Production temperature control method for wire drawing machine

Abstract

The invention belongs to the technical field of intelligent control, and particularly relates to a wire drawing machine production temperature control method. The method comprises the following steps: acquiring multi-temperature-zone temperature data of the wire drawing machine and determining a temperature control set range; coupling interference in the temperature zones is processed through an improved neural network decoupling algorithm, and independent controlled temperature values of all the temperature zones are obtained; an orthogonal mode PID controller is constructed, a unique mapping relation between PID three parameters and an orthogonal mode is established, a single-mode core characteristic quantity is extracted through adaptive orthogonal mode decomposition, and three-parameter non-coupling adaptive setting is completed; and a sparrow search algorithm of an improved elite retention variation mechanism is adopted, global optimization is carried out on control parameters to obtain an optimal parameter set, and real-time temperature control of the temperature zone is achieved. The problems of multi-temperature-zone coupling interference and PID parameter setting difficulty of the wire drawing machine are effectively solved, the temperature control precision, the production stability and the product forming quality are remarkably improved, and the high-precision temperature control requirement of high-end wire drawing production is met.

Description

Technical Field

[0001] This invention belongs to the field of intelligent control technology, and in particular relates to a method for controlling the production temperature of a wire drawing machine. Background Technology

[0002] Wire drawing machines are core equipment in the production of metal wires and polymer materials. The temperature control accuracy of their multi-zone heating system directly determines the forming quality, mechanical properties, and production stability of the product. High-precision temperature control of each zone has become a core requirement for high-end wire drawing production processes. Therefore, independent temperature adjustment is required for each zone, making precise and uncoupled control of multi-zone temperatures a key aspect of the development of wire drawing machine temperature control systems. Summary of the Invention

[0003] In view of the technical problems existing in the background art, the present invention proposes a method for controlling the production temperature of a wire drawing machine.

[0004] To achieve the above objectives, the technical solution adopted by the present invention includes the following steps:

[0005] S1. Obtain the temperature data of the target wire drawing machine in multiple temperature zones and determine the setting range of temperature control for each temperature zone;

[0006] S2. Based on temperature data from multiple temperature zones, the coupling interference between each temperature zone is decoupled by an improved neural network decoupling algorithm to obtain independent controlled temperature values ​​for each temperature zone.

[0007] S3. For the independently controlled temperature values ​​of each temperature zone after decoupling, construct an orthogonal modal PID controller with one-to-one binding of modal parameters, and establish a unique mapping relationship between the three PID parameters and the orthogonal modes to achieve decoupling of control parameters;

[0008] S4. Perform adaptive orthogonal mode decomposition on the real-time temperature response data of each temperature zone, and extract the single-mode core feature quantities that are bound to each PID parameter based on the temperature setpoint and real-time output data.

[0009] S5. Based on the extracted core features of the corresponding modes, complete the uncoupled adaptive tuning of the three parameters of proportional, integral, and derivative, and obtain the improved temperature controller.

[0010] S6. Using the sparrow search algorithm and an improved elite retention mutation mechanism, the control parameters of the improved temperature controller are globally optimized to obtain the optimal set of control parameters, and the corresponding temperature zone of the wire drawing machine is controlled in real time.

[0011] Preferably, in step S2, based on the temperature data from multiple temperature zones, the coupling interference between each temperature zone is decoupled using an improved neural network decoupling algorithm to obtain independent controlled temperature values ​​for each temperature zone. The specific implementation of this method is as follows:

[0012] S21. Set the temperature control system of the wire drawing machine to include n independent heating temperature zones. Take the heating control quantity of each temperature zone as the system input and the real-time temperature sampling value of each temperature zone as the system output. Construct a multi-input multi-output system model in the discrete domain that reflects the temperature coupling characteristics between temperature zones.

