Machining heat treatment air duct pressure control optimization method

Abstract

The invention discloses a machining heat treatment air duct pressure control optimization method, which comprises the steps of constructing a machining heat treatment air duct pressure PID control system, introducing an improved neuron synaptic optimization algorithm, constructing an improved neuron synaptic optimization algorithm module, and optimizing the machining heat treatment air duct pressure based on the machining heat treatment air duct pressure PID control system. An improved neuron synaptic optimization algorithm is used for optimizing to obtain an optimal control parameter, and the optimal control parameter is mapped to the air duct pressure PID controller module, so that the machining heat treatment air duct pressure is better controlled. An improved neuron synaptic optimization algorithm is introduced, a machining heat treatment air duct pressure PID control system is constructed, the improved neuron synaptic optimization algorithm is used for carrying out offline iterative calculation in a preset feature search space, optimal control parameters are obtained through optimization and mapped to an air duct pressure PID controller module, and therefore the optimal control parameters are obtained. And more excellent control over the pressure of the machining heat treatment air duct is achieved.

Description

Technical Field

[0001] This invention belongs to the technical field of control optimization, and in particular relates to a method for optimizing the pressure control of air ducts in machining and heat treatment. Background Technology

[0002] In the heat treatment process of machining, the air duct system is a key component for controlling the circulation of atmosphere and heat transfer within the furnace. Maintaining stable air duct pressure is a fundamental requirement for ensuring uniform heating and consistent cooling rates of metal workpieces. Air duct pressure not only directly affects the spatial distribution of the temperature field within the furnace but also influences process aspects such as the flow direction of protective gas, heat exchange efficiency, and chemical reactions on the workpiece surface. Excessive fluctuations in air duct pressure can easily lead to quality defects such as workpiece deformation, uneven hardness, and surface oxidation, directly impacting the reliability of the entire machine. Currently, in the heat treatment field, variable frequency speed control technology or valve opening adjustment are commonly used to intervene in air duct pressure. A closed-loop control circuit is constructed using hardware devices such as centrifugal fans, electric regulating valves, and airflow sensors to achieve real-time pressure tracking and control. However, problems such as insufficient adjustment accuracy, slow dynamic response, and weak anti-interference capabilities often exist, making it difficult to quickly smooth out pressure fluctuations during transient conditions such as workpiece loading / unloading or furnace door opening / closing. Summary of the Invention

[0003] To overcome the technical problems described in the background section, this invention provides an optimization method for controlling airflow pressure in machining heat treatment ducts. It introduces an improved swarm intelligence algorithm, specifically an improved neuron synaptic optimization algorithm. By constructing a PID control system for the airflow pressure in machining heat treatment ducts, the improved neuron synaptic optimization algorithm performs offline iterative calculations within a preset feature search space to find the optimal control parameters. These optimal control parameters are then mapped to the airflow pressure PID controller module. The optimized parameter combination enables better control of the airflow pressure in machining heat treatment ducts.

[0004] The technical solution of this invention is as follows: Based on the above technical solution, the technical effect of the present invention is as follows: A method for optimizing the pressure control of air ducts in machining heat treatment includes the following steps: Step S1: Construct a PID control system for air duct pressure in machining and heat treatment, including an air duct pressure error calculation module, an air duct pressure PID controller module, an improved neuron synapse optimization algorithm module, an air duct pressure regulation module, and an air duct pressure monitoring module. Step S2: Introduce an improved neuron synapse optimization algorithm. The improvement strategy used is as follows: Step S21: In the initialization phase, the Tent chaotic mapping initialization strategy is introduced. The Tent chaotic mapping generates a chaotic sequence with ergodicity and randomness, which improves the uniformity of the initial neuron population in the multidimensional search space. Step S22: In the signal strength normalization and synaptic plasticity adjustment operation, a normalized Euclidean distance calculation strategy is introduced. By mapping the parameter dimensions with different value ranges to a unified scale, the dominant influence of the larger value range on the synaptic connection weight is eliminated, ensuring that the influence of each PID parameter dimension on the synaptic connection strength is balanced. Step S23: In the position update stage, a long-range axon pulse projection position update strategy is introduced. By utilizing the random alternation of long and short step sizes of the Levy flight mechanism, the algorithm's ability to escape local optima in the search space is enhanced and the global search efficiency is improved. Step S24: After the position update, an inhibitory signal reflection verification strategy is introduced to establish a reverse solution exploration mechanism. By probabilistically generating and evaluating the reverse synaptic feature solution of the current neuron position, the effective search range is expanded and the algorithm convergence is accelerated. Step S3: Use the improved neuron synapse optimization algorithm to perform offline iterative calculations in the feature search space to find the optimal control parameters; Step S4: Map the optimal control parameters to the duct pressure PID controller module, and use the optimized parameter combination to achieve better control of the duct pressure in machining and heat treatment.

