is mapped to the current universe. The Gaussian membership function calculation formula is called to calculate the function value of the input data at the center value of the Gaussian membership function as the membership degree. Using preset membership matching fuzzy rules for inference, the absolute value of the difference between the output value of the adaptive variable universe fuzzy controller and the preset target temperature value is taken to obtain the temperature control error. The temperature control error is multiplied by the preset energy consumption weight coefficient to obtain the energy consumption term. The value of the temperature control error is added to the value of the energy consumption term, and the sum is used as the fitness function value of the individual.

[0030] In this embodiment, S5 specifically includes: S51. Perform the difference operation on the corresponding dimension elements of the position vectors of the two individuals, add the squares of the difference results and take the square root to obtain the Euclidean distance between the two individuals. Multiply the gravitational constant by the product of the inertial masses of the two individuals and divide by the Euclidean distance to obtain the magnitude of the gravitational force between the two individuals. Multiply the magnitude of the gravitational force by the position of the second individual minus the difference of the position vector of the first individual and divide by the Euclidean distance to obtain the direction vector of the gravitational force of the second individual on the first individual. S52. In the dimensional space, the gravitational direction vectors of all other individuals in the population that the individual experiences are added and accumulated in the corresponding dimension. A random number generation function is called to generate a random number between -1 and 1 and superimposed on the accumulation result. The random number is a preset interval distribution value used to simulate random disturbances in the search process, and the resultant force vector of the individual is obtained. S53. Divide the resultant force vector by the inertial mass of each individual to obtain the acceleration vector of each individual. Multiply the acceleration vector by a preset time step coefficient and add it to the current velocity vector of the individual to update the velocity vector of the individual. Multiply the updated velocity vector by the preset time step coefficient and add it to the current position vector of the individual to update the position vector of the individual. S54. Check whether the updated position vector of an individual exceeds the optimization range of the preset scaling factor. If the value of the position vector is greater than or less than the upper or lower bound of the optimization range, the position vector is forcibly assigned the upper or lower bound of the optimization range. The scaling factor of the adaptive variable universe fuzzy controller is updated using the updated position vector. It is determined whether the maximum number of iterations has been reached. If it has, the output is the global optimal position vector. S55. Input the global optimal position vector as the optimal scaling factor into the adaptive variable universe fuzzy controller. Multiply the optimal scaling factor by the boundary value of the initial fuzzy universe to obtain the transformed fuzzy universe. Remap the real heat storage state feature vector to the transformed fuzzy universe. For each preset membership degree matching fuzzy rule output set, read the recalculated membership degree as the weight coefficient. S56. Multiply the center value of the output set of each preset membership degree matching fuzzy rule by the corresponding weight coefficient to obtain a weighted product, and add them to obtain the sum of the numerators. Add the weight coefficients to obtain the sum of the denominators. Divide the sum of the numerators by the sum of the denominators to calculate the precise output control quantity. According to the magnitude and polarity of the output control quantity, map it to the heating power adjustment quantity, cooling intensity adjustment quantity and speed compensation command respectively, and combine them to generate a refined control command set.

[0031] In this embodiment, S6 specifically includes: parsing the refined control instruction set and obtaining the heating power adjustment amount, cooling intensity adjustment amount and speed compensation instruction; sending drive signals to the heating execution unit, cooling execution unit and frequency conversion drive unit respectively to execute the corresponding operation; synchronously collecting real-time feedback data and converting and encapsulating it with the target parameters; and calling the display interface to render the processed data to the console interface in real time.

[0032] Example 1: To verify the feasibility of this invention in the field of industrial thermal control, it was applied to an intelligent temperature control system for a fully enclosed, long-distance heat-resistant steel wire rope rubber conveyor belt belonging to a large energy and chemical group. This conveyor belt is 12.8 kilometers long, running through the raw material storage area in the north of the plant, the core reaction workshop, and the finished product wharf in the south. It is responsible for continuously transporting high-temperature calcined materials to the processing stage, with a designed conveying capacity of 3500 tons per hour and a belt speed set at 4.5 meters per second. Due to the long conveying distance and the crossing of three drastically different thermal environments—outdoor open-air section, semi-enclosed trestle section, and indoor constant-temperature section—coupled with an initial material temperature exceeding 160°C, the rubber belt is highly susceptible to vulcanization reduction due to localized overheating or tearing due to uneven thermal expansion and contraction during operation. The site is equipped with over 400 distributed infrared temperature probes, 48 ​​environmental temperature and humidity sensors, and high-precision tension and speed detection devices, generating over 850GB of temperature control and operational data daily. Traditional PID control strategies rely on a single temperature threshold for on / off heating or cooling. When faced with sudden cold waves, drastic changes in sunlight, or fluctuations in material flow, the adjustment response is severely lagging. The surface temperature of the conveyor belt often fluctuates by more than 15°C in a short period of time, leading to accelerated belt aging and high energy consumption.