[0013] S22. Construct a decoupled neural network consisting of an input layer, a hidden layer, and a decoupled output layer. The input layer has 2n nodes, and the input consists of the real-time temperature sampling values ​​of each temperature zone and the temperature sampling values ​​of the previous time step. The hidden layer adopts a sparse connection structure with temperature coupling prior, retaining only the connections between corresponding nodes in adjacent temperature zones. The connection weights between corresponding nodes in non-adjacent temperature zones are fixed at 0 and do not participate in training updates. The nonlinear characteristics of temperature coupling are fitted by a hyperbolic tangent activation function, and the hidden layer vector is output. The decoupled output layer has n nodes. First, a strict diagonalized linear mapping is established from the hidden layer output to the output layer. Only the non-zero weights between the i-th node of the hidden layer and the i-th node of the output layer are retained, and the weights of the other cross-node connections are fixed at 0. Then, the linear mapping result is subjected to amplitude-limited saturation activation processing with upper and lower limits, and finally the decoupled compensation control quantity of each temperature zone is output.

[0014] S23. Define a relative decoupling index for single-temperature zones in response to temperature coupling, and construct a loss function for the decoupled neural network based on this relative decoupling index.

[0015] S24. Collect temperature sample sets of multiple temperature zones under all working conditions during the historical production process of the wire drawing machine as training data, pre-train the constructed decoupled neural network until the loss function converges to the preset convergence threshold or reaches the set maximum number of training times, complete the pre-training and obtain the pre-trained decoupled neural network model.

[0016] S25. The decoupling compensation control quantities of each temperature zone output in real time from the decoupling neural network are superimposed with the heating control quantity input of the corresponding temperature zone according to the one-to-one correspondence of the temperature zones to generate the total control input that is finally input to the heating actuator of each temperature zone of the wire drawing machine. The total control input is substituted into the aforementioned temperature coupling system model, and the transfer function matrix is ​​decomposed into diagonal inherent characteristic terms and off-diagonal coupling characteristic terms. The compensation terms and coupling interference terms are completely canceled by the decoupling compensation control quantities. The final output of the total control input of the heating actuator of each temperature zone is the independent controlled temperature setpoint of each temperature zone.

[0017] Preferably, in step S3, for the independently controlled temperature values ​​of each decoupled temperature zone, an orthogonal modal PID controller with one-to-one binding of modal parameters is constructed, and a unique mapping relationship between the three PID parameters and the orthogonal modes is established to achieve the decoupling of control parameters. Specifically, this is implemented as follows:

[0018] S31. For the i-th temperature zone after decoupling, use the independent controlled temperature setpoint of that temperature zone as the target value. The output value is the real-time temperature sampled at time k in this temperature range. Construct a single-input single-output closed-loop control loop for this temperature zone, where k is the discrete sampling time, i is the temperature zone number, and n is the total number of independent heating temperature zones of the wire drawing machine;

[0019] S32. Establish a unique mapping relationship between the three parameters of PID (proportional, integral, and derivative) and the orthogonal mode, and output the real-time temperature at time k in the i-th temperature zone. The unique decomposition is the sum of three pairwise orthogonal modal components. Orthogonality is strictly constrained by setting the vector inner product to zero within a sliding time window of length M. The decomposition formula in the discrete domain is: ,and ,in, This refers to the vector dot product operation within the sliding time window. These are the temperature sequence vectors for the reference tracking mode, oscillation suppression mode, and drift compensation mode within the sliding window, respectively.

[0020] Preferably, in step S4, adaptive orthogonal mode decomposition is performed on the real-time temperature response data of each temperature zone. Based on the temperature setpoint and real-time output data, the single-mode core feature quantities that are bound to each PID parameter are extracted as follows:

[0021] S41. For the i-th temperature region, set a sliding time window of length M, and construct the temperature response data vector within the window: Temperature target value vector: ;

[0022] S42. Using the temperature target value trend as a benchmark, the benchmark tracking mode vector is obtained through least squares orthogonal projection. The core feature quantity of the baseline tracking mode is extracted as the baseline tracking error: ;

[0023] S43, Regarding the residual vector The alternating oscillation components are extracted by high-pass orthogonal projection to obtain the oscillation suppression mode vector. The core characteristic quantity is the oscillation mode energy. ;

[0024] S44, For the final residual vector The core feature quantity is the cumulative amount of drift modes. , where m is the loop variable for the summation operation.

