Intelligent agent system for high polymer material production and intelligent control method

Abstract

The invention relates to the technical field of intelligent control for high polymer material production, and discloses an intelligent agent system for high polymer material production and an intelligent control method. Obtaining the length of a reactor and the number of sensors, and axially and uniformly segmenting and mapping; performing node coordinate normalization to construct a Chebyshev orthogonal basis, and performing composite trapezoidal integration to generate and correct an inner product matrix and a projection vector; based on correction matrix augmentation and offline Gauss-Jordan inversion, extracting a spectrum expansion pre-calculation coefficient; calculating a spectral coefficient on line according to a sampling interval, reconstructing a temperature field, calculating heating power in combination with thermal parameters, and issuing and iterating in real time; and after the batch is finished, adaptively correcting the temperature curve according to the number-average molecular weight deviation to form closed-loop optimization. The method ensures that monitoring corresponds to a model, matrix numerical value is stable, operation is efficient, whole-field uniform heating and cross-batch performance are consistent, and temperature acquisition accuracy, heating efficiency and product quality are improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of intelligent control for high polymer material production, in particular to an intelligent agent system and an intelligent control method for high polymer material production. BACKGROUND

[0002] At present, the accurate control of temperature field in the production of high polymer materials is directly related to the kinetic process of polymerization reaction and the performance indicators of the final product, such as molecular weight distribution, polymerization degree and molecular chain structure, etc. In the traditional process, the reactor often adopts several temperature sensors arranged equidistantly along the axial direction, and the PID control or PI control is used to adjust the heaters of each section to maintain the pre-set temperature curve.

[0003] Firstly, the PID / PI-based controller mostly only considers single-point feedback, lacking global observation and modeling of the entire temperature field; when the reactor scale is large or the flow distribution is uneven, the feedback given by the temperature sensors at different positions is often inconsistent, leading to overheating or underheating of some heating sections, and ultimately causing the high polymer reaction rate to present oscillation or overshoot in the local, affecting the consistency and stability of the polymer quality. Secondly, although the commonly used model predictive control (MPC) can predict future temperature changes using global states and optimize multi-variable coupled reactions to some extent, its core often relies on linearization or low-order approximation models, which are difficult to accurately depict the nonlinear thermal coupling and chain growth kinetics in the high polymerization process. At the same time, the online solving of optimization problems by MPC requires high computing resources, making it difficult to achieve millisecond-level real-time control in multi-sensor, high-dimensional scenarios. Thirdly, existing academic research has proposed robust control strategies based on fuzzy logic or neural networks, using expert experience or online training models for production environments with large changes in operating conditions, but they rely on a large amount of historical data or expert knowledge accumulation, the model training process is time-consuming and difficult to explain, lacks strict mathematical reversibility guarantee, and when the operating conditions exceed the training scope, there is a risk of performance degradation or failure. In addition, in the batch production mode, traditional control methods often only focus on temperature closed-loop adjustment within a single batch, lacking cross-batch performance evaluation and self-adaptive correction mechanisms. Once the performance of a batch product (such as number-average molecular weight, polymerization degree) deviates from the target, it can only rely on manual or offline test to adjust the temperature curve for the next batch, and cannot achieve online, automatic performance feedback loop, thereby prolonging the production optimization cycle and increasing raw material and energy consumption. Finally, most existing methods still rely on empirical or coarse-grained division for the spatial discretization and modeling of the temperature field, lacking mathematical orthogonal expansion and spectral coefficient solving ideas. They fail to take advantage of the excellent approximation characteristics of orthogonal polynomials in the standard interval to efficiently and accurately reconstruct the entire field; at the same time, in key areas such as numerical integration and matrix inversion, online correction methods that can guarantee symmetry and reversibility are not used, leading to cumulative numerical errors and making it difficult to meet the requirements of high-precision control.

[0004] Therefore, this case aims to propose an intelligent system and intelligent control method for polymer material production, covering the entire process of polymer material production, realizing a complete system from spatial structure modeling, thermal parameter pre-calculation, online coefficient rapid updating to batch performance closed-loop optimization. First, the internal space of the reactor is uniformly discretized and mapped one by one with the field sensors. Then, an orthogonal function set is constructed on the standard interval and the core coefficient matrix is ​​obtained through numerical integration. After symmetry and reversibility correction, an augmented matrix is ​​generated and the expansion coefficients are pre-calculated. Next, online spectral coefficient updates are realized by combining discrete-time grids. Temperature field reconstruction and precise power allocation are completed at each monitoring node, and control commands are issued in real time iteratively. Finally, the temperature curve is dynamically corrected through an adaptive algorithm based on the product number-average molecular weight as feedback, realizing continuous optimization of performance between production batches. Summary of the Invention

[0005] This invention provides an intelligent body system and intelligent control method for the production of polymer materials, which helps to solve the problems mentioned in the background art.