[0033] In practical deployment, the method of this invention aligns and resamples the data from the aforementioned multi-source heterogeneous sensors according to timestamps, constructing a spatiotemporal feature map that includes environmental meteorological parameters, operating condition data, and surface temperature field distribution. Utilizing the powerful hierarchical feature extraction capabilities of the SwinTransformer V2 network, the model performs high-precision inversion of the thermal inertia and real-time enthalpy change rate of different sections of the long-distance conveyor belt, capturing the nonlinear evolution trend of the temperature field that is difficult to identify using traditional methods. Based on this, the system calculates the current heat storage state feature vector in real time and inputs it into the adaptive variable universe fuzzy controller. An improved gravity search algorithm incorporating a reverse learning strategy runs continuously in the background, dynamically optimizing the scaling factor of the fuzzy controller, enabling the control universe to automatically shrink or expand according to the current temperature deviation and rate of change, thereby outputting refined control commands including heating power, cooling airflow, and speed compensation. This mechanism realizes a shift from passive adjustment to proactive prediction, especially under complex operating conditions such as sudden increases in material flow or sudden drops in ambient temperature, enabling early intervention to maintain the dynamic balance of the belt temperature. Table 1 details a comparison of key performance data between the method of this invention and traditional PID control during this period, in terms of conveyor belt temperature stability, energy consumption level, and equipment maintenance frequency. Table 1. Comparison of measured data between the present invention and traditional PID control in temperature control performance of long-distance conveyor belts.

[0034] Based on the comparative data shown in Table 1, it can be seen that the intelligent temperature control method based on the Swin Transformer V2 network and the improved gravity search algorithm proposed in this invention has significant performance advantages over traditional PID control in long-distance conveyor belt temperature control, especially in key indicators such as temperature stability, response speed, energy consumption control and equipment protection.

[0035] Regarding temperature control accuracy, this invention maintains temperature fluctuations within ±2.1℃ in all four typical test conditions, far superior to the average fluctuation level of ±8.9℃ of traditional systems. For example, in the "cold wave cooling condition" with severe environmental interference, traditional methods can only achieve a fluctuation range of ±12.5℃, frequently triggering over-temperature alarms; while this invention, through precise thermal inertia inversion and adaptive adjustment of the universe of discourse, strictly controls fluctuations within ±2.1℃, effectively ensuring the stability of belt operation.

[0036] In terms of control response and energy saving, this invention optimizes the scaling factor through an improved gravity search algorithm, achieving dynamic optimization of the control strategy. Data shows that the integral of absolute temperature deviation (IAE) and unit transmission energy consumption of this invention are significantly reduced under various operating conditions, and the number of equipment start-ups and shutdowns is reduced by approximately 80% compared to traditional methods. Particularly under the "full load high-temperature operation" condition, this invention reduces unit energy consumption to 0.63 kWh / t and the number of start-ups and shutdowns to 25 times, representing reductions of 12.5% ​​and 86.6% respectively compared to traditional methods, greatly improving the system's operating efficiency and economy.

[0037] Regarding equipment health protection, this invention effectively avoids belt damage caused by drastic temperature changes. Under all operating conditions, the number of abnormal temperature rises on the belt surface is essentially zero, while traditional methods resulted in a cumulative 29 abnormalities. Overall, this invention effectively solves the problems of control lag and energy waste in long-distance conveyor belts under complex operating conditions, and possesses extremely high practical application value.

[0038] The above description is only a preferred embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any equivalent substitutions or modifications made by those skilled in the art within the scope of the technology disclosed in the present invention, based on the technical solution and inventive concept of the present invention, should be covered within the scope of protection of the present invention.