[0025] Preferably, step S5, based on the extracted corresponding modal core feature quantities, completes the uncoupled adaptive tuning of the proportional, integral, and derivative parameters, which are completely independent, to obtain the specific implementation of the improved temperature controller, including: firstly, calculating the temperature tracking deviation rate: The proportional adaptive tuning is updated using a piecewise function as follows: ,in, As the base value of the proportional coefficient, the integral adaptive tuning is updated as follows: ,in These are the base values ​​for the integral coefficients. This is the maximum calibrated value of the cumulative drift compensation mode under rated operating conditions in the i-th heating temperature zone. To prevent integral saturation, the controller output is set to 1 when it does not reach the upper limit of the control quantity of the wire drawing machine heating actuator; and to 0 when it does. The adaptive differential tuning is then updated as follows: ,in, These are the basic values ​​of the differential coefficients. To obtain the maximum value of the current oscillation mode energy and the minimum threshold, and to ensure that the denominator is always a non-zero valid value, The minimum threshold for oscillation mode energy is set; finally, the output of the improved temperature controller is: .

[0026] Preferably, the improved elite retention mutation mechanism in step S6 involves sorting the individuals in the population according to their fitness values ​​and dividing them into three levels: elite individuals, ordinary individuals, and inferior individuals. Differentiated retention and mutation rules are executed for the characteristics of individuals in different levels. After each iteration, the elite retention is first executed to lock the high-quality parameter genes that are suitable for the temperature control of the wire drawing machine. Then, the differentiated mutation operation is executed according to the level to update the population. After iterating to the preset termination condition, the optimal control parameter set is output to control the corresponding temperature zone of the wire drawing machine in real time.

[0027] As a preferred approach, the specific rules for hierarchical division and differentiated mutation are as follows: the top 10% of individuals with fitness values ​​after each iteration are classified as elite individuals, and these elite individuals are directly retained in their entirety for the next iteration without any mutation operations; the middle 30% to 70% of individuals are classified as ordinary individuals, and mild single-parameter random mutation is performed on them, randomly adjusting only one of the three parameters of PID, thereby enhancing the accuracy of local optimization while preserving the high-quality genes of the main body; the bottom 20% of individuals are classified as inferior individuals, and full-parameter reset mutation is performed on them to completely regenerate new individuals that meet the parameter stability boundary.

[0028] Compared with existing technologies, the advantages and positive effects of this invention lie in its ability to precisely eliminate coupling interference between multiple temperature zones in a wire drawing machine by improving the neural network decoupling algorithm, achieving independent temperature control for each zone and fundamentally solving the problem of mutual influence between temperature zones. The use of an orthogonal modal PID controller uniquely binds the three PID parameters to the temperature control mode, enabling uncoupled adaptive tuning and significantly improving temperature tracking accuracy and oscillation suppression capabilities. Combined with an improved sparrow search algorithm with an elite-preserving mutation mechanism, global optimization of control parameters is achieved, balancing optimization efficiency and parameter stability. This solution effectively solves the problems of low temperature control accuracy, difficult parameter tuning, and large coupling interference in traditional wire drawing machines, significantly improving production stability and product forming quality, and meeting the high-precision temperature control requirements of high-end wire drawing production. Attached Figure Description

[0029] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the following description of the embodiments will be briefly introduced. Obviously, the drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0030] Figure 1 This is a schematic diagram of the structural process of a temperature control method for wire drawing machine production. Detailed Implementation

[0031] To better understand the above-mentioned objectives, features, and advantages of the present invention, the present invention will be further described below in conjunction with the accompanying drawings and embodiments. It should be noted that, unless otherwise specified, the embodiments and features described in these embodiments can be combined with each other.

[0032] Numerous specific details are set forth in the following description in order to provide a full understanding of the invention. However, the invention may also be practiced in other ways than those described herein, and therefore the invention is not limited to the specific embodiments disclosed in the following specification.

[0033] This embodiment proposes a method for controlling the production temperature of a wire drawing machine. For example... Figure 1 As shown, the specific implementation is as follows:

[0034] First, obtain the temperature data of the target wire drawing machine in multiple temperature zones to determine the setting range of temperature control for each zone.