[0006] This invention provides the following technical solution: an intelligent control method for polymer material production, comprising: S1. Obtain the total length of the reactor and the number of temperature sensors to be arranged, and determine the number of spatial division intervals and step size based on the two. Establish a one-dimensional coordinate system with the feed end as the origin, number each node and measure its physical position, establish a mapping between each node and the corresponding sensor number, and divide the node into several heating sections and number each section node. S2. Map the physical coordinates of the nodes to the normalized interval of 0 to 1, set the highest order of the orthogonal basis, and construct Chebyshev basis functions of each order based on the normalized coordinates to generate the basis function value matrix. S3. Calculate the inner product matrix and the projection vector of the constant term of the basis function matrix using the composite trapezoidal integral method, and perform symmetry correction and invertibility correction on the obtained inner product matrix in sequence. S4. Construct an augmented matrix based on the corrected inner product matrix, use the Gauss-Jordan elimination method to obtain the inverse matrix, and extract the pre-calculated coefficient vector for spectral expansion from the inverse matrix. S5. Discretize the production process according to the controller sampling interval, obtain the target temperature setpoint at each sampling time, and calculate the spectral expansion coefficient online based on the target temperature value and the pre-calculated coefficient vector. S6. Based on the obtained spectral expansion coefficients, the nodal temperature field is reconstructed at each sampling time, and the corresponding heating power is calculated by combining the material mass, specific heat capacity and heat exchange parameters of each heating section. S7. The calculated heating power is sent as a control command to the field heater, and the control parameters are continuously adjusted iteratively based on the real-time temperature feedback data. S8. After each production batch is completed, the number-average molecular weight of that batch is measured, the deviation from the preset target value is calculated, adaptive correction is performed based on the deviation and sensitivity coefficient, and the temperature curve setting value for the next batch is updated.

[0007] Optionally, the process of obtaining the total length of the reactor and the number of temperature sensors to be deployed, and determining the number of spatial division intervals and step size based on these two factors, establishing a one-dimensional coordinate system with the feed end as the origin, numbering each node and measuring its physical position, mapping each node to its corresponding sensor number, and dividing the reactor into several heating sections according to the nodes and numbering each section's nodes, specifically includes: Obtain the total length of the reactor, denoted as . ; Obtain the total number of temperature sensors that need to be deployed, and denote it as... ,and ; Calculate the number of spatial intervals and the step size: , ;in, It represents the number of spatial intervals; The spatial step size; Establish a one-dimensional linear coordinate system with the reactor feed end as the origin and the axis pointing towards the discharge outlet as the positive direction. ; Calculate the physical coordinates of each node index: ;in, For node indexing; For the first Each node position; Each node on site Establish a mapping table with the corresponding temperature sensor number; Set the total number of heating sections to ; For each index segment, let the index of the corresponding node segment be... ;in, Index for heating section; For the first The node index corresponding to the segment.

[0008] Optionally, the step of mapping the node physical coordinates to a normalized interval of 0 to 1, setting the highest order of the orthogonal basis, and constructing Chebyshev basis functions of various orders sequentially based on the normalized coordinates to generate a basis function value matrix specifically includes: Mapping physical coordinates to a dimensionless interval: ;in, For the first Normalized coordinates of each node; Let the highest order of the orthogonal basis be... and satisfy , ;in, It is a positive integer; Construct the first The Chebyshev basis functions are as follows: ;in, For polynomial order index; Dimensionless spatial coordinates; For the first Chebyshev polynomial function; Constructing the basis function value matrix For each pair of indexes calculate ;in, ; ; Let be the element of the basis function matrix, representing the th At the node The values ​​of the first-order basis functions; Let be a matrix space over the real number field with dimension . OK, List.