[0005] Furthermore, in the mechanical processing and heat treatment air duct pressure PID control system constructed in step S1, the air duct pressure monitoring module collects the actual pressure signal in real time and feeds it back to the error calculation module. It compares the signal with the target setpoint to generate a deviation signal and inputs it to the PID controller module. At the same time, the improved neuron synapse optimization algorithm module performs static parameter configuration on the controller based on the parameters obtained by offline optimization of the controlled object model. Then, the PID controller module calculates the precise control increment based on the deviation signal and outputs it to the air duct pressure regulation module to complete the full closed-loop linear regulation of the air duct pressure.

[0006] Furthermore, the Tent chaotic mapping initialization strategy in step S21 includes the following steps: Step S211: For each neuron The Dimensional features, generating initial chaotic sequence values ; Step S212: Iteratively update according to the Tent mapping rules The mapping logic is as follows: , In the formula This represents the internal iteration index of the chaotic map. Indicates the first The first neuron 3D features in the 1st dimension State values ​​during the next chaotic evolution This indicates the state value for the next evolution; Step S213: Generate chaotic values Mapped to the physical search space, , In the formula Indicates the first The first neuron 3D positional components, express The first in the vector One element, express The first in the vector One element, This represents the corresponding chaotic variable.

[0007] Furthermore, the normalized Euclidean distance calculation strategy in step S22 includes the following steps: Step S221, determine the first The search space for 3D neuron synaptic features, i.e., the range of the 3D neuron synaptic features. Upper bound of 3D search space and the lower world The difference; Step S222: Calculate neurons With neurons In the Normalized position difference components on the dimension , , In the formula Represents the normalized position difference components. Represents neurons In the Positional components of the dimension, Represents neurons In the Positional components of the dimension, Indicates the first Upper bound of the 3D search space Indicates the first Lower bound of the dimensional search space; Step S223: Based on the normalized position difference components of all dimensions Calculate the normalized Euclidean distance This distance is then substituted into the Hebb learning rule formula to update the synaptic connection strength. , In the formula This represents the normalized Euclidean distance. Indicates the PID parameter dimension. This indicates the result calculated in step S222. Normalized position difference components; Next, update the synaptic connection strength. , In the formula This indicates the updated synaptic connection strength. Indicates the current synaptic connection strength. Indicates the basic learning rate. Represents neurons The normalized signal strength, Represents neurons The normalized signal strength, This represents the distance attenuation factor.

[0008] Furthermore, the long-range axon pulse projection position update strategy in step S23 includes the following steps: Step S231: Calculate the Levy flight step vector , , In the formula This represents the Lévy flight step size vector. This indicates that the standard deviation is 1. A normally distributed random variable, Let represent a random variable that follows a normal distribution with a standard deviation of 1. The Levi Flight Index indicates that... It is calculated from a formula that includes the gamma function; Step S232: Introduce the Levy flight step scaling factor. Combining the sum of synaptic traction forces with the current optimal position The guiding role; Step S233: Perform position update calculation. , In the formula Indicates the updated neuron location. Indicates the neuron position before the update. Indicates that it comes from the population excluding neurons The sum of synaptic traction forces of all other neurons, This represents the scaling factor for Levy's flight step size. This indicates the location of the currently most activated neuron.

[0009] Furthermore, the suppression signal reflection verification strategy in step S24 includes the following steps: Step S241: Generate an interval Random numbers within, and inverse learning probabilities Compare; Step S242: If the condition is met, calculate the reverse synaptic feature solution for the current neuron position. , , In the formula This represents the characteristic solution of the reverse synapse. and These represent the lower and upper bound vectors of the search space, respectively. Indicates the neuron's position after the position update; Step S243: Evaluate the antisynaptic feature solution The signal strength, if The signal strength is better than at the original location The signal strength is then used replace .