[0035] Then, based on the temperature data of multiple temperature zones, the coupling interference between each temperature zone is decoupled by improving the neural network decoupling algorithm, so as to obtain the independent controlled temperature value of each temperature zone.

[0036] Specifically, we first define the temperature control system of the wire drawing machine as having n independent heating zones. We then use the heating control parameters of each zone as system inputs and the real-time temperature sampling values ​​of each zone as system outputs to construct a multi-input multi-output system model in the discrete domain that reflects the temperature coupling characteristics between the zones. Where k is the discrete sampling time, , where is the temperature output vector. Let be the real-time temperature sample value of the i-th temperature zone at time k. Here is the input vector for the heating control quantity, where Let be the heating control quantity for the i-th temperature zone at time k. For the transfer function matrix, For the shift operator, each element in the transfer function matrix is ​​the coupling transfer function of the heating control quantity of the j-th temperature zone to the temperature output of the i-th temperature zone, where the diagonal elements are the self-transfer functions of the i-th temperature zone. Let be the temperature perturbation vector, where Let be the ambient temperature disturbance value of the i-th temperature zone at time k.

[0037] A decoupled neural network is constructed, consisting of an input layer, a hidden layer, and a decoupled output layer. The input layer has a total of 2n nodes, and the input vector contains the real-time temperature sampling values ​​of each temperature zone and the temperature sampling value of the previous moment. The input vector is: The hidden layer employs a sparse connection structure with temperature coupling prior, retaining only connections between nodes corresponding to adjacent temperature zones. Connection weights between nodes corresponding to non-adjacent temperature zones are fixed at 0 and do not participate in training updates. The output of the hidden layer is calculated as follows: ,in, Let k be the output vector of the hidden layer at time k. This is a hyperbolic tangent activation function used to fit the nonlinear characteristics of temperature coupling. The weight matrix from the input layer to the hidden layer satisfies the temperature coupling sparsity constraint: the weight value is non-zero only when the temperature region corresponding to the node is an adjacent temperature region; all other elements are fixed to 0. Here is the recursive weight matrix of the hidden layer. This is the bias vector of the hidden layer. The total number of nodes in the decoupled output layer is n, and the output is the decoupling compensation control quantity for each temperature zone. .

[0038] Furthermore, the output vector from the hidden layer Decoupling compensation control quantity for each temperature zone The calculation is as follows: Establish a linear mapping relationship from the hidden layer output to the output layer, and the calculation formula is: ,in, Let be the connection weight matrix from the hidden layer to the output layer. It is an n×n strictly diagonal matrix, with only diagonal elements being trainable non-zero weights, corresponding to the connection weights from the i-th node in the hidden layer to the i-th node in the output layer; all off-diagonal elements are fixed to 0. This is the bias vector for the output layer.

[0039] The linear mapping result is subjected to amplitude limiting activation processing to obtain the final decoupling compensation control quantity. ,in, For the output layer's band-limited saturated linear activation function, the calculation rule for a single node is as follows:

[0040] ,in, , These are the upper and lower limits of the decoupling compensation control quantity, respectively, ultimately yielding the final output decoupling compensation control quantity at time k. .

[0041] Define a single-temperature-zone relative decoupling index for temperature coupling: ,in, This represents the relative decoupling degree of the i-th temperature zone. A larger value indicates that the temperature of the i-th temperature zone is less affected by the coupling interference from other temperature zones. Let be the discrete transfer function of the self-control channel in the i-th temperature zone. Let j be the discrete transfer function of the cross-coupling channel from temperature zone j to temperature zone i; and let the loss function be constructed based on the relative decoupling index. The temperature sample set of multiple temperature zones under all working conditions during the historical production process of the wire drawing machine is collected as training data. The constructed decoupled neural network is pre-trained until the loss function J converges to the preset convergence threshold or reaches the set maximum number of times to complete the pre-training, and the pre-trained decoupled neural network model is obtained.

