Iot-based ptfe fiber membrane production line process parameter optimization control method

Abstract

The present application relates to the technical field of PTFE fiber membrane production, in particular to a PTFE fiber membrane production line process parameter optimization control method based on the Internet of Things, comprising: obtaining a real-time data set of Internet of Things sensors of multiple process links on the production line, including primary extrusion process parameters, drafting process timing parameters, sintering process temperature and pressure data, and secondary extrusion composite process parameters; performing feature extraction and fusion on the data to generate a production line comprehensive state feature matrix; calling a particle swarm optimization algorithm improved according to the physical coupling relationship of each process parameter to analyze the matrix and generate a process parameter optimization suggestion vector; generating an adjustment instruction set through a process controller collaborative calculation model and issuing it to the extruder, drafting machine and sintering furnace actuator to drive their collaborative adjustment. This method realizes accurate optimization of process parameters and collaborative control of actuators, ensuring stable operation of the production line.

Description

Technical Field

[0001] This invention relates to the field of PTFE fiber membrane production technology, and in particular to a method for optimizing and controlling process parameters of a PTFE fiber membrane production line based on the Internet of Things. Background Technology

[0002] PTFE fiber membranes, with their excellent high-temperature resistance, corrosion resistance, and air permeability, are widely used in environmental protection, chemical, and other fields. Their production quality is closely related to the precision of the production line's process parameters. Currently, the process parameter control of PTFE fiber membrane production lines largely relies on manual monitoring or single sensor data acquisition. Standard particle swarm optimization algorithms are used to analyze these fragmented process parameters and generate parameter adjustment suggestions. However, the sensor data from each process stage is not effectively fused, only reflecting the operating status of a single stage and failing to comprehensively present the overall operating condition of the production line.

[0003] Standard particle swarm optimization algorithms fail to consider the physical coupling relationships between various process parameters in a PTFE fiber membrane production line, including primary extrusion, drawing, sintering, and secondary extrusion compounding. This leads to a disconnect between the generated process parameter optimization suggestions and actual production conditions, resulting in insufficient accuracy. Furthermore, in existing technologies, actuators such as extruders, drawing machines, and sintering furnaces are often adjusted independently, lacking coordination. This makes it difficult to match parameter adjustments across different process stages, easily leading to production instability and significant product quality fluctuations. It is necessary to address the issues of scattered and unfused sensor data, algorithms failing to consider parameter coupling relationships leading to inaccurate optimization, and the lack of coordinated adjustment among actuators. Summary of the Invention

[0004] The purpose of this invention is to address the shortcomings of existing technologies by proposing an IoT-based method for optimizing and controlling process parameters in PTFE fiber membrane production lines.

[0005] To achieve the above objectives, the present invention adopts the following technical solution: a method for optimizing and controlling process parameters of a PTFE fiber membrane production line based on the Internet of Things, comprising:

[0006] Acquire real-time data sets from IoT sensors at multiple process stages on the PTFE fiber membrane production line, including primary extrusion process parameters, drawing process timing parameters, sintering and molding process temperature and pressure data, and secondary extrusion composite process parameters;

[0007] Feature extraction and fusion processing are performed on the real-time data set of the IoT sensors to generate a comprehensive production line status feature matrix;

[0008] An improved particle swarm optimization algorithm is used to analyze the comprehensive state feature matrix of the production line, generating optimization suggestion vectors for process parameters across multiple process stages. This improved particle swarm optimization algorithm modifies the standard particle swarm optimization algorithm based on the physical coupling relationships between process parameters in the PTFE fiber membrane production line, including:

[0009] Based on the standard particle swarm optimization algorithm, a process coupling constraint term calculated from a physical correlation model of known process parameters is introduced;

[0010] In each iteration of updating the particle velocity, the process coupling constraint term is added as an additional velocity component to guide the particle's flight direction, causing it to tend toward the region that satisfies the physical coupling relationship between process parameters.

[0011] An adaptive weighting coefficient is set for the process coupling constraint term, and the adaptive weighting coefficient is dynamically adjusted according to the current particle position's distance from the boundary of the feasible process parameter region.

[0012] By introducing the process coupling constraint term and adaptive weight coefficient, the improved particle swarm optimization algorithm can more efficiently search for the global optimal solution or near-optimal solution that satisfies the physical constraints of the actual production process in the high-dimensional parameter space.

[0013] Based on the process parameter optimization suggestion vector, a specific set of process parameter adjustment instructions is generated through the process controller collaborative calculation model;

[0014] The set of process parameter adjustment instructions is sent to the extruder, drawing machine and sintering furnace actuators corresponding to the PTFE fiber membrane production line, driving them to make coordinated adjustments.

[0015] As a further aspect of the present invention, feature extraction and fusion processing are performed on the real-time data set of the IoT sensors to generate a comprehensive production line status feature matrix, including:

[0016] The primary extrusion process parameters are normalized to generate a normalized extrusion feature vector, wherein the primary extrusion process parameters include extrusion pressure, extrusion temperature and push speed.

[0017] The timing parameters of the drawing process are extracted to generate a dynamic feature sequence of the drawing process. The timing parameters of the drawing process include the speed timing and tension timing of the multi-stage drawing machine.

[0018] Multi-scale noise reduction and feature point recognition are performed on the temperature and pressure data of the sintering process to generate a sintering stability feature vector.

[0019] A coupling relationship analysis was performed on the two-stage extrusion composite process parameters to generate an interactive feature vector of the composite process. The two-stage extrusion composite process parameters include the two-stage extrusion pressure, the base film traction speed and the composite interface temperature.

[0020] The normalized extrusion feature vector, the dynamic feature sequence of the drawing process, the sintering stability feature vector, and the composite process interaction feature vector are input into the spatiotemporal feature fusion network. The spatiotemporal feature fusion network performs a time-dimensional convolution on the dynamic feature sequence of the drawing process and splices and fuses the remaining static feature vectors in the spatial dimension, finally outputting the comprehensive state feature matrix of the production line.

[0021] As a further aspect of the present invention, the step of calling the improved particle swarm optimization algorithm to analyze the comprehensive state feature matrix of the production line and generating process parameter optimization suggestion vectors for multiple process stages includes:

[0022] The comprehensive state feature matrix of the production line is mapped to the search space of the improved particle swarm optimization algorithm, where each process parameter corresponds to one dimension of the search space;

[0023] Initialize the particle swarm, where each particle represents a set of candidate process parameter combinations. The position vector of the particle is composed of the primary extrusion pressure, the various stages of drawing speed, the sintering temperature, and the secondary extrusion pressure.

[0024] The fitness value of each particle is calculated by a predefined membrane performance prediction model based on the particle's position vector. The input of the membrane performance prediction model is process parameters, and the output is the predicted membrane porosity and separation comprehensive score.

[0025] The velocity and position of the particle swarm are updated according to the improved particle swarm optimization algorithm. During the update process, the velocity update of the particles not only considers the individual historical best position and the group historical best position, but is also subject to the correction of the physical coupling constraint between process parameters.

[0026] When the algorithm iterates to a preset number of times or the fitness value converges, the particle position vector corresponding to the historical best position of the population is decoded to generate the process parameter optimization suggestion vector.