[0010] Furthermore, step S3, which involves tuning the PID control parameters of the air duct pressure in the machining heat treatment air duct pressure PID control system using an improved neural synaptic optimization algorithm, includes the following detailed steps: Step S31: Initialize system model and algorithm parameters: Establish a mathematical model of the duct pressure control system and set the dimensions of the PID parameters to be optimized. Set the size of the neuron population Set the maximum number of neural remodeling cycles. And setting the search space boundaries for PID parameters; Step S32: Initialize the synaptic connection strength matrix For elements in the matrix Assign initial values; Step S33: Execute the Tent chaotic mapping initialization strategy as described in claim 3 to generate an initial neuron population. The signal strength of the initial population was assessed to identify the most strongly activated neurons. ; Step S34: Enter the iterative optimization process, and execute the following sub-steps in each iteration: Sub-step S341: Calculate the normalized signal intensity of each neuron in the current population; Sub-step S342: Execute the normalized Euclidean distance calculation strategy as described in claim 4, and update the synaptic connection strength matrix according to the Heb learning rule. ; Sub-step S343: Execute the long-range axonal pulse projection position update strategy as described in claim 5 to update the neuron position; Sub-step S344: Execute the inhibitory signal reflection verification strategy as described in claim 6, and perform reverse learning optimization on the updated position; Sub-step S345: Perform synaptic pruning and reinforcement operations, removing synaptic pruning elements below the synaptic pruning threshold. Weak connections are established, and strong connections are reinforced by displacement. Sub-step S346: Evaluate the updated neuron population and update the globally most activated neurons. ; Step S35: Determine whether the current iteration count has reached the maximum number of neural remodeling cycles. If the condition is met, the iteration terminates and the output is completed. If the optimal PID control parameters are used, then proceed to step S34.

[0011] 1. This invention introduces an improved swarm intelligence algorithm and an improved neuron synapse optimization algorithm, which can perform in-depth optimization of the PID parameters of the air duct pressure control system, thereby improving the control performance of the mechanical processing and heat treatment air duct pressure PID control system. 2. This invention lays the foundation for the distribution diversity of global search by constructing an initial population with high ergodicity using the Tent chaotic mapping strategy during the initialization phase; 3. This invention effectively eliminates the imbalance of parameter dimensions with different value ranges on synaptic connection weights by using a normalized Euclidean distance calculation strategy, ensuring that the contribution of each PID parameter is balanced in the optimization process. 4. This invention combines the Levy flight mechanism in the long-range axon pulse projection position update strategy, giving the algorithm the ability to escape local traps in a complex solution space, which greatly enhances the global search efficiency; 5. The present invention expands the effective search boundary and significantly accelerates the convergence process of the algorithm by introducing a reverse solution exploration mechanism through the suppression signal reflection verification strategy. Attached Figure Description

[0012] Figure 1 This is a flowchart of the optimization method for air duct pressure control in machining and heat treatment according to the present invention.

[0013] Figure 2 This is a comparison of the optimal fitness convergence curves of the neuron synapse optimization algorithm of this invention and the improved neuron synapse optimization algorithm.

[0014] Figure 3 This is a comparison of the system response curves of the neuron synapse optimization algorithm of this invention and the improved neuron synapse optimization algorithm.

[0015] Figure 4 This is a comparison chart of the PID control parameter optimization curves of the neuron synapse optimization algorithm of this invention and the improved neuron synapse optimization algorithm. Detailed Implementation

[0016] Example 1: As Figure 1 As shown, the present invention provides a method for optimizing the pressure control of air ducts in machining and heat treatment, comprising the following steps: Step S1: Construct a PID control system for air duct pressure in machining and heat treatment, including an air duct pressure error calculation module, an air duct pressure PID controller module, an improved neuron synapse optimization algorithm module, an air duct pressure regulation module, and an air duct pressure monitoring module. The air duct pressure monitoring module collects the actual pressure signal in real time and feeds it back to the error calculation module. It compares the error with the target setpoint to generate a deviation signal and inputs it to the PID controller module. At the same time, the improved neuron synapse optimization algorithm module performs static parameter configuration on the controller based on the parameters obtained by offline optimization of the controlled object model. Then, the PID controller module calculates the precise control increment based on the deviation signal and outputs it to the air duct pressure regulation module to complete the full closed-loop linear regulation of the air duct pressure. Step S2: Introduce an improved neuron synapse optimization algorithm and construct an improved neuron synapse optimization algorithm module: Step S3: Based on the PID control system for air duct pressure in machining and heat treatment, the improved neuron synapse optimization algorithm is used to perform offline iterative calculations in the feature search space to find the optimal control parameters. Step S4: Map the optimal control parameters to the duct pressure PID controller module, and use the optimized parameter combination to achieve better control of the duct pressure in machining and heat treatment.