[0042] Finally, the decoupling compensation control quantities of each temperature zone output in real time from the decoupled neural network are... With heating control input The final control input is generated by superimposing the temperature zones one by one and then inputting it to the heating actuators of each temperature zone of the wire drawing machine. ,Will Substituting into the temperature-coupled system model, the transfer function matrix is ​​decomposed into diagonal and off-diagonal coupling terms: ,in, It is a diagonal matrix with diagonal lines. It is a non-diagonal matrix, and the control quantity is compensated through decoupling. The compensation effect causes the compensation term and the coupling interference term to satisfy the following cancellation relationship: The output is the total control input for the heating actuators in each temperature zone. These are independent controlled temperature values ​​for each temperature zone, i.e., temperature setpoints.

[0043] Next, for the independently controlled temperature values ​​of each temperature zone after decoupling, an orthogonal modal PID controller with one-to-one binding of modal parameters is constructed, and a unique mapping relationship between the three PID parameters and the orthogonal modes is established to achieve decoupling of control parameters.

[0044] Specifically, for the i-th temperature zone after decoupling, the target value is the independent controlled temperature setpoint of that temperature zone. The output value is the real-time temperature sampled at time k in this temperature range. A single-input, single-output closed-loop control loop is constructed for this temperature zone, where k is the discrete sampling time, i is the temperature zone number, and n is the total number of independent heating temperature zones in the wire drawing machine. A unique mapping relationship is established between the PID proportional, integral, and derivative parameters and the orthogonal mode, and the real-time temperature output of the i-th temperature zone at time k is recorded. The unique decomposition is the sum of three pairwise orthogonal modal components. Orthogonality is strictly constrained by setting the vector inner product to zero within a sliding time window of length M. The decomposition formula in the discrete domain is: ,and ,in, This refers to the vector dot product operation within the sliding time window. These are the temperature sequence vectors for the reference tracking mode, oscillation suppression mode, and drift compensation mode within the sliding window, respectively. The baseline is used to track the modal components, which correspond to the proportional gain only. The core control objective is to achieve trend tracking of the temperature range towards the target value. To suppress the modal components of oscillation, the core control objective corresponding to the differential gain is to suppress the alternating oscillations and overshoot of the temperature dynamic process. To compensate for drift modal components, a unique integral gain is assigned, with the core control objective being to eliminate steady-state temperature deviations and slowly varying disturbances. Based on this unique mapping and binding relationship, the three PID parameters are naturally decoupled; adjustments to each parameter only affect the corresponding bound modal component, without interfering with other modal components.

[0045] Furthermore, adaptive orthogonal mode decomposition is performed on the real-time temperature response data of each temperature zone. Based on the temperature setpoint and real-time output data, single-mode core feature quantities that are bound to each PID parameter are extracted.

[0046] Specifically, for the i-th temperature zone, a sliding time window of length M is set, and a temperature response data vector within the window is constructed: Temperature target value vector: Based on the trend of the target temperature value, the reference tracking mode vector is obtained through least squares orthogonal projection. The core feature quantity of the baseline tracking mode is extracted as the baseline tracking error: For the residual vector The alternating oscillation components are extracted by high-pass orthogonal projection to obtain the oscillation suppression mode vector. The core characteristic quantity is the oscillation mode energy. ; for the final residual vector The core feature quantity is the cumulative amount of drift modes. , where m is the loop variable for the summation operation. In this embodiment, all the aforementioned feature data (including residual vector, oscillation suppression mode vector, oscillation mode energy, etc.) are pre-normalized to convert the original sensor data with physical dimensions into dimensionless standardized values. The control logic makes decisions based only on the relative differences and trends of the dimensionless data, eliminating the problem of inconsistent dimensions and ensuring the universality and robustness of the algorithm.

[0047] Then, based on the extracted core features of the corresponding modes, the proportional, integral, and derivative parameters are completely independent and uncoupled adaptively tuned to obtain an improved temperature controller. Specifically, the temperature tracking error rate is first calculated: The proportional adaptive tuning is updated using a piecewise function as follows: ,in, As the base value of the proportional coefficient, the integral adaptive tuning is updated as follows: ,in These are the base values ​​for the integral coefficients. This is the maximum calibrated value of the cumulative drift compensation mode under rated operating conditions in the i-th heating temperature zone. To prevent integral saturation, the controller output is set to 1 when it does not reach the upper limit of the control quantity of the wire drawing machine heating actuator; and to 0 when it does. The adaptive differential tuning is then updated as follows: ,in, These are the basic values ​​of the differential coefficients. To obtain the maximum value of the current oscillation mode energy and the minimum threshold, and to ensure that the denominator is always a non-zero valid value, The minimum threshold for oscillation mode energy is set; finally, the output of the improved temperature controller is: .