[0009] Optionally, the step of using the composite trapezoidal integral method to calculate the inner product matrix and the projection vector of the constant term of the basis function matrix respectively, and performing symmetry correction and invertibility correction on the obtained inner product matrix in sequence, specifically includes: Calculate the inner product matrix using the composite trapezoidal rule. and projection vector Specifically: ; ; in, For polynomial order index; For the basis function Rank and first The inner product of the order; Let be a square matrix space over the real number field, representing all A set of real matrices of dimension 1; For the basis function Order pair constant The projection; Let be a column vector space over the real number field, representing all vectors with dimension 1. The set of real vectors; Perform steps S310 to S320 to perform the symmetry correction branch: S310, Setting Tolerances ; S320, Traverse all : S321, if It remains unchanged; S322, Otherwise, perform forced symmetry: And record the symmetry correction mark; Perform steps S330 to S360 to perform the reversibility correction branch: S330, Calculate the determinant ; S340, Set the reversibility threshold ; S350, if ,jump over; S360, Otherwise execute Regularization, specifically: , ;in, For regularization parameters; For matrix traces; for Identity matrix.

[0010] Optionally, the step of constructing an augmented matrix based on the corrected inner product matrix, obtaining the inverse matrix using the Gauss-Jordan elimination method, and extracting the pre-calculated coefficient vector for spectral expansion from the inverse matrix specifically includes: Constructing augmented matrices ;in, For augmented matrix elements; The row number has a range of values. ; The column number has a range of values. ; Perform Gauss-Jordan elimination to find the inverse, specifically: For each primary row index : S401, Row Exchange: If Then in line Select the first one Exchange lines With Action ; S402, Principal Unity: ; S403, Column Elimination: For each row... ,make , Again ; Extracting the inverse matrix ; ; Calculate the pre-calculated coefficient vector : ;in, The first in the spectrum unfolding Pre-calculated value of the product of the order coefficient and the projection vector.

[0011] Optionally, the step of discretizing the production process according to the controller sampling interval to obtain the target temperature setpoint at each sampling time, and calculating the spectral expansion coefficients online based on the target temperature value and the pre-calculated coefficient vector, specifically includes: The total process design cycle time is obtained and denoted as... , and ensure ;in, This represents the total number of time steps. The sampling time interval of the controller; Constructing the time series: , ;in, For the first Each sampling time; Number the sampling time; Obtain the discrete values ​​of the target temperature curve, denoted as ;in, For the first The set temperature at any given time; Calculate the spectral expansion coefficients: ;in, For the first Time of the first coefficients of the order polynomial.

[0012] Optionally, based on the obtained spectral expansion coefficients, the nodal temperature field is reconstructed at each sampling time, and the corresponding heating power is calculated by combining the material mass, specific heat capacity, and heat exchange parameters of each heating section. Specifically, this includes: Reconstructing the temperature field: , , ;in, For the node ,time The reconstructed temperature; Input parameters for each segment ;in, For the first Section material quality; For the first Specific heat capacity of the section; For the first Section heat exchange coefficient; The ambient temperature; Calculate the power at the initial moment : ;in, For the first Heating power at the first moment of the segment; right Step-by-step calculation: S601, Temperature-Time Derivative: ; S602, Heating Power: .

[0013] Optionally, the step of sending the calculated heating power as a control command to the field heater and continuously adjusting the control parameters based on real-time temperature feedback specifically includes: right ,Will Issued to the on-site heaters and executed; like Then let If the condition is met, return to step S5; otherwise, proceed to step S8.

[0014] Optionally, after each production batch is completed, the number-average molecular weight of that batch is measured, the deviation from the preset target value is calculated, adaptive correction is performed based on the deviation and sensitivity coefficient, and the temperature profile setting value for the next batch is updated. Specifically, this includes: After the batch ends, at each Sampling and determination of number-average molecular weight The target number-average molecular weight is given by the process. ; Calculate the deviation ; Perform steps S801 to S803 to conduct the initial calibration: S801, Let the temperature increase be... ; S802, run the test batch and measure the batch-average molecular weight. ; S803, Calculate the sensitivity coefficient: ; Adaptive update: , ,make ;in, For the updated target temperature curve at time The set value; Set the tolerance threshold as ; For all examine: If all checkpoints are satisfied, the iteration ends; otherwise, return to step S5 and continue with the next batch of adaptive iteration.

[0015] An intelligent agent system for implementing the intelligent control method for polymer material production includes: Temperature sensor module: used to collect temperature signals from each node; Data acquisition and processing module: used to receive and preprocess sensor signals; Controller module: Used to perform online calculation and iterative control of spectral expansion coefficients based on preprocessing results; Heating execution module: used to receive control commands and drive the on-site heating equipment; Evaluation and calibration module: Used to evaluate the number-average molecular weight and adaptively update the target temperature profile after the batch is completed.