[0017] Based on the PID control system for air duct pressure in machining and heat treatment, the PID controller outputs signals. The mathematical expression for the time-domain PID with differential filtering is as follows: , In the formula Represents continuous-time variables in the simulation. Indicates time The controller output signal, Indicates time The system pressure deviation value, Represents the proportionality coefficient. Represents the integral coefficient. Represents the integral variable. This represents the differential term output after low-pass filtering; The differential term output Satisfies the following differential equation: , In the formula Represents the differential coefficient. This represents the differential filter coefficient parameter with a value of 0.02, where the differential filter coefficient... The purpose of setting it to a fixed value is to achieve a trade-off between high-frequency noise suppression and system response speed, and to ensure the numerical stability of the differential term. Therefore, it is not used as a variable to be optimized. The corresponding neuron signal strength The formula is: , In the formula This represents the time multiplied by the integral term of the absolute error. This indicates the overshoot penalty. This indicates the adjustment time penalty item. This represents the control energy constraint term. This indicates the weight of the integral term and has a value of 10. This indicates the overshoot penalty weight, with a value of 5000. This indicates that the time penalty weight is adjusted and the value is 500. This indicates that the energy weight is controlled and has a value of 5. Where time is multiplied by the integral term of absolute error The formula is: , In the formula Represents the total number of sampling points and equal to Divide by , Indicates the first The normalized systematic error at each sampling time point Indicates the sampling step size; Overshoot penalty item The formula is: , In the formula This represents the overshoot of the actual system output. This parameter represents the maximum allowable overshoot ratio and has a value of 0.05. Among them, the time adjustment penalty item The formula is: , In the formula This indicates the actual adjustment time of the system output. This indicates the maximum allowable adjustment time parameter, with a value of 12.0; Among them, the energy control term The formula is: , In the formula , , These represent the current proportional, integral, and differential coefficients, respectively. This represents the energy scaling factor parameter and its value is... .