[0027] As a further aspect of the present invention, based on the process parameter optimization suggestion vector, a specific set of process parameter adjustment instructions is generated through a process controller collaborative calculation model, including:

[0028] The process parameter optimization suggestion vector is input into the process controller collaborative calculation model, which pre-sets the control logic and response characteristics of each actuator.

[0029] The collaborative computing model of the process controller first performs a feasibility check on the input process parameter optimization suggestion vector. The feasibility check is based on the current state, physical limits and adjustment rate of each actuator.

[0030] For parameter recommendations that pass the feasibility verification, the process controller collaborative calculation model decomposes them into sub-instructions for each independent actuator;

[0031] The process controller collaborative computing model coordinates the execution order and timing of sub-instructions based on the execution timing dependencies and potential interference between sub-instructions, generating a conflict-free and time-optimal sequence of process parameter adjustment instructions, i.e., the process parameter adjustment instruction set.

[0032] As a further aspect of the present invention, the process controller collaborative computing model, based on the execution timing dependencies and potential interference between sub-instructions, collaboratively plans the execution order and timing of sub-instructions to generate a conflict-free and time-optimal sequence of process parameter adjustment instructions, including:

[0033] Construct an instruction dependency graph with sub-instructions as nodes and inter-instruction dependencies as directed edges, wherein the dependencies include sequential dependencies, resource mutual exclusion, and time window constraints;

[0034] On the instruction dependency graph, an instruction scheduling algorithm based on the critical path method is applied to calculate the earliest start time, latest start time, and relaxation time of each sub-instruction;

[0035] Prioritize scheduling sub-instructions on the critical path and allocate them specific time slots;

[0036] For sub-instructions on non-critical paths, within their relaxation time range, combined with the idle time window of the actuator, a heuristic algorithm is used for filling scheduling, with the goal of minimizing the overall process adjustment cycle.

[0037] The scheduling results are converted into a sequence of instructions with timestamps, forming the conflict-free and time-optimal process parameter adjustment instruction sequence.

[0038] As a further aspect of the present invention, the set of process parameter adjustment instructions is issued to the extruder, drawing machine, and sintering furnace actuators corresponding to the PTFE fiber membrane production line, driving them to perform coordinated adjustments, including:

[0039] The set of process parameter adjustment instructions is converted into underlying control protocol instructions that can be recognized by the controllers of each actuator through an industrial IoT gateway;

[0040] According to the timestamps in the instruction sequence, the underlying control protocol instructions are sent to the corresponding extruder controller, drawing machine controller and sintering furnace controller in a timely manner through timing event triggers;

[0041] After the instruction is issued, the feedback signals of each actuator are monitored in real time. The feedback signals include the deviation between the actual execution parameters and the instruction target value.

[0042] When the deviation is detected to continuously exceed the preset threshold, the adjustment anomaly handling process is triggered. The adjustment anomaly handling process includes pausing the issuance of subsequent instructions, recording the abnormal status, and sending an alarm to the upper-level system.

[0043] As a further aspect of the present invention, the adjusted anomaly handling process includes pausing the issuance of subsequent instructions, recording the anomaly status, and sending an alarm to the upper-level system, including:

[0044] When the deviation exceeds the preset threshold, a pause signal is immediately sent to the timing event trigger to interrupt the issuance of all current and subsequent unexecuted instructions;

[0045] Meanwhile, the abnormal status log records the identifier of the executing agency that caused the abnormality, the content of the abnormal instruction, the target value and the actual feedback value, as well as the timestamp of the abnormality.

[0046] Based on the recorded abnormal status information, a structured abnormal alarm message is generated, which includes the abnormal type, the scope of impact, and the suggested handling measures.

[0047] The abnormal alarm message is pushed to the upper-level monitoring system of the production line in real time through a preset communication link, and then waits for further instructions from the upper-level system.

[0048] As a further aspect of the present invention, the step of adding a process coupling constraint term as an additional velocity component during each iteration of particle velocity update to guide the particle's flight direction towards a region that satisfies the physical coupling relationship between process parameters includes:

[0049] In each iteration, based on the predefined physical correlation model between process parameters, the coupling constraint force between any two process parameters in the current particle position vector is calculated. For two process parameters that are physically correlated, the magnitude of the coupling constraint force is proportional to the difference between the current value of the two process parameters and the expected value defined by the correlation model, and the direction of the coupling constraint force is to reduce the difference.

[0050] The calculated coupling constraint forces are vector synthesized for all associated process parameters in the current particle position vector to generate the total process coupling constraint force vector acting on the current particle.

[0051] The additional velocity component is obtained by multiplying the total process coupling constraint force vector by the current adaptive weight coefficient.

[0052] The additional velocity component is vector-superimposed with the original velocity update amount calculated based on the individual historical best position and the group historical best position to obtain the final velocity update amount of the particle in this iteration.

[0053] The velocity of the current particle is updated using the final velocity update amount, and the position of the current particle is updated based on the updated velocity.

[0054] As a further aspect of the present invention, the collaborative computing model of the process controller first performs a feasibility verification on the input process parameter optimization suggestion vector. This feasibility verification is based on the current state, physical limits, and adjustment rate of each actuator, and includes:

[0055] The status database of each actuator is queried to obtain the current status parameters of the extruder, drawing machine and sintering furnace. The current status parameters include the current operating mode, current process parameter values ​​and equipment health indicators.

[0056] Read the physical limit parameters of each actuator from the equipment physical limit configuration table. The physical limit parameters include the maximum working pressure and temperature range of the first-stage extruder, the maximum and minimum speeds of each stage of the drawing machine, and the maximum withstand temperature and pressure of the sintering furnace.

[0057] The adjustment rate parameters of each actuator are obtained from the dynamic characteristic model of the equipment. The adjustment rate parameters include the maximum rate of change of extruder pressure, the maximum acceleration of speed adjustment of drawing machine, and the maximum rate of heating and cooling of sintering furnace.

[0058] The target parameter value suggested for each actuator in the process parameter optimization suggestion vector is compared with its corresponding current state parameter, physical limit parameter and adjustment rate parameter, and boundary check, change rate feasibility check and state compatibility check are performed in sequence.

[0059] If any check fails, the process parameter optimization suggestion vector is determined to be infeasible in whole or in part, and a feasibility verification failure report containing the specific reasons for failure and the actuators involved is generated based on the check results.

[0060] If all checks pass, the proposed process parameter optimization vector is deemed feasible, and a flag indicating successful verification and a subset of parameters that can be safely executed are output.

[0061] Compared with the prior art, the advantages and positive effects of the present invention are as follows:

[0062] Based on the physical coupling relationships between various process parameters in a PTFE fiber membrane production line, the standard particle swarm optimization algorithm is improved. This improved algorithm analyzes the comprehensive state characteristic matrix of the production line, which integrates primary extrusion process parameters, drawing process timing parameters, sintering temperature and pressure data, and secondary extrusion composite process parameters. This generates optimization suggestion vectors for multiple process stages. Compared to the conventional standard particle swarm optimization algorithm, which only analyzes scattered parameters, this method more accurately captures the inherent relationships between process parameters, avoiding a disconnect between optimization suggestions and actual production conditions. This makes process parameter optimization more aligned with the actual operating needs of the production line, reducing the randomness of parameter optimization and making the optimization suggestions more targeted and feasible.