Claims

1. A method for intelligent control of mixing temperature in a rubber conveyor belt based on fuzzy control, characterized in that, Includes the following steps: S1. Synchronously collect environmental heat distribution data and rubber conveyor belt operating status data for the entire rubber conveyor belt route; S2. Construct an environment prediction window based on spatiotemporal alignment. Based on the current running speed and position coordinates of the rubber conveyor belt, map the environmental heat distribution data onto the future movement trajectory of the rubber conveyor belt, and calculate the expected heat loss gradient of the target area to be entered within a preset time period in the future. S3. Based on the operating status data of the rubber conveyor belt, construct a spatiotemporal feature map and input it into the Swin Transformer V2 network. Utilize hierarchical shift window and global self-attention mechanism to mine the nonlinear correlation of the temperature field, invert and calculate the real-time dynamic thermal inertia and enthalpy change rate, and fuse them with the temperature deviation to extract and output the feature vector of the real heat storage state. S4. Input the actual heat storage state feature vector and the expected heat loss gradient into the adaptive variable universe fuzzy controller based on the improved gravity search algorithm, and map them to the initial fuzzy universe. Call the Gaussian membership function to calculate the membership degree, initialize the population, introduce a reverse learning strategy to select the optimal individual, and calculate and output the inertial mass and gravitational constant. S5. Based on the gravitational constant and inertial mass, update the population position vector and perform global optimization on the scaling factor to obtain the global optimal position vector. Input and adjust the fuzzy universe of discourse, use fuzzy inference to decouple the mixed interference, and directly map the calculation results into a refined control instruction set. S6. Based on the refined control instruction set, drive the heating execution unit, cooling execution unit and frequency conversion drive unit distributed at key nodes of the rubber conveyor belt to work together and display the results in real time on the control console.

2. The intelligent temperature control method for mixing rubber conveyor belts based on fuzzy control according to claim 1, characterized in that, S1 specifically includes: S11. Synchronously trigger the distributed fiber optic sensors and infrared thermal imagers deployed along the entire rubber conveyor belt to collect environmental heat distribution data along the entire rubber conveyor belt. The environmental heat distribution data includes the surface temperature field distribution and the meteorological environment temperature of different areas along the route. S12. Read the data from the tension sensor deployed on the rubber conveyor belt to obtain the load tension of the rubber conveyor belt, read the feedback signal from the frequency converter to obtain the real-time running speed of the rubber conveyor belt, read the preset physical property database of rubber material to obtain the real-time thermal conductivity coefficient of the rubber material, and combine them to generate the running status data of the rubber conveyor belt.

3. The intelligent control method for mixing temperature of a rubber conveyor belt based on fuzzy control according to claim 1, characterized in that, S2 specifically includes: S21. Calculate the displacement of the rubber conveyor belt in a future preset time period. Calculate the target position coordinates of the rubber conveyor belt in the future preset time period by using the sum of the current real-time position coordinates of the rubber conveyor belt and the displacement. Based on the target position coordinates, retrieve and obtain the meteorological environmental temperature of different areas along the route corresponding to the target position coordinates from the environmental heat distribution data. S23. Calculate the difference between the meteorological ambient temperature at the target location coordinates and the current surface temperature field distribution. Divide the difference by the displacement to calculate the expected heat loss gradient of the target area that the rubber conveyor belt will enter in the future preset time period.