[0018] The step S3, which involves tuning the control parameters of the fluid pressure control system using an improved neuron synapse optimization algorithm, is as follows: Step 1: Perform the chaotic neuron pre-activation initialization step; Step 11: Transfer function model of the PID control system for pressure in the machining and heat treatment air duct. for: , In the formula Represents the Laplace operator. This represents the system gain parameter. Indicates the time constant parameter. This represents the pure time delay parameter of the system, after system identification. =1.5, =2.5, =0.4; Neuron synaptic feature dimensions to be optimized The value is 3 and corresponds to the proportionality coefficient respectively. Integral coefficient and differential coefficients Set the size of the neuron population The value is set to 30, which represents the maximum number of neural remodeling cycles. The value is set to 100, and the upper bound vector of the neuron synaptic feature search space is set to... And the lower bound vector is , =[10,2,1], =[0,0,0]; Step 12: Initialize algorithm control parameters and set chaos parameters. Set the base learning rate to 0.7. The value is set to 0.05, which is the distance attenuation factor. A value of 1.0 is used to set the synaptic pruning threshold. The value is set to 0.2, which is the enhancement factor. The value is set to 0.1, which sets the Levi Flight Index. The value is set to 1.5, which sets the reverse learning probability. The value is 0.3; Step 13: Initialize the synaptic connection strength matrix Generate a OK A matrix of columns, where each element... The initial value is in the interval The initial value is randomly generated within the time frame, and this initial value represents the strength of random connections between neurons at the initial time step. Step 14: Generate the initial neuron population using Tent chaotic mapping. For each neuron, the first The first step is to generate an initial chaotic seed value in the interval (0,1) using a uniformly distributed random function. Then, the Tent mapping formula is used for 50 iterations to eliminate transient effects. The mapping logic is as follows: , In the formula This represents the internal iteration index of the chaotic map. Indicates the first The first neuron 3D features in the 1st dimension State values ​​during the next chaotic evolution This indicates the state value for the next evolution; Take the final value And mapped to the physical search space: , In the formula Indicates the first The first neuron 3D positional components, This represents the corresponding chaotic variable. express The first in the vector One element, express The first in the vector One element; Step 15: Calculate the signal strength of each neuron in the initial population. It identifies the neuron with the lowest current signal strength as the most activated neuron. ; Step 2: Enter the neural remodeling cycle optimization step, setting the current neural remodeling cycle count to [number]. ,when Less than or equal to the maximum number of neural remodeling cycles At that time, the following sub-steps are executed repeatedly: Step 21: Perform signal strength normalization and synaptic plasticity adjustment operations. First, calculate the normalized signal strength. , , In the formula This represents the original signal strength of the current neuron. This represents the minimum signal strength value in the current population. This represents the maximum signal strength value in the current population; Next, calculate the neurons. With neurons Normalized Euclidean distance between , , In the formula Represents neurons The Synaptic characteristics, Represents neurons The Synaptic characteristics, and Indicates the first Dimensional search boundary; The neurons are then updated according to Hebb's learning rule. For neurons Synaptic connection strength , , In the formula This indicates the updated synaptic connection strength. This indicates the synaptic connection strength before the update. Indicates the basic learning rate. Represents neurons The normalized signal strength, Represents neurons The normalized signal strength, Represents the distance attenuation factor. Represents neurons With neurons The normalized Euclidean distance between them; Updated synaptic connection strength Restricted to the range Inside; Step 22: Perform long-range axonal pulse projection location update operation, using the Lévy flight strategy to update neuron locations. , In the formula Indicates the updated neuron location. Indicates the neuron position before the update. Indicates that it comes from the population excluding neurons The sum of the synaptic traction forces of all other neurons besides, This indicates the location of the currently most activated neuron. This represents the Lévy flight step size vector. This represents the scaling factor for Levi's flight step size, with a value of 0.01. Where Levy's flight step vector The formula is: , In the formula Let represent a random variable that follows a normal distribution. Let represent a random variable that follows a standard normal distribution. Indicates the Levi Flight Index; random variable Standard deviation The calculation formula is: , In the formula Represents the standard gamma function. Represents pi; Step 23: Perform the inhibitory signal reflection verification operation. For each neuron, generate a range within the interval... A random number within the range; if the random number is less than the reverse learning probability... Then calculate the reverse synaptic feature solution of the neuron. , , In the formula This indicates the location of the characteristic solution of the reverse synapse. and These represent the lower and upper bound vectors of the search space, respectively. Indicates the neuron position after step 22 update; Step 23 then evaluates the antisynaptic characteristic solution. The signal strength, if The signal strength is better than The signal strength is then used replace ; Step 24: Perform synaptic pruning and strengthening operations, first traversing the synaptic connection strength matrix. All those less than the synaptic pruning threshold Synaptic connection strength Reset to 0; The retained strong synaptic connections were then reinforced by displacement. , In the formula Indicates the final neuron location after enhancement. Indicates the strengthening factor. Represents all neurons Establish a strong connection and satisfy Greater than or equal to the synaptic pruning threshold A collection of neurons; Step 25: Update the neuron population state and evaluate the signal strength of all updated neurons. If a better signal strength value is found, the strongest activated neuron is updated. and its corresponding minimum signal strength ; Step 3: Execute the termination and output steps, and determine the current number of neural remodeling cycles. Has the maximum number of neural remodeling cycles been reached? If so, stop optimization and output the strongest activated neuron. As the optimal PID control parameter for the fluid pressure control system, otherwise let Increase by 1 and return to step 21 to continue execution.