[0063] Based on the process parameter optimization suggestion vector, a specific set of process parameter adjustment instructions is generated through a collaborative calculation model of the process controller. This set of instructions is then sent to the actuators of the extruder, drawing machine, and sintering furnace corresponding to the PTFE fiber membrane production line, driving them to make coordinated adjustments. Compared with the independent adjustment mode of each actuator in conventional technology, this method can achieve synchronous matching of parameters in various process stages such as extrusion, drawing, sintering, and secondary extrusion compounding. This avoids overall process imbalance caused by adjusting parameters in a single stage, improves the coordination and consistency of process parameter adjustments, reduces unstable factors in the production process, and lowers product quality fluctuations. Attached Figure Description

[0064] Figure 1 This is a flowchart of the IoT-based process parameter optimization and control method for PTFE fiber membrane production lines according to the present invention.

[0065] Figure 2 A flowchart for generating feature matrices through feature extraction and fusion processing;

[0066] Figure 3 A flowchart illustrating the introduction of physical coupling constraints to improve the particle swarm optimization algorithm. Detailed Implementation

[0067] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention.

[0068] In the description of this invention, it should be understood that the terms "length," "width," "upper," "lower," "front," "rear," "left," "right," "vertical," "horizontal," "top," "bottom," "inner," and "outer," etc., indicating orientation or positional relationships, are based on the orientation or positional relationships shown in the accompanying drawings and are only for the convenience of describing the invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation, and therefore should not be construed as a limitation of the invention. Furthermore, in the description of this invention, "a plurality of" means two or more, unless otherwise explicitly specified.

[0069] See Figure 1 This invention provides a method for optimizing and controlling process parameters in a PTFE fiber membrane production line based on the Internet of Things (IoT). The overall implementation scheme is as follows:

[0070] By deploying IoT sensors at multiple process stages on the production line, real-time data are collected on primary extrusion process parameters, drawing process timing parameters, sintering temperature and pressure data, and secondary extrusion composite process parameters, forming a real-time data set from the IoT sensors. Feature extraction and fusion processing are performed on this data set to generate a comprehensive production line status feature matrix representing the overall operating status of the production line. An improved particle swarm optimization algorithm is then used to analyze this matrix. This algorithm improves upon the standard particle swarm optimization algorithm based on the physical coupling relationships between various process parameters in the PTFE fiber membrane production line, thereby generating process parameter optimization suggestion vectors for multiple process stages. These suggestion vectors are input into the process controller collaborative computing model, and after feasibility verification and collaborative planning, a specific set of process parameter adjustment instructions is generated. This set of instructions is then sent to the corresponding extruders, drawing machines, and sintering furnace actuators on the production line, driving them to perform collaborative adjustments and achieving closed-loop optimization control of the production process parameters.

[0071] In one embodiment of the present invention, the process of performing feature extraction and fusion processing on a real-time data set of IoT sensors to generate a comprehensive production line status feature matrix is ​​implemented as follows: (See reference) Figure 2The system normalizes the primary extrusion process parameters, including extrusion pressure, extrusion temperature, and pushing speed, to eliminate the influence of different dimensions and generate a normalized extrusion feature vector. It extracts temporal features from the drawing process timing parameters, including the speed and tension timing of the multi-stage drawing machine, to capture the dynamic changes in the drawing process and generate a dynamic feature sequence. It performs multi-scale noise reduction and feature point identification on the temperature and pressure data of the sintering process, filtering out noise interference and extracting key features reflecting sintering stability to generate a sintering stability feature vector. It analyzes the coupling relationship of the secondary extrusion composite process parameters, including secondary extrusion pressure, base film traction speed, and composite interface temperature, quantifying the interaction between parameters and generating a composite process interaction feature vector. The normalized extrusion feature vector, the dynamic feature sequence of the drawing process, the sintering stability feature vector, and the composite process interaction feature vector are then input into a spatiotemporal feature fusion network. This network performs a convolution operation on the dynamic feature sequence of the drawing process in the time dimension to extract temporal correlation and splices and fuses the remaining static feature vectors in the spatial dimension, outputting a comprehensive production line state feature matrix that fully characterizes the overall state of the production line.

[0072] In specific implementation, the process of performing feature extraction and fusion processing on the real-time data set of IoT sensors to generate a comprehensive state feature matrix of the production line is carried out in the following way: In specific implementation, the first-level extrusion process parameters include extrusion pressure, extrusion temperature and pushing speed. For the collected extrusion pressure value sequence, extrusion temperature value sequence and pushing speed value sequence, the maximum and minimum values ​​are normalized respectively, so that the value of each parameter is linearly transformed to the [0,1] interval, thereby generating a normalized extrusion feature vector composed of normalized extrusion pressure, extrusion temperature and pushing speed, so as to eliminate the influence of different physical dimensions on subsequent feature fusion.

[0073] In some embodiments, the timing parameters of the drawing process include the speed timing and tension timing of the multi-stage drawing machine. For the continuously collected speed timing data and tension timing data of the multi-stage drawing machine, a sliding window mechanism is used to extract a fixed-length timing segment. Within each timing segment, the mean, variance, and first-order difference mean are calculated as statistical features. The statistical features of the multi-stage drawing machine are arranged in chronological order to generate a dynamic feature sequence of the drawing process, which is used to characterize the dynamic law of speed and tension evolution over time during the drawing process.

[0074] It is understood that the temperature and pressure data in the sintering process are high-frequency sampled data with noise. Through multi-scale noise reduction processing, filtering algorithms are applied at different time scales to remove random fluctuations, and feature points are identified on the noise-reduced data curves, including the peak points of the temperature curve and the starting points of the steady-state plateau of the pressure curve. The values ​​of these feature points and their relative time positions are combined to generate a sintering stability feature vector, which reflects the degree of control and stability of the sintering process.

[0075] In some embodiments, the secondary extrusion composite process parameters include secondary extrusion pressure, base film traction speed, and composite interface temperature. By performing correlation analysis and regression modeling on historical data of these three parameters, a functional relationship between secondary extrusion pressure and base film traction speed, as well as a response surface model of composite interface temperature as a function of the former two, are established. Interaction coefficients representing the mutual influence between parameters are extracted from the model, and these coefficients constitute the composite process interaction feature vector, which is used to quantify the coupling effect between different parameters in the secondary extrusion composite stage.

[0076] In practical implementation, the normalized extrusion feature vector, the dynamic feature sequence of the drawing process, the sintering stability feature vector, and the composite process interaction feature vector are input into the spatiotemporal feature fusion network. The spatiotemporal feature fusion network first performs a one-dimensional convolution operation on the dynamic feature sequence of the drawing process, with the convolution kernel sliding along the time dimension to extract local temporal patterns, compressing the convolution output into a low-dimensional time feature vector. Then, the normalized extrusion feature vector, the sintering stability feature vector, and the composite process interaction feature vector are concatenated in the spatial dimension to form a high-dimensional static feature vector. The time feature vector and the static feature vector are concatenated again, and the dimensions are adjusted through a fully connected layer to output a unified comprehensive state feature matrix of the production line. It can be understood that the convolution operation in the time dimension of the spatiotemporal feature fusion network can be represented as:

[0077] ,in: This represents the dynamic characteristic sequence of the stretching process. and These represent the kernel weights and biases, respectively. This represents the convolution operation. This represents the output time feature vector.