[0048] Finally, the sparrow search algorithm and the improved elite preservation mutation mechanism are used to globally optimize the control parameters of the improved temperature controller, obtain the optimal control parameter set, and perform real-time control of the corresponding temperature zone of the wire drawing machine.

[0049] Specifically, the Sparrow Search Algorithm (SSA) is a novel swarm intelligence optimization algorithm. Its core inspiration comes from the cooperative behavior of sparrows in their natural foraging and anti-predation. Through biomimetic group role division and dynamic interaction, it achieves efficient solutions to complex optimization problems.

[0050] This invention uses the proportional, integral, and derivative parameters of an improved temperature PID controller as optimization variables. Combined with the engineering constraints of the heating actuator in the temperature zone of a wire drawing machine, the physical stability boundary of the three parameters is first determined (the parameter values ​​all meet the control input range of the temperature controller actuator, and there are no invalid parameters). Then, a random uniform initialization method is used to generate an initial population. Each individual in the population is a three-dimensional parameter vector. The population size is set according to the optimization requirements of the multi-temperature zone temperature control of the wire drawing machine, ensuring that the initial individuals uniformly cover the parameter solution space and guarantee the initial foundation for global optimization.

[0051] After population initialization, this invention enters an iterative evolution process, deeply integrating the improved elite retention mutation mechanism into each iteration cycle of the sparrow search algorithm. The iterations are structured around a core closed-loop process: fitness value calculation and sorting, population hierarchy division, elite retention locking, hierarchical differential mutation, population update, and termination condition judgment. Fitness value calculation is a prerequisite for each iteration, using a constructed fitness function to quantitatively evaluate the parameter combinations of each individual in the population. Population hierarchy division and differential mutation are the core of this invention's improved mechanism. After each iteration, individuals are sorted by fitness value and divided into three levels: elite individuals, ordinary individuals, and inferior individuals. Differentiated retention and mutation rules are applied based on the characteristics of individuals at different levels. After each iteration, elite retention is performed to lock high-quality parameter genes suitable for the temperature control of the wire drawing machine. Then, differential mutation operations are performed hierarchically to update the population. After iterating to the preset termination condition, the optimal control parameter set is output for real-time regulation of the corresponding temperature zone of the wire drawing machine.

[0052] Furthermore, the specific rules for hierarchical division and differentiated mutation are as follows: the top 10% of individuals in terms of fitness value after each iteration are classified as elite individuals, and these elite individuals are directly retained in their entirety for the next iteration without any mutation operation; the middle 30% to 70% of individuals are classified as ordinary individuals, and mild single-parameter random mutation is performed on them, randomly adjusting only one of the three parameters of PID, thereby enhancing the accuracy of local optimization while preserving the high-quality genes of the main body; the bottom 20% of individuals are classified as inferior individuals, and full-parameter reset mutation is performed on them to completely regenerate new individuals that meet the parameter stability boundary.

[0053] The above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention in any other way. Any person skilled in the art may make changes or modifications to the above-disclosed technical content to create equivalent embodiments for application in other fields. However, any simple modifications, equivalent changes, and modifications made to the above embodiments based on the technical essence of the present invention without departing from the scope of the present invention shall still fall within the protection scope of the present invention.