[0019] The improved neuron synapse optimization algorithm and the neuron synapse optimization algorithm were used to perform offline iterative calculations in the search space. Experiments were conducted on the Matlab / Simulink simulation platform, and the simulation time was set. The value is set to 40.0 seconds, which is the simulation step size. The value is set to 0.05 seconds to set the initial system pressure. The value is 500.0, which sets the target value. The value is 800.0, such as Figures 2-4 As shown, experimental results indicate that the optimal fitness value of the neuron synapse optimization algorithm is 4.6465, while the fitness value of the improved neuron synapse optimization algorithm is reduced to 2.4303, improving the optimization accuracy by approximately 47.7%. Furthermore, the improved neuron synapse optimization algorithm achieves smooth convergence after approximately 40 iterations, effectively avoiding the premature convergence phenomenon that occurs around the 20th iteration of the neuron synapse optimization algorithm, and obtaining the optimal control parameters. =2.9491、 =1.1253、 =0.4591, mapped to the duct pressure PID controller module, utilizes the optimized parameter combination to achieve better control of the duct pressure in machining and heat treatment, resulting in a shorter settling time and lower overshoot in the step response of the duct pressure system. Therefore, the improved neuron synapse optimization algorithm improves the neuron synapse optimization algorithm through a multi-dimensional mechanism, thereby improving the control accuracy and system stability of the duct pressure in machining and heat treatment.

Claims

1. A method for optimizing pressure control in air ducts used in machining and heat treatment, characterized in that: Includes the following steps: Step S1: Construct a PID control system for air duct pressure in machining and heat treatment, including an air duct pressure error calculation module, an air duct pressure PID controller module, an improved neuron synapse optimization algorithm module, an air duct pressure regulation module, and an air duct pressure monitoring module. Step S2: Introduce an improved neuron synapse optimization algorithm. The improvement strategy used is as follows: Step S21: In the initialization phase, the Tent chaotic mapping initialization strategy is introduced. The Tent chaotic mapping generates a chaotic sequence with ergodicity and randomness, which improves the uniformity of the initial neuron population in the multidimensional search space. Step S22: In the signal strength normalization and synaptic plasticity adjustment operation, a normalized Euclidean distance calculation strategy is introduced. By mapping the parameter dimensions with different value ranges to a unified scale, the dominant influence of the larger value range on the synaptic connection weight is eliminated, ensuring that the influence of each PID parameter dimension on the synaptic connection strength is balanced. Step S23: In the position update stage, a long-range axon pulse projection position update strategy is introduced. By utilizing the random alternation of long and short step sizes of the Levy flight mechanism, the algorithm's ability to escape local optima in the search space is enhanced and the global search efficiency is improved. Step S24: After the position update, an inhibitory signal reflection verification strategy is introduced to establish a reverse solution exploration mechanism. By probabilistically generating and evaluating the reverse synaptic feature solution of the current neuron position, the effective search range is expanded and the algorithm convergence is accelerated. Step S3: Use the improved neuron synapse optimization algorithm to perform offline iterative calculations in the feature search space to find the optimal control parameters; Step S4: Map the optimal control parameters to the duct pressure PID controller module, and use the optimized parameter combination to achieve better control of the duct pressure in machining and heat treatment.

2. The method for optimizing air duct pressure control in machining and heat treatment according to claim 1, characterized in that: In the mechanical processing and heat treatment air duct pressure PID control system constructed in step S1, the air duct pressure monitoring module collects the actual pressure signal in real time and feeds it back to the error calculation module. It compares the signal with the target setpoint to generate a deviation signal and inputs it to the PID controller module. At the same time, the improved neuron synapse optimization algorithm module performs static parameter configuration on the controller based on the parameters obtained by offline optimization of the controlled object model. Then, the PID controller module calculates the precise control increment based on the deviation signal and outputs it to the air duct pressure regulation module to complete the full closed-loop linear regulation of the air duct pressure.

3. The method for optimizing air duct pressure control in machining and heat treatment according to claim 1, characterized in that, The Tent chaotic mapping initialization strategy in step S21 includes the following steps: Step S211: For each neuron The Dimensional features, generating initial chaotic sequence values ; Step S212: Iteratively update according to the Tent mapping rules The mapping logic is as follows: ， In the formula This represents the internal iteration index of the chaotic map. Indicates the first The first neuron 3D features in the 1st dimension State values ​​during the next chaotic evolution. This indicates the state value for the next evolution; Step S213: Generate chaotic values Mapped to the physical search space, ， In the formula Indicates the first The first neuron 3D positional components, express The first in the vector One element, express The first in the vector One element, This represents the corresponding chaotic variable.