[0078] In one embodiment of the present invention, the process of analyzing the comprehensive state feature matrix of the production line and generating process parameter optimization suggestion vectors by calling an improved particle swarm optimization algorithm is implemented as follows: the comprehensive state feature matrix of the production line is mapped to the search space of the improved particle swarm optimization algorithm, wherein each process parameter corresponds to a dimension in the search space; the particle swarm is initialized so that each particle represents a set of candidate process parameter combinations, and the position vector of the particle is composed of specific parameters such as primary extrusion pressure, primary drawing speed, sintering temperature, and secondary extrusion pressure; the fitness value of each particle is calculated, which is calculated by a predefined membrane performance prediction model based on the position vector of the particle. The input of the membrane performance prediction model is the process parameters, and the output is the predicted membrane porosity and separation comprehensive score; the velocity and position of the particle swarm are updated according to the improved particle swarm optimization algorithm. During the update process, the velocity update of the particle not only considers the individual historical best position and the group historical best position, but is also subject to the correction of the physical coupling constraint between process parameters; when the algorithm iteration reaches a preset number of times or the fitness value converges, the particle position vector corresponding to the group historical best position is decoded and restored to the actual process parameter value to generate the process parameter optimization suggestion vector.

[0079] In practice, the process of using the improved particle swarm optimization algorithm to analyze the comprehensive state feature matrix of the production line and generate optimization suggestion vectors for process parameters for multiple process stages is implemented as follows: In practice, each feature dimension of the comprehensive state feature matrix of the production line is mapped to the coordinate axis of the search space of the improved particle swarm optimization algorithm. Each dimension of the search space corresponds to a process parameter to be optimized, including the first-stage extrusion pressure, the first-stage drawing speed, the second-stage drawing speed, the sintering furnace temperature, the sintering furnace pressure, and the second-stage extrusion pressure. This establishes a one-to-one mapping relationship between the abstract features in the comprehensive state feature matrix of the production line and the specific physical process parameters. When initializing the particle swarm optimization algorithm, the number of particles is set to a fixed value. The position vector of each particle is a six-dimensional real vector. The first element of the position vector represents a candidate value for the first-stage extrusion pressure, the second element represents a candidate value for the first-stage drawing speed, the third element represents a candidate value for the second-stage drawing speed, the fourth element represents a candidate value for the sintering furnace temperature, the fifth element represents a candidate value for the sintering furnace pressure, and the sixth element represents a candidate value for the second-stage extrusion pressure. The initial position of the particles is randomly generated within the allowable range of each process parameter, and the initial velocity of the particles is set to a zero vector or a small-amplitude random perturbation.

[0080] In some embodiments, when calculating the fitness value of each particle, the predefined membrane performance prediction model adopts an artificial neural network model trained on historical production data. The number of nodes in the input layer of the membrane performance prediction model is the same as the dimension of the particle position vector. The six nodes in the input layer sequentially receive the candidate values ​​of the first-level extrusion pressure, the first-level drawing speed, the second-level drawing speed, the sintering furnace temperature, the sintering furnace pressure, and the second-level extrusion pressure from the particle position vector. The output layer of the membrane performance prediction model contains two nodes. The first output node outputs the predicted membrane porosity value, and the second output node outputs the predicted membrane separation comprehensive score value. The membrane porosity value and the membrane separation comprehensive score value are weighted and summed, and the result is used as the fitness value of the current particle. The weight coefficients of the membrane porosity and the membrane separation comprehensive score are pre-configured according to the emphasis on product performance indicators in actual production.

[0081] It is understandable that, in the process of updating the velocity and position of a particle swarm based on the improved particle swarm optimization algorithm, the velocity update not only considers the individual's historical best position and the swarm's historical best position, but is also subject to correction by the physical coupling constraints between process parameters. In one iteration, for a given particle in the swarm, its velocity and position updates follow specific mathematical rules. The formula for updating the particle velocity during the update process is:

[0082] ,in: This represents the velocity component of the i-th particle in the d-th dimension during the (k+1)-th iteration. Let w represent the velocity component of the i-th particle in the d-th dimension at the k-th iteration, and w represent the inertia weighting coefficient. and Represents the acceleration constant. and Represents a random number in the range [0,1]. This represents the historical best position component of the i-th particle in the d-th dimension up to the k-th iteration. This represents the current position component of the i-th particle in the d-th dimension at the k-th iteration. This represents the historical best position component of the entire particle swarm in the d-th dimension up to the k-th iteration. This represents the constraint force component in the d-th dimension, calculated based on the physical coupling relationship between process parameters. The introduction of this constraint force component guides the particle during flight by the physical correlation between parameters in the actual production process.

[0083] In some embodiments, the improved particle swarm optimization algorithm sets a maximum number of iterations as the termination condition. When the algorithm reaches the preset maximum number of iterations, the iteration process stops. At this point, the particle position vector corresponding to the historical best position of the swarm is taken as the final optimization result. This particle position vector is a six-dimensional real number vector. The particle position vector corresponding to the historical best position of the swarm is decoded, that is, the values ​​of its various dimensions are converted into the values ​​of actual process parameters according to a preset mapping relationship. The first dimension value is converted into the working pressure setting value of a specific model of twin-screw extruder, the second and third dimensions value is converted into the linear velocity setting value of the two-stage drawing rollers, the fourth and fifth dimensions value is converted into the temperature setting value of the sintering furnace zone and the furnace pressure setting value, and the sixth dimension value is converted into the working pressure setting value of another single-screw extruder, thereby generating a process parameter optimization suggestion vector containing six specific process parameter setting values.

[0084] Optionally, in addition to setting the maximum number of iterations as the termination condition, the improved particle swarm optimization algorithm also sets a fitness value convergence threshold. After each iteration, the difference between the current optimal fitness value of the population and the optimal fitness value of the population in the previous iteration is calculated. If the absolute value of the difference is less than the preset convergence threshold and continues for several iterations, the fitness value is considered to have converged, the algorithm iteration process is terminated in advance, and the current historical optimal position of the population is decoded to generate a process parameter optimization suggestion vector.