4. The intelligent control method for mixing temperature of a rubber conveyor belt based on fuzzy control according to claim 1, characterized in that, S3 specifically includes: S31. Align and stitch together the real-time running speed, load tension, and real-time thermal conductivity coefficient of the rubber conveyor belt in the rubber conveyor belt running status data according to the timestamp and the surface temperature field distribution of the rubber conveyor belt to construct a spatiotemporal feature map that integrates multi-source information. S32. Input the spatiotemporal feature map that integrates multi-source information into the PatchPartition layer of the Swin Transformer V2 network, divide the spatiotemporal feature map into several non-overlapping image blocks according to a preset size, flatten each image block into a one-dimensional vector, and linearly map the one-dimensional vector to a preset dimension to obtain the feature sequence. S33. Input the feature sequence into the Swing Transformer Block of the Swing Transformer V2 network, and use the hierarchical shift window mechanism to divide the feature sequence into multiple local windows according to the preset size. Calculate the query vector, key vector and value vector corresponding to all pixels in each local window. S34. Multiply the query vector by the transpose of the key vector and divide by the square root of the total number of pixels to obtain the attention score. Input the attention score into the Softmax function to normalize and obtain the weight coefficients, i.e. the multi-head self-attention weights. Multiply the weight coefficients by the value vector, add all the product results in the local window, and update the feature vector in the local window. S35. Using a hierarchical shift window mechanism, perform a cyclic shift operation on the feature vectors within the updated local window, connect and merge the adjacent local windows after shifting to form an extended window, calculate the new query vector, key vector, and value vector on the merged feature vectors within the extended window, multiply the new query vector by the transpose of the new key vector and divide by the square root of the total number of pixels to obtain the extended attention score. S36. Input the extended attention score into the Softmax function to normalize and obtain the recalculated multi-head self-attention weights. Multiply the feature vector within the extended window with the recalculated multi-head self-attention weights to generate a feature vector that has extracted non-linear correlations. S37. Input the feature vector with extracted nonlinear correlation into the PatchMerging layer of the Swin Transformer V2 network, perform a concatenation operation between four adjacent feature blocks to form a concatenated feature block, multiply the concatenated feature block by a weight matrix of a preset size, perform dimensionality reduction processing, and construct high-level semantic features. S38. Input the high-level semantic features into the global self-attention layer of the Swin Transformer V2 network, calculate the correlation score matrix between any two high-level semantic features, multiply the high-level semantic features with the correlation score matrix to obtain the weighted global feature vector, and input it into the preset multilayer perceptron for nonlinear mapping to obtain the real-time dynamic thermal inertia and enthalpy change rate. S39. The real-time dynamic thermal inertia and enthalpy change rate are numerically added to obtain the thermal inertia characteristic value. The surface temperature field distribution of the rubber conveyor belt is read and the difference is calculated with the preset temperature threshold to obtain the current temperature deviation. The thermal inertia characteristic value and the current temperature deviation are spliced ​​in the channel dimension and input into the preset fully connected layer for linear transformation, and the real heat storage state feature vector is output.

5. The intelligent control method for mixing temperature of a rubber conveyor belt based on fuzzy control according to claim 1, characterized in that, S4 specifically includes: S41. Construct an adaptive variable universe fuzzy controller based on an improved gravity search algorithm. Input the actual heat storage state feature vector of the rubber conveyor belt and the expected heat loss gradient as input data to the data input port of the adaptive variable universe fuzzy controller and map them to the preset initial fuzzy universe. S42. Call the Gaussian membership function calculation formula, take the difference between the input data and the center value of the Gaussian membership function as the numerator of the exponent term, take the square of the width of the Gaussian membership function as the denominator of the exponent term, calculate the membership degree of the input data on the fuzzy subset, initialize the scaling factor of the adaptive variable universe fuzzy controller, and preset the optimization range of the scaling factor. S43. Preset the population size parameter of the improved gravity search algorithm. Within the optimization range of the scaling factor, call the random number generation function to generate random numbers between zero and one. Add the lower bound of the optimization range to the random number multiplied by the difference between the upper and lower bounds of the optimization range to obtain the initial position vector of each individual. Generate an initial velocity vector with a value of zero. The position vector of each individual represents a set of scaling factor solutions of the adaptive variable universe fuzzy controller. Establish a one-to-one correspondence between the individual position and the scaling factor value of the adaptive variable universe fuzzy controller. S44. Introduce a reverse learning strategy to calculate the reverse position vector of each individual in the preset optimization space. The calculation formula is the upper bound value of the optimization space plus the lower bound value of the optimization space and minus the value of the initial position vector of the current individual. Calculate the fitness function value of the initial position vector and the reverse position vector of each individual. Compare the fitness function values ​​and retain the individuals with better fitness function values ​​to form the initial population. S45. Calculate the fitness value of each individual in the population, identify the individual with the smallest fitness function value in the current population as the current best individual, and identify the individual with the largest fitness function value as the current worst individual; S46. Calculate the difference between the best fitness value and the worst fitness value of the population at the current iteration number, calculate the difference between the fitness value and the worst fitness value of each individual, divide the difference between the fitness value and the worst fitness value by the difference between the best fitness value and the worst fitness value to obtain the intermediate ratio, and normalize the intermediate ratio to obtain the inertial mass of each individual. S47. Preset the initial value and decay rate of the gravitational constant, divide the current iteration number by the maximum iteration number to obtain the iteration ratio, calculate the natural constant by exponentiation of the iteration ratio, subtract the decay rate by the exponentiation result from the initial value, and calculate the gravitational constant under the current iteration number, so that the gravitational constant gradually decreases as the iteration number increases.