4. The method for optimizing air duct pressure control in machining and heat treatment according to claim 1, characterized in that, The normalized Euclidean distance calculation strategy in step S22 includes the following steps: Step S221, determine the first The search space for 3D neuron synaptic features, i.e., the range of the 3D neuron synaptic features. Upper bound of 3D search space and the lower world The difference; Step S222: Calculate neurons With neurons In the Normalized position difference components on the dimension , ， In the formula Represents the normalized position difference components. Represents neurons In the Positional components of the dimension, Represents neurons In the Positional components of the dimension, Indicates the first Upper bound of the 3D search space Indicates the first Lower bound of the dimensional search space; Step S223: Based on the normalized position difference components of all dimensions Calculate the normalized Euclidean distance This distance is then substituted into the Hebb learning rule formula to update the synaptic connection strength. ， In the formula This represents the normalized Euclidean distance. Indicates the PID parameter dimension. This indicates the result calculated in step S222. Normalized position difference components; Next, update the synaptic connection strength. ， In the formula This indicates the updated synaptic connection strength. Indicates the current synaptic connection strength. Indicates the basic learning rate. Represents neurons The normalized signal strength, Represents neurons The normalized signal strength, This represents the distance attenuation factor.

5. The method for optimizing air duct pressure control in machining and heat treatment according to claim 1, characterized in that, The long-range axon pulse projection position update strategy in step S23 includes the following steps: Step S231: Calculate the Levy flight step vector , ， In the formula This represents the Lévy flight step size vector. This indicates that the standard deviation is 1. A normally distributed random variable, Let represent a random variable that follows a normal distribution with a standard deviation of 1. The Levi Flight Index indicates that... It is calculated from a formula that includes the gamma function; Step S232: Introduce the Levy flight step scaling factor. Combining the sum of synaptic traction forces with the current optimal position The guiding role; Step S233: Perform position update calculation. ， In the formula Indicates the updated neuron location. Indicates the neuron position before the update. Indicates that it comes from the population excluding neurons The sum of synaptic traction forces of all other neurons, This represents the scaling factor for Levy's flight step size. This indicates the location of the currently most activated neuron.

6. The method for optimizing air duct pressure control in machining and heat treatment according to claim 1, characterized in that, The suppression signal reflection verification strategy in step S24 includes the following steps: Step S241: Generate an interval Random numbers within, and inverse learning probabilities Compare; Step S242: If the condition is met, calculate the reverse synaptic feature solution for the current neuron position. , ， In the formula This represents the characteristic solution of the reverse synapse. and Let these represent the lower and upper bound vectors of the search space, respectively. Indicates the neuron's position after the position update; Step S243: Evaluate the antisynaptic feature solution The signal strength, if The signal strength is better than at the original location The signal strength is then used replace .

7. The method for optimizing air duct pressure control in machining and heat treatment according to claim 1, characterized in that, Step S3, which involves tuning the PID control parameters of the air duct pressure in the PID control system for machining heat treatment using an improved neural network synaptic optimization algorithm, includes the following detailed steps: Step S31: Initialize system model and algorithm parameters: Establish a mathematical model of the duct pressure control system and set the dimensions of the PID parameters to be optimized. Set the size of the neuron population Set the maximum number of neural remodeling cycles. And setting the search space boundaries for PID parameters; Step S32: Initialize the synaptic connection strength matrix For elements in the matrix Assign initial values; Step S33: Execute the Tent chaotic mapping initialization strategy as described in claim 3 to generate an initial neuron population. The signal strength of the initial population was assessed to identify the most strongly activated neurons. ; Step S34: Enter the iterative optimization process, and execute the following sub-steps in each iteration: Sub-step S341: Calculate the normalized signal intensity of each neuron in the current population; Sub-step S342: Execute the normalized Euclidean distance calculation strategy as described in claim 4, and update the synaptic connection strength matrix according to the Heb learning rule. ; Sub-step S343: Execute the long-range axonal pulse projection position update strategy as described in claim 5 to update the neuron position; Sub-step S344: Execute the inhibitory signal reflection verification strategy as described in claim 6, and perform reverse learning optimization on the updated position; Sub-step S345: Perform synaptic pruning and reinforcement operations, removing synaptic pruning elements below the synaptic pruning threshold. Weak connections are established, and strong connections are reinforced by displacement. Sub-step S346: Evaluate the updated neuron population and update the globally most activated neurons. ; Step S35: Determine whether the current iteration count has reached the maximum number of neural remodeling cycles. If the condition is met, the iteration terminates and the result is output. If the optimal PID control parameters are used, then proceed to step S34.

Images