[0085] In one embodiment of the present invention, the improved particle swarm optimization algorithm improves the standard particle swarm optimization algorithm based on the physical coupling relationship between various process parameters in the PTFE fiber membrane production line. The specific implementation method is as follows: (See attached document) Figure 3Based on the standard particle swarm optimization algorithm, a process coupling constraint term calculated from a known physical correlation model between process parameters is introduced. During each iteration when updating the particle velocity, this process coupling constraint term is added as an additional velocity component to guide the particle's flight direction, causing it to tend towards the region satisfying the physical coupling relationship between process parameters. According to the predefined physical correlation model between process parameters, the coupling constraint force between any two process parameters in the current particle position vector is calculated. For two physically correlated process parameters, the magnitude of the coupling constraint force is proportional to the difference between the current value of the two process parameters and the expected value defined by the correlation model, and its direction points towards reducing this difference. The calculated coupling constraint force is applied to all correlated process parameters. Vector synthesis generates the total process coupling constraint force vector acting on the current particle. An adaptive weighting coefficient is set for this process coupling constraint force term, which is dynamically adjusted according to the distance between the current particle position and the boundary of the feasible process parameter region. The total process coupling constraint force vector is multiplied by the current adaptive weighting coefficient to obtain an additional velocity component. This additional velocity component is vector-superimposed with the original velocity update amount calculated based on the individual's historical best position and the group's historical best position to obtain the final velocity update amount of the particle in this iteration. The velocity of the current particle is updated using the final velocity update amount, and the position of the current particle is updated according to the updated velocity, thereby searching for the global optimal solution or near-optimal solution that satisfies the physical constraints of the actual production process more efficiently in the high-dimensional parameter space.

[0086] In its implementation, the improved particle swarm optimization algorithm modifies the standard particle swarm optimization algorithm based on the physical coupling relationships between various process parameters in the PTFE fiber membrane production line. Building upon the standard algorithm, a process coupling guidance term calculated from a known physical correlation model between process parameters is introduced. This term serves as an additional guiding variable in the particle velocity update process. During each iteration of particle velocity updates, the pairwise coupling guidance components between any two process parameters in the current particle position vector are calculated according to the predefined physical correlation model. For two physically correlated process parameters, such as a positively correlated material conservation relationship between primary extrusion pressure and pushing speed, or a thermodynamic equilibrium relationship between sintering furnace temperature and sintering furnace pressure, the magnitude of the coupling guidance component is proportional to the difference between the current value of these two process parameters and the expected value defined by the correlation model. The direction of the coupling guidance component always points in a direction that reduces this difference, i.e., it drives the parameter values ​​to adjust in a direction that satisfies the physical correlation relationship.

[0087] In some embodiments, the calculated pairwise coupling guidance components for all associated process parameters in the current particle position vector are vector synthesized. All coupling guidance components are then summed according to vector addition rules to generate a total process coupling guidance vector acting on the current particle. This total process coupling guidance vector reflects the overall correction requirement of the physical coupling relationship between all process parameters on the current particle position. An adaptive scaling factor is set for the process coupling guidance term. This adaptive scaling factor is dynamically adjusted based on the distance between the current particle position and the boundary of the feasible process parameter region. When the particle position is close to the boundary of the feasible region, the adaptive scaling factor is increased to enhance the guiding effect of process coupling and prevent the particle from flying out of the feasible solution space; when the particle position is inside the feasible region, the adaptive scaling factor is decreased to preserve the particle's own exploration capability.

[0088] In practice, the total process coupling steering vector is multiplied by the current adaptive scaling factor to obtain an additional velocity component used to correct the particle's flight direction. This additional velocity component is then vector-superimposed with the original velocity update calculated based on the individual and group historical best positions to obtain the final velocity update for the particle in this iteration. The formula for calculating the final velocity update is as follows:

[0089] ,in: This represents the final velocity update of the i-th particle in the t-th iteration, and has velocity dimensions. This represents the inertia weight, which is a dimensionless coefficient. This represents the velocity vector of the i-th particle at the (t-1)-th iteration; and These represent individual learning factors and social learning factors, respectively, and are dimensionless constants. This represents the individual historical best position vector of the i-th particle; This represents the position vector of the i-th particle at the (t-1)-th iteration; Represents the historical best position vector of the group; represents the adaptive scaling factor at the t-th iteration, which is a dimensionless weight; Let represent the total process coupling guidance vector acting on the particle at the t-th iteration. This vector is synthesized from the coupling guidance components of all associated process parameter pairs and is constructed as a physical quantity with velocity dimensions to ensure that the dimensions on both sides of the formula are consistent. See Table 1.

[0090] Table 1: Illustrated Rules for Adaptive Scaling Factor Adjustment

[0091] ,

[0092] It can be understood that the current particle's velocity vector is updated using the calculated final velocity update, and the current particle's position vector is updated according to the displacement formula based on the updated velocity vector. This allows the particle to be guided not only by its own and the group's experience during the search process, but also by the inherent physical laws of the production process, ensuring it flies in a direction that satisfies the physical constraints of the actual production process in the high-dimensional parameter space. In this way, the improved particle swarm optimization algorithm can effectively avoid searching for invalid solutions that violate physical laws, accelerate the convergence speed, and thus more efficiently search for the globally optimal or near-optimal solution that satisfies the physical constraints of the actual production process. Optionally, the dynamic adjustment strategy of the adaptive scaling factor can be adjusted according to the sensitivity of specific process parameters. For parameter sets with strong physical coupling, the baseline level of the adaptive scaling factor is appropriately increased to enhance the guiding effect of the guiding term, while for parameter sets with weak physical coupling, the baseline level is decreased to give particles greater freedom to explore.

[0093] In one embodiment of the present invention, the process of generating a set of process parameter adjustment instructions through a process controller collaborative calculation model based on the process parameter optimization suggestion vector is implemented as follows: the process parameter optimization suggestion vector is input into a process controller collaborative calculation model that has preset control logic and response characteristics of each actuator; the model performs a feasibility check on the input process parameter optimization suggestion vector, queries the state database of each actuator to obtain the current state parameters of the extruder, drawing machine and sintering furnace, reads the physical limit parameters of each actuator from the equipment physical limit configuration table, and obtains the adjustment rate parameters of each actuator from the equipment dynamic characteristic model; the suggested target parameter value is compared with the corresponding current state parameter, physical limit parameter and adjustment rate parameter, and boundary check, change rate feasibility check and state compatibility check are performed in sequence. If any check fails... If the check fails, a feasibility verification failure report is generated; if all checks pass, the suggested vector is deemed feasible. For parameter suggestions that pass the verification, the model decomposes them into sub-instructions for each independent actuator. An instruction dependency graph is constructed, with sub-instructions as nodes and directed edges for instruction order dependencies, resource mutual exclusion, and time window constraints. An instruction scheduling algorithm based on the critical path method is applied to this graph to calculate the earliest start time, latest start time, and relaxation time of each sub-instruction. Sub-instructions on the critical path are prioritized for scheduling and allocated specific time slots. For sub-instructions on non-critical paths, a heuristic algorithm is used to fill the scheduling within their relaxation time range, combined with the idle time window of the actuator, with the goal of minimizing the overall process adjustment cycle. The scheduling results are transformed into a sequence of timestamped instructions, forming a conflict-free and time-optimal set of process parameter adjustment instructions.