6. The intelligent temperature control method for mixing rubber conveyor belts based on fuzzy control according to claim 5, characterized in that, The calculation of the fitness function values ​​for each individual's initial position vector and reverse position vector specifically includes: The numerical values ​​of the initial position vector or reverse position vector corresponding to the individual are assigned to the scaling factor of the adaptive variable universe fuzzy controller. The current universe is calculated by multiplying the scaling factor by the preset boundary value of the initial fuzzy universe. The feature vector of the real heat storage state is mapped to the current universe. The Gaussian membership function calculation formula is called to calculate the function value of the input data at the center value of the Gaussian membership function as the membership degree. Using preset membership matching fuzzy rules for inference, the absolute value of the difference between the output value of the adaptive variable universe fuzzy controller and the preset target temperature value is taken to obtain the temperature control error. The temperature control error is multiplied by the preset energy consumption weight coefficient to obtain the energy consumption term. The value of the temperature control error is added to the value of the energy consumption term, and the sum is used as the fitness function value of the individual.

7. The intelligent temperature control method for mixing rubber conveyor belts based on fuzzy control according to claim 1, characterized in that, S5 specifically includes: S51. Perform the difference operation on the corresponding dimension elements of the position vectors of the two individuals, add the squares of the difference results and take the square root to obtain the Euclidean distance between the two individuals. Multiply the gravitational constant by the product of the inertial masses of the two individuals and divide by the Euclidean distance to obtain the magnitude of the gravitational force between the two individuals. Multiply the magnitude of the gravitational force by the position of the second individual minus the difference of the position vector of the first individual and divide by the Euclidean distance to obtain the direction vector of the gravitational force of the second individual on the first individual. S52. In the dimensional space, the gravitational direction vectors of all other individuals in the population that the individual experiences are added together in the corresponding dimension. The random number generation function is called to generate a random number between -1 and 1 and then superimposed on the accumulation result to obtain the resultant force vector of the individual. S53. Divide the resultant force vector by the inertial mass of each individual to obtain the acceleration vector of each individual. Multiply the acceleration vector by a preset time step coefficient and add it to the current velocity vector of the individual to update the velocity vector of the individual. Multiply the updated velocity vector by the preset time step coefficient and add it to the current position vector of the individual to update the position vector of the individual. S54. Check whether the updated position vector of an individual exceeds the optimization range of the preset scaling factor. If the value of the position vector is greater than or less than the upper or lower bound of the optimization range, the position vector is forcibly assigned the upper or lower bound of the optimization range. The scaling factor of the adaptive variable universe fuzzy controller is updated using the updated position vector. It is determined whether the maximum number of iterations has been reached. If it has, the output is the global optimal position vector. S55. Input the global optimal position vector as the optimal scaling factor into the adaptive variable universe fuzzy controller. Multiply the optimal scaling factor by the boundary value of the initial fuzzy universe to obtain the transformed fuzzy universe. Remap the real heat storage state feature vector to the transformed fuzzy universe. For each preset membership degree matching fuzzy rule output set, read the recalculated membership degree as the weight coefficient. S56. Multiply the center value of the output set of each preset membership degree matching fuzzy rule by the corresponding weight coefficient to obtain a weighted product, and add them to obtain the sum of the numerators. Add the weight coefficients to obtain the sum of the denominators. Divide the sum of the numerators by the sum of the denominators to calculate the precise output control quantity. According to the magnitude and polarity of the output control quantity, map it to the heating power adjustment quantity, cooling intensity adjustment quantity and speed compensation command respectively, and combine them to generate a refined control command set.

8. The intelligent temperature control method for mixing rubber conveyor belts based on fuzzy control according to claim 1, characterized in that, S6 specifically includes: parsing the refined control instruction set and obtaining the heating power adjustment amount, cooling intensity adjustment amount and speed compensation instruction; sending drive signals to the heating execution unit, cooling execution unit and frequency conversion drive unit respectively to execute the corresponding operation; synchronously collecting real-time feedback data and converting and encapsulating it with the target parameters; and calling the display interface to render the processed data to the console interface in real time.

Images