[0016] The present invention has the following beneficial effects: 1. By combining the actual length of the reactor with the number of on-site temperature sensors, the reactor is evenly divided into multiple monitoring nodes along its axis, and a corresponding sensor number is assigned to each node, establishing a one-to-one mapping between virtual model nodes and on-site equipment. This design effectively solves the problems of mismatch between model points and on-site equipment and difficulty in unifying data channels in traditional methods, enabling accurate acquisition and real-time mapping of temperature data from each monitoring node, thus improving data quality and model reliability. Compared to existing methods that only deploy sensors at a few key locations, this solution achieves comprehensive coverage of temperature information across the entire field without increasing on-site modification costs, laying a solid foundation for subsequent high-precision reconstruction and control.

[0017] 2. After mapping the physical coordinates of each monitoring node to a standard interval, a set of Chebyshev orthogonal functions is constructed according to a preset order, and the corresponding basis function value matrix is ​​generated. Unifying coordinate transformation and orthogonal function construction ensures the orthogonality and numerical regularity of the function set within the standard interval, avoiding numerical instability caused by inconsistent coordinate scales in traditional polynomial expansions. Compared to existing techniques using piecewise polynomial interpolation or simple linear interpolation, this method not only unifies the mathematical expression but also improves expansion efficiency through orthogonality, providing a high-quality set of basic functions for subsequent inner product calculations and matrix inversions, thus enhancing the accuracy and numerical stability of temperature field reconstruction.

[0018] 3. An improved composite trapezoidal numerical integration method is employed to calculate the inner product matrix between the basis function sets and the projection vectors of the basis functions onto constants, while simultaneously performing rigorous symmetry and invertibility corrections. Symmetry and condition number corrections are integrated into the post-integration processing. Numerical error adjustment strategies and regularization techniques ensure that the matrix is ​​both strictly symmetric and possesses good invertibility conditions, thus avoiding the ill-conditioned or singular problems caused by direct inversion after traditional direct integration. In practical applications, this module improves the success rate and numerical accuracy of matrix inversion, providing a reliable guarantee for subsequent augmented matrix inversion and pre-calculation of coefficients, ensuring system stability during long-cycle, high-precision computations.

[0019] 4. The corrected inner product matrix and the projection vector are merged to construct an augmented matrix, and an efficient row-column elimination algorithm is used to obtain the inverse matrix in one step. Simultaneously, the pre-calculated coefficient vector for spectral expansion is extracted from this matrix. This tightly integrates pre-calculation with linear algebraic solving, moving the most time-consuming matrix operations offline in advance. This allows for one-time calculation and storage of the coefficient vector, enabling the expansion coefficients to be obtained online with only simple queries and a few vector operations. Compared to existing online real-time solutions to entire systems of linear equations, this scheme reduces the online computational burden, meets the real-time requirements of high-frequency sampling and fast control, and improves the overall system response speed.

[0020] 5. The production process is discretized into a discrete time sequence according to the controller's sampling interval. Combined with offline pre-calculated coefficient vectors, simple operations such as vector multiplication are used to quickly update the spectral expansion coefficients online. This seamless integration of time discretization and pre-calculated data allows for high-precision spectral coefficients obtained with only minimal computational complexity during online calculations. Compared to traditional empirical models or online control based on simple proportional-integral algorithms, this method has a physical model foundation, maintaining both computational efficiency and numerical accuracy at high sampling frequencies, thus meeting the requirements for online temperature control under varying operating conditions.

[0021] 6. Temperature field reconstruction across all nodes is achieved using online-updated spectral coefficients combined with orthogonal basis functions. Precise heating power allocation is calculated based on material mass, heat capacity, and heat exchange parameters of each heating section. Temperature field modeling and power calculation are coupled into a single system. The introduction of physical thermodynamic parameters imbues the calculation results with sound engineering physics significance. Compared to methods that rely solely on single-point temperature feedback or empirical curves for power allocation, this approach enables precise power control across the entire field, effectively eliminating temperature gradients and hot spots, and improving material performance consistency and production yield.