[0094] In practical implementation, the process of generating a specific set of process parameter adjustment instructions based on the process parameter optimization suggestion vector and through the process controller collaborative calculation model is carried out as follows: In practical implementation, the process controller collaborative calculation model pre-sets the control logic and response characteristics of each actuator. The process controller collaborative calculation model performs feasibility verification on the input process parameter optimization suggestion vector and queries the status database of each actuator to obtain the current status parameters of the extruder, drawing mill, and sintering furnace. The current status parameters include the extruder currently being in automatic operation mode, the current extrusion pressure being 12.5 MPa, the current linear speed of the drawing mill being 15 m / min, the current temperature of the sintering furnace being 385 degrees Celsius, and all equipment health indicators being [missing information]. Normal; Read the physical limit parameters of each actuator from the equipment physical limit configuration table, including the maximum working pressure of the first-stage extruder is 25 MPa, the minimum working temperature is 280 degrees Celsius, and the maximum working temperature is 360 degrees Celsius; the maximum speed of each stage of the drawing mill is 35 m / min, and the minimum speed is 5 m / min; the maximum withstand temperature of the sintering furnace is 420 degrees Celsius, and the maximum withstand pressure is 0.8 MPa; Obtain the adjustment rate parameters of each actuator from the equipment dynamic characteristic model, including the maximum rate of pressure change of the extruder is 0.5 MPa per second, the maximum acceleration of the drawing mill speed adjustment is 0.2 m / s², the maximum heating rate of the sintering furnace is 5 degrees Celsius per minute, and the maximum cooling rate is 3 degrees Celsius per minute.

[0095] In some embodiments, the target parameter values ​​suggested for each actuator in the process parameter optimization suggestion vector are compared with the corresponding current state parameters, physical limit parameters, and adjustment rate parameters. Boundary checks, rate-of-change feasibility checks, and state compatibility checks are performed sequentially. The boundary check determines whether the target parameter exceeds the range of the physical limit parameter; the rate-of-change feasibility check determines whether the rate of change required to adjust from the current value to the target value exceeds the limit of the adjustment rate parameter; and the state compatibility check determines whether the target parameter matches the current operating mode and health status of the actuator. If any check fails, the process parameter optimization suggestion vector is deemed infeasible in whole or in part, and a feasibility verification failure report containing the specific reasons for failure and the relevant actuators is generated based on the check results. If all checks pass, the process parameter optimization suggestion vector is deemed feasible, and a verification pass flag and a subset of parameters that can be safely executed are output.

[0096] Understandably, for parameter recommendations that have passed feasibility verification, the process controller collaborative calculation model decomposes them into sub-instructions for each independent actuator. A complete process parameter optimization recommendation vector contains three target setpoints: extruder target pressure of 16 MPa, drawing machine target speed of 18 m / min, and sintering furnace target temperature of 395 degrees Celsius. After decomposition, three independent sub-instructions are generated: the first sub-instruction is directed to the extruder controller, which adjusts the pressure to 16 MPa; the second sub-instruction is directed to the drawing machine controller, which adjusts the speed to 18 m / min; and the third sub-instruction is directed to the sintering furnace controller, which adjusts the temperature to 395 degrees Celsius.

[0097] In practical implementation, the collaborative computing model of the process controller constructs an instruction dependency graph with sub-instructions as nodes and directed edges representing inter-instruction dependencies. These dependencies include sequential dependencies, resource exclusivity, and time window constraints. For example, the extruder pressure adjustment sub-instruction must be completed before the drawing machine speed adjustment sub-instruction because the drawing tension is sensitive to changes in extrusion pressure; therefore, they have a sequential dependency. The drawing machine speed adjustment sub-instruction and the sintering furnace temperature adjustment sub-instruction share the same power supply resources within the same time period, resulting in resource exclusivity. The sintering furnace temperature adjustment sub-instruction must be initiated within three minutes to meet production cycle requirements, thus exhibiting a time window constraint. On the instruction dependency graph, an instruction scheduling algorithm based on the critical path method is applied to calculate the earliest start time, latest start time, and relaxation time for each sub-instruction. Sub-instructions on the critical path are prioritized for scheduling, and a defined time slot is allocated to them. Delays in sub-instructions on the critical path will extend the overall adjustment cycle (see Table 2).

[0098] Table 2: Calculation Results of Subinstruction Scheduling Time Parameters

[0099] ,

[0100] Optionally, for sub-instructions on non-critical paths, within their relaxation time range, a heuristic algorithm is used for filling scheduling, combined with the idle time window of the actuator, with the goal of minimizing the overall process adjustment cycle. Non-critical sub-instructions are scheduled as early as possible to reduce the waiting time of subsequent processes. The scheduling results are transformed into a sequence of timestamped instructions, each instruction accompanied by a precise start time, forming a conflict-free and time-optimal sequence of process parameter adjustment instructions, i.e., the process parameter adjustment instruction set. It can be understood that when the process controller collaborative computing model performs a feasibility check on the rate of change, it calculates the minimum time required to adjust from the current value to the target value and compares it with the adjustment rate parameter. Its judgment logic can be expressed by the following inequality:

[0101] ,in: Indicates the minimum time required to complete the adjustment. This represents the suggested target parameter values ​​in the process parameter optimization suggestion vector. This represents the current state parameter value retrieved from the state database. This represents the adjustment rate parameter obtained from the device's dynamic characteristic model. This indicates the maximum allowable adjustment time for the production plan.

[0102] In one embodiment of the present invention, the process of issuing a set of process parameter adjustment instructions to the actuators of the PTFE fiber membrane production line and driving them to adjust in a coordinated manner is implemented as follows: the set of process parameter adjustment instructions is converted into underlying control protocol instructions that can be recognized by the controllers of each actuator through an industrial IoT gateway; according to the timestamp in the instruction sequence, the underlying control protocol instructions are sent to the corresponding extruder controller, drawing machine controller, and sintering furnace controller in a timely manner through a timing event trigger; after the instructions are issued, the feedback signals of each actuator are monitored in real time, and the signals contain the deviation between the actual execution parameters and the instruction target values; when the deviation is detected to continuously exceed a preset threshold, the adjustment anomaly handling process is triggered, and a pause signal is immediately sent to the timing event trigger to interrupt the issuance of all current and subsequent unexecuted instructions; at the same time, the actuator identifier, abnormal instruction content, target value, actual feedback value, and timestamp of the anomaly are recorded in the anomaly status log; based on the recorded anomaly status information, a structured anomaly alarm message containing the anomaly type, scope of impact, and suggested handling measures is generated; the message is pushed to the upper-level monitoring system of the production line in real time through a preset communication link, and further instructions from the upper-level system are awaited.

[0103] In practice, the process parameter adjustment instruction set is issued to the extruder, drawing machine, and sintering furnace actuators corresponding to the PTFE fiber membrane production line, driving them to perform coordinated adjustments in the following manner: The process parameter adjustment instruction set includes multiple control instructions with timestamps, such as "adjust the pressure of extruder No. 1 to 16.0 MPa at 10:05:00 on 2026-04-22", "adjust the speed of the main drawing machine to 18.2 m / min at 10:06:30 on 2026-04-22", and "adjust the temperature of zone three of the sintering furnace to 398℃ at 10:07:15 on 2026-04-22". Through an industrial IoT gateway, these high-level instructions are converted into ModbusTCP or PROFINETIO control protocol instructions recognizable by the underlying actuator controllers. During the conversion process, the floating-point parameters in the instructions are packaged according to the device register address mapping table, and the timestamp information is converted into timed trigger events within the device controller. According to the timestamp in the instruction sequence, the underlying control protocol instructions are sent to the corresponding controller on time through the timing event trigger. The timing event trigger maintains a high-precision clock. When the clock reaches the instruction timestamp, the corresponding Modbus write instruction or PROFINET output frame is immediately sent to the IP address or bus node of the target device to ensure that the instruction is executed within the millisecond error range.