Claims

1. A method for controlling the production temperature of a wire drawing machine, characterized in that, Includes the following steps: S1. Obtain the temperature data of the target wire drawing machine in multiple temperature zones and determine the setting range of temperature control for each temperature zone; S2. Based on temperature data from multiple temperature zones, the coupling interference between each temperature zone is decoupled by an improved neural network decoupling algorithm to obtain independent controlled temperature values ​​for each temperature zone. S3. For the independently controlled temperature values ​​of each temperature zone after decoupling, construct an orthogonal modal PID controller with one-to-one binding of modal parameters, and establish a unique mapping relationship between the three PID parameters and the orthogonal modes to achieve decoupling of control parameters; S4. Perform adaptive orthogonal mode decomposition on the real-time temperature response data of each temperature zone, and extract the single-mode core feature quantities that are bound to each PID parameter based on the temperature setpoint and real-time output data. S5. Based on the extracted core features of the corresponding modes, complete the uncoupled adaptive tuning of the three parameters of proportional, integral, and derivative, and obtain the improved temperature controller. S6. Using the sparrow search algorithm and an improved elite retention mutation mechanism, the control parameters of the improved temperature controller are globally optimized to obtain the optimal set of control parameters, and the corresponding temperature zone of the wire drawing machine is controlled in real time.

2. The method for controlling the production temperature of a wire drawing machine according to claim 1, characterized in that, In step S2, based on the temperature data from multiple temperature zones, the coupling interference between each temperature zone is decoupled using an improved neural network decoupling algorithm to obtain independent controlled temperature values ​​for each zone. The specific implementation of this method is as follows: S21. Set the temperature control system of the wire drawing machine to include n independent heating temperature zones. Take the heating control quantity of each temperature zone as the system input and the real-time temperature sampling value of each temperature zone as the system output. Construct a multi-input multi-output system model in the discrete domain that reflects the temperature coupling characteristics between temperature zones. S22. Construct a decoupled neural network consisting of an input layer, a hidden layer, and a decoupled output layer. The input layer has 2n nodes, and the input consists of the real-time temperature sampling values ​​of each temperature zone and the temperature sampling values ​​of the previous time step. The hidden layer adopts a sparse connection structure with temperature coupling prior, retaining only the connections between corresponding nodes in adjacent temperature zones. The connection weights between corresponding nodes in non-adjacent temperature zones are fixed at 0 and do not participate in training updates. The nonlinear characteristics of temperature coupling are fitted by a hyperbolic tangent activation function, and the hidden layer vector is output. The decoupled output layer has n nodes. First, a strict diagonalized linear mapping is established from the hidden layer output to the output layer. Only the non-zero weights between the i-th node of the hidden layer and the i-th node of the output layer are retained, and the weights of the other cross-node connections are fixed at 0. Then, the linear mapping result is subjected to amplitude-limited saturation activation processing with upper and lower limits, and finally the decoupled compensation control quantity of each temperature zone is output. S23. Define a relative decoupling index for single-temperature zones in response to temperature coupling, and construct a loss function for the decoupled neural network based on this relative decoupling index. S24. Collect temperature sample sets of multiple temperature zones under all working conditions during the historical production process of the wire drawing machine as training data, pre-train the constructed decoupled neural network until the loss function converges to the preset convergence threshold or reaches the set maximum number of training times, complete the pre-training and obtain the pre-trained decoupled neural network model. S25. The decoupling compensation control quantities of each temperature zone output in real time from the decoupling neural network are superimposed with the heating control quantity input of the corresponding temperature zone according to the one-to-one correspondence of the temperature zones to generate the total control input that is finally input to the heating actuator of each temperature zone of the wire drawing machine. The total control input is substituted into the aforementioned temperature coupling system model, and the transfer function matrix is ​​decomposed into diagonal inherent characteristic terms and off-diagonal coupling characteristic terms. The compensation terms and coupling interference terms are completely canceled by the decoupling compensation control quantities. The final output of the total control input of the heating actuator of each temperature zone is the independent controlled temperature setpoint of each temperature zone.