[0104] In some embodiments, during the execution process after the instruction is issued, the feedback signals of each actuator are monitored in real time. The feedback signals include the deviation between the actual execution parameters and the target value of the instruction. For the extruder, the analog feedback value of the pressure sensor is read in real time; for the drawing machine, the actual linear velocity measured by the encoder is read in real time; for the sintering furnace, the actual temperature of each temperature zone measured by the thermocouple is read in real time. The actual feedback value is compared with the target value in the instruction in real time, the instantaneous deviation is calculated, and the duration of the deviation is continuously tracked. When the absolute value of the deviation between the actual feedback value and the target value of an actuator is detected to continuously exceed a preset threshold and the duration reaches a set time, the adjustment anomaly handling process is triggered. A pause signal is immediately sent to the timing event trigger to interrupt the currently issued instruction and the sending of all subsequent unexecuted instructions, preventing continuous misoperation under erroneous operating conditions from causing equipment damage or product scrapping.

[0105] It is understandable that adjusting the anomaly handling process also includes recording the identifier of the executing mechanism that caused the anomaly, the content of the anomaly instruction, the target value and the actual feedback value, as well as the timestamp of the anomaly occurrence in the anomaly status log. For example, recording "Anomalous equipment: No. 2 drawing machine, instruction content: speed up to 18.2 m / min, target value: 18.2, actual value: 17.5, deviation: 0.7 m / min, threshold: 0.5 m / min, occurrence time: 2026-04-22 10:06:32.105". Based on the recorded anomaly status information, a structured anomaly alarm message is generated. The anomaly alarm message is encapsulated in JSON or XML format, including anomaly type field, impact scope field, and suggested handling measure field. The anomaly alarm message is pushed to the upper-level monitoring system of the production line in real time through a preset MQTT or OPCUA communication link, and is displayed in a prominent pop-up window in the system, waiting for manual confirmation or automatic processing instructions from the upper-level system. Optionally, in the deviation monitoring stage, to distinguish between instantaneous fluctuations and persistent anomalies, a weighted moving average method with a forgetting factor is used to process the deviation sequence, smoothing short-term fluctuations and highlighting persistent deviation trends. The calculation method is as follows:

[0106] ,in: This represents the instantaneous deviation between the actual feedback value and the target value at the nth sampling time. This represents the smoothed bias at the nth sampling time. A forgetting factor between 0 and 1, used to adjust the weighting of recent data. This factor is applied after smoothing the bias. An abnormal state is determined only when the absolute value of the signal exceeds a preset threshold and remains so for a specified period, effectively filtering out false alarms caused by sensor noise or instantaneous load fluctuations.

[0107] The above are merely preferred embodiments of the present invention and are 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 that can be applied to 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 optimizing and controlling process parameters in a PTFE fiber membrane production line based on the Internet of Things, characterized in that, The method includes: The real-time data set of IoT sensors for multiple process stages on the PTFE fiber membrane production line is obtained. The real-time data set of IoT sensors includes primary extrusion process parameters, drawing process timing parameters, temperature and pressure data of sintering molding process, and secondary extrusion composite process parameters. Feature extraction and fusion processing are performed on the real-time data set of the IoT sensors to generate a comprehensive production line status feature matrix; An improved particle swarm optimization algorithm is invoked to analyze the comprehensive state feature matrix of the production line, generating process parameter optimization suggestion vectors for multiple process links. The improved particle swarm optimization algorithm improves the standard particle swarm optimization algorithm based on the physical coupling relationship between various process parameters in the PTFE fiber membrane production line. Based on the process parameter optimization suggestion vector, a specific set of process parameter adjustment instructions is generated through the process controller collaborative calculation model; The set of process parameter adjustment instructions is sent to the actuators of the extruder, drawing machine and sintering furnace corresponding to the PTFE fiber membrane production line, driving them to make coordinated adjustments. The improved particle swarm optimization algorithm improves the standard particle swarm optimization algorithm based on the physical coupling relationship between various process parameters in the PTFE fiber membrane production line, including: Based on the standard particle swarm optimization algorithm, a process coupling constraint term is introduced, which is calculated by a known physical correlation model between process parameters. In each iteration of updating the particle velocity, the process coupling constraint term is used as an additional velocity component. This additional velocity component is used to guide the particle's flight direction, causing it to tend towards the region that satisfies the physical coupling relationship between the process parameters. An adaptive weighting coefficient is set for the process coupling constraint term, and the adaptive weighting coefficient is dynamically adjusted according to the current particle position's distance from the boundary of the feasible process parameter region. By introducing the process coupling constraint term and adaptive weight coefficient, the improved particle swarm optimization algorithm can more efficiently search for the global optimal solution or near-optimal solution that satisfies the physical constraints of the actual production process in the high-dimensional parameter space.

2. The method for optimizing and controlling process parameters of a PTFE fiber membrane production line based on the Internet of Things as described in claim 1, characterized in that, Feature extraction and fusion processing are performed on the real-time data set of the IoT sensors to generate a comprehensive production line status feature matrix, including: The primary extrusion process parameters are normalized to generate a normalized extrusion feature vector, wherein the primary extrusion process parameters include extrusion pressure, extrusion temperature and push speed. The timing parameters of the drawing process are extracted to generate a dynamic feature sequence of the drawing process. The timing parameters of the drawing process include the speed timing and tension timing of the multi-stage drawing machine. Multi-scale noise reduction and feature point recognition are performed on the temperature and pressure data of the sintering process to generate a sintering stability feature vector. A coupling relationship analysis was performed on the two-stage extrusion composite process parameters to generate an interactive feature vector of the composite process. The two-stage extrusion composite process parameters include the two-stage extrusion pressure, the base film traction speed and the composite interface temperature. The normalized extrusion feature vector, the dynamic feature sequence of the drawing process, the sintering stability feature vector, and the composite process interaction feature vector are input into the spatiotemporal feature fusion network. The spatiotemporal feature fusion network performs a time-dimensional convolution on the dynamic feature sequence of the drawing process and splices and fuses the remaining static feature vectors in the spatial dimension, finally outputting the comprehensive state feature matrix of the production line.

3. The method for optimizing and controlling process parameters of a PTFE fiber membrane production line based on the Internet of Things as described in claim 1, characterized in that, The improved particle swarm optimization algorithm is invoked to analyze the comprehensive state feature matrix of the production line, generating process parameter optimization suggestion vectors for multiple process stages, including: The comprehensive state feature matrix of the production line is mapped to the search space of the improved particle swarm optimization algorithm, where each process parameter corresponds to one dimension of the search space; Initialize the particle swarm, where each particle represents a set of candidate process parameter combinations. The position vector of the particle is composed of the primary extrusion pressure, the various drawing speeds, the sintering temperature, and the secondary extrusion pressure. The fitness value of each particle is calculated by a predefined membrane performance prediction model based on the particle's position vector. The input of the membrane performance prediction model is process parameters, and the output is the predicted membrane porosity and separation comprehensive score. The velocity and position of the particle swarm are updated according to the improved particle swarm optimization algorithm. During the update process, the velocity update of the particles not only considers the individual historical best position and the group historical best position, but is also subject to the correction of the physical coupling constraint between process parameters. When the algorithm iterates to a preset number of times or the fitness value converges, the particle position vector corresponding to the historical best position of the population is decoded to generate the process parameter optimization suggestion vector.

4. The method for optimizing and controlling process parameters of a PTFE fiber membrane production line based on the Internet of Things as described in claim 1, characterized in that, Based on the process parameter optimization suggestion vector, a specific set of process parameter adjustment instructions is generated through the process controller collaborative calculation model, including: The process parameter optimization suggestion vector is input into the process controller collaborative calculation model, which pre-sets the control logic and response characteristics of each actuator. The collaborative computing model of the process controller first performs a feasibility check on the input process parameter optimization suggestion vector. The feasibility check is based on the current state, physical limits and adjustment rate of each actuator. For parameter recommendations that pass the feasibility verification, the process controller collaborative calculation model decomposes them into sub-instructions for each independent actuator; The process controller collaborative computing model coordinates the execution order and timing of sub-instructions based on the execution timing dependencies and potential interference between sub-instructions, generating a conflict-free and time-optimal sequence of process parameter adjustment instructions, i.e., the process parameter adjustment instruction set.

5. The method for optimizing and controlling process parameters of a PTFE fiber membrane production line based on the Internet of Things according to claim 4, characterized in that, The process controller collaborative computing model, based on the execution timing dependencies and potential interference between sub-instructions, collaboratively plans the execution order and timing of sub-instructions to generate a conflict-free and time-optimal sequence of process parameter adjustment instructions, including: Construct an instruction dependency graph with sub-instructions as nodes and inter-instruction dependencies as directed edges, wherein the dependencies include sequential dependencies, resource mutual exclusion, and time window constraints; On the instruction dependency graph, an instruction scheduling algorithm based on the critical path method is applied to calculate the earliest start time, latest start time, and relaxation time of each sub-instruction; Prioritize scheduling sub-instructions on the critical path and allocate them specific time slots; For sub-instructions on non-critical paths, within their relaxation time range, combined with the idle time window of the actuator, a heuristic algorithm is used for filling scheduling, with the goal of minimizing the overall process adjustment cycle. The scheduling results are converted into a sequence of instructions with timestamps, forming the conflict-free and time-optimal process parameter adjustment instruction sequence.

6. The method for optimizing and controlling process parameters of a PTFE fiber membrane production line based on the Internet of Things according to claim 5, characterized in that, The set of process parameter adjustment instructions is sent to the actuators of the extruder, drawing machine, and sintering furnace corresponding to the PTFE fiber membrane production line, driving them to perform coordinated adjustments, including: The set of process parameter adjustment instructions is converted into underlying control protocol instructions that can be recognized by the controllers of each actuator through an industrial IoT gateway; According to the timestamps in the instruction sequence, the underlying control protocol instructions are sent to the corresponding extruder controller, drawing machine controller and sintering furnace controller in a timely manner through timing event triggers; After the instruction is issued, the feedback signals of each actuator are monitored in real time. The feedback signals include the deviation between the actual execution parameters and the instruction target value. When the deviation is detected to continuously exceed the preset threshold, the adjustment anomaly handling process is triggered. The adjustment anomaly handling process includes pausing the issuance of subsequent instructions, recording the abnormal status, and sending an alarm to the upper-level system.

7. The method for optimizing and controlling process parameters of a PTFE fiber membrane production line based on the Internet of Things as described in claim 6, characterized in that, The adjusted anomaly handling process includes pausing the issuance of subsequent instructions, recording the anomaly status, and sending an alarm to the upper-level system, including: When the deviation exceeds the preset threshold, a pause signal is immediately sent to the timing event trigger to interrupt the issuance of all current and subsequent unexecuted instructions; Meanwhile, the abnormal status log records the identifier of the executing agency that caused the abnormality, the content of the abnormal instruction, the target value and the actual feedback value, as well as the timestamp of the abnormality. Based on the recorded abnormal status information, a structured abnormal alarm message is generated, which includes the abnormal type, the scope of impact, and the suggested handling measures. The abnormal alarm message is pushed to the upper-level monitoring system of the production line in real time through a preset communication link, and then waits for further instructions from the upper-level system.

8. The method for optimizing and controlling process parameters of a PTFE fiber membrane production line based on the Internet of Things according to claim 7, characterized in that, The step of updating the particle velocity in each iteration by incorporating the process coupling constraint term as an additional velocity component, which guides the particle's flight direction toward a region that satisfies the physical coupling relationship between process parameters, includes: In each iteration, based on the predefined physical correlation model between process parameters, the coupling constraint force between any two process parameters in the current particle position vector is calculated. For two process parameters that are physically correlated, the magnitude of the coupling constraint force is proportional to the difference between the current value of the two process parameters and the expected value defined by the correlation model, and the direction of the coupling constraint force is to reduce the difference. The calculated coupling constraint forces are vector synthesized for all associated process parameters in the current particle position vector to generate the total process coupling constraint force vector acting on the current particle. The additional velocity component is obtained by multiplying the total process coupling constraint force vector by the current adaptive weight coefficient. The additional velocity component is vector-superimposed with the original velocity update amount calculated based on the individual historical best position and the group historical best position to obtain the final velocity update amount of the particle in this iteration. The velocity of the current particle is updated using the final velocity update amount, and the position of the current particle is updated based on the updated velocity.

9. The method for optimizing and controlling process parameters of a PTFE fiber membrane production line based on the Internet of Things according to claim 8, characterized in that, The collaborative computing model of the process controller first performs a feasibility check on the input process parameter optimization suggestion vector. The feasibility check is based on the current state, physical limits, and adjustment rate of each actuator, including: The status database of each actuator is queried to obtain the current status parameters of the extruder, drawing machine and sintering furnace. The current status parameters include the current operating mode, current process parameter values ​​and equipment health indicators. Read the physical limit parameters of each actuator from the equipment physical limit configuration table. The physical limit parameters include the maximum working pressure and temperature range of the first-stage extruder, the maximum and minimum speeds of each stage of the drawing machine, and the maximum withstand temperature and pressure of the sintering furnace. The adjustment rate parameters of each actuator are obtained from the dynamic characteristic model of the equipment. The adjustment rate parameters include the maximum rate of change of extruder pressure, the maximum acceleration of speed adjustment of drawing machine, and the maximum rate of heating and cooling of sintering furnace. The target parameter value suggested for each actuator in the process parameter optimization suggestion vector is compared with its corresponding current state parameter, physical limit parameter and adjustment rate parameter, and boundary check, change rate feasibility check and state compatibility check are performed in sequence. If any check fails, the process parameter optimization suggestion vector is determined to be infeasible in whole or in part, and a feasibility verification failure report containing the specific reasons for failure and the actuators involved is generated based on the check results. If all checks pass, the proposed process parameter optimization vector is deemed feasible, and a flag indicating successful verification and a subset of parameters that can be safely executed are output.

Images