3. The method for controlling the production temperature of a wire drawing machine according to claim 1, characterized in that, In step S3, for the independently controlled temperature values ​​of each temperature zone after decoupling, an orthogonal modal PID controller with one-to-one binding of modal parameters is constructed, and a unique mapping relationship between the three PID parameters and the orthogonal modes is established to achieve the decoupling of control parameters. The specific implementation is as follows: S31. For the i-th temperature zone after decoupling, use the independent controlled temperature setpoint of that temperature zone as the target value. The output value is the real-time temperature sampled at time k in this temperature range. Construct a single-input single-output closed-loop control loop for this temperature zone, where k is the discrete sampling time, i is the temperature zone number, and n is the total number of independent heating temperature zones of the wire drawing machine; S32. Establish a unique mapping relationship between the three parameters of PID (proportional, integral, and derivative) and the orthogonal mode, and output the real-time temperature at time k in the i-th temperature zone. The unique decomposition is the sum of three pairwise orthogonal modal components. Orthogonality is strictly constrained by setting the vector inner product to zero within a sliding time window of length M. The decomposition formula in the discrete domain is: ,and ,in, This refers to the vector dot product operation within the sliding time window. These are the temperature sequence vectors for the reference tracking mode, oscillation suppression mode, and drift compensation mode within the sliding window, respectively.

4. The method for controlling the production temperature of a wire drawing machine according to claim 1, characterized in that, In step S4, adaptive orthogonal mode decomposition is performed on the real-time temperature response data of each temperature zone. Based on the temperature setpoint and real-time output data, the core single-mode feature quantities that are bound to each PID parameter are extracted as follows: S41. For the i-th temperature region, set a sliding time window of length M, and construct the temperature response data vector within the window: Temperature target value vector: ; S42. Using the temperature target value trend as a benchmark, the benchmark tracking mode vector is obtained through least squares orthogonal projection. The core feature quantity of the baseline tracking mode is extracted as the baseline tracking error: ; S43, Regarding the residual vector The alternating oscillation components are extracted by high-pass orthogonal projection to obtain the oscillation suppression mode vector. The core characteristic quantity is the oscillation mode energy. ; S44, For the final residual vector The core feature quantity is the cumulative amount of drift modes. , where m is the loop variable for the summation operation.

5. The method for controlling the production temperature of a wire drawing machine according to claim 4, characterized in that, In step S5, based on the extracted core features of the corresponding mode, the proportional, integral, and derivative parameters are completely independent and uncoupled adaptively tuned. The specific implementation of the improved temperature controller includes: first, calculating the temperature tracking deviation rate: The proportional adaptive tuning is updated using a piecewise function as follows: ,in, As the base value of the proportional coefficient, the integral adaptive tuning is updated as follows: ,in These are the base values ​​for the integral coefficients. This is the maximum calibrated value of the cumulative drift compensation mode under rated operating conditions in the i-th heating temperature zone. To prevent integral saturation, the controller output is set to 1 when it does not reach the upper limit of the control quantity of the wire drawing machine heating actuator; and to 0 when it does. The adaptive differential tuning is then updated as follows: ,in, These are the basic values ​​of the differential coefficients. To obtain the maximum value of the current oscillation mode energy and the minimum threshold, and to ensure that the denominator is always a non-zero valid value, The minimum threshold for oscillation mode energy is set; finally, the output of the improved temperature controller is: .

6. The method for controlling the production temperature of a wire drawing machine according to claim 1, characterized in that, In step S6, the improved elite retention and mutation mechanism is to sort the individuals in the population according to their fitness values ​​and divide them into three levels: elite individuals, ordinary individuals, and inferior individuals. Differentiated retention and mutation rules are executed for the characteristics of individuals in different levels. After each iteration, the elite retention is first executed to lock the high-quality parameter genes that are suitable for the temperature control of the wire drawing machine. Then, the differentiated mutation operation is executed according to the level to update the population. After iterating to the preset termination condition, the optimal control parameter set is output to control the corresponding temperature zone of the wire drawing machine in real time.

7. The method for controlling the production temperature of a wire drawing machine according to claim 6, characterized in that, The specific rules for hierarchical division and differentiated mutation are as follows: the top 10% of individuals with fitness values ​​after each iteration are classified as elite individuals, and these elite individuals are directly retained in their entirety for the next iteration without any mutation operations; the middle 30% to 70% of individuals are classified as ordinary individuals, and mild single-parameter random mutation is performed on them, randomly adjusting only one of the three parameters of PID, thereby enhancing the accuracy of local optimization while preserving the main high-quality genes; the bottom 20% of individuals are classified as inferior individuals, and full-parameter reset mutation is performed on them to completely regenerate new individuals that meet the parameter stability boundary.

Images