[0022] 7. The calculated power command is sent to the field heater in real time, and the spectral coefficients and control parameters are iteratively updated based on the temperature data fed back from the field, forming a closed-loop control. An iterative framework combining physical model-driven and real-time feedback-driven approaches is constructed. When faced with sensor errors, environmental disturbances, or batch differences in materials, the system can autonomously correct its control strategy, improving control robustness and accuracy. Compared with traditional fixed-parameter or manually adjusted control methods, this solution achieves automated, adaptive, and precise control, reducing manual intervention and debugging costs.

[0023] 8. After each production batch, key performance indicators of the product samples are measured, and the deviation from the target value is calculated. Combined with a pre-calibrated sensitivity coefficient, the temperature setpoint curve for the next batch is adaptively adjusted. Production results are fed back into the algorithm loop, and control parameters are dynamically updated driven by performance deviations, achieving batch-level adaptive optimization. Compared to existing batch adjustment methods that rely solely on manual experience or fixed correction strategies, this solution, based on physical models and sensitivity analysis, can correct deviations more scientifically and quickly, continuously improving product consistency and process stability. Attached Figure Description

[0024] Figure 1 This is a schematic diagram of the process of the present invention. Detailed Implementation

[0025] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0026] Example, refer to Figure 1 A smart control method for polymer material production, comprising: S1. Obtain the total length of the reactor and the number of temperature sensors to be arranged, and determine the number of spatial division intervals and step size based on the two. Establish a one-dimensional coordinate system with the feed end as the origin, number each node and measure its physical position, establish a mapping between each node and the corresponding sensor number, and divide the node into several heating sections and number each section node. S2. Map the physical coordinates of the nodes to the normalized interval of 0 to 1, set the highest order of the orthogonal basis, and construct Chebyshev basis functions of each order based on the normalized coordinates to generate the basis function value matrix. S3. Calculate the inner product matrix and the projection vector of the constant term of the basis function matrix using the composite trapezoidal integral method, and perform symmetry correction and invertibility correction on the obtained inner product matrix in sequence. S4. Construct an augmented matrix based on the corrected inner product matrix, use the Gauss-Jordan elimination method to obtain the inverse matrix, and extract the pre-calculated coefficient vector for spectral expansion from the inverse matrix. S5. Discretize the production process according to the controller sampling interval, obtain the target temperature setpoint at each sampling time, and calculate the spectral expansion coefficient online based on the target temperature value and the pre-calculated coefficient vector. S6. Based on the obtained spectral expansion coefficients, the nodal temperature field is reconstructed at each sampling time, and the corresponding heating power is calculated by combining the material mass, specific heat capacity and heat exchange parameters of each heating section. S7. The calculated heating power is sent as a control command to the field heater, and the control parameters are continuously adjusted iteratively based on the real-time temperature feedback data. S8. After each production batch is completed, the number-average molecular weight of that batch is measured, the deviation from the preset target value is calculated, adaptive correction is performed based on the deviation and sensitivity coefficient, and the temperature curve setting value for the next batch is updated.

[0027] A complete intelligent temperature field control method is proposed, covering the entire process from on-site parameter acquisition to closed-loop calibration of production batches. First, the actual length of the reactor and the number of temperature sensors are uniformly modeled, and several monitoring nodes are evenly divided along the axis, with a corresponding sensor assigned to each node. This step ensures accurate correspondence between physical nodes and sensor data, providing reliable input for subsequent modeling. Next, the physical coordinates of the monitoring nodes are mapped to a unified standard interval, and a Chebyshev orthogonal function set is constructed. Inner product matrices and projection vectors are generated through numerical integration, and the matrices are then corrected for symmetry and invertibility, resulting in a high-quality mathematical foundation. Then, an augmented matrix is ​​constructed based on the corrected matrix and inverted offline to extract pre-calculated coefficients for spectral expansion, enabling offline large-scale matrix operations. Next, by discretizing the production process into sampling times, spectral expansion coefficients are calculated online to reconstruct the temperature field of each node and calculate the precise heating power in conjunction with material thermal parameters. Finally, heating power commands are issued in real time and iterative control is implemented based on temperature feedback. After the production batch ends, the set curve is adaptively corrected based on sample performance deviations, forming a closed-loop optimization system from modeling to quality feedback. By establishing the steps of reactor spatial discretization and sensor mapping, the problems of mismatched monitoring points and incomplete temperature data in traditional control were solved, ensuring the comprehensiveness and accuracy of temperature acquisition. Through the steps of node coordinate standardization and orthogonal function construction, numerical instability caused by coordinate scale differences was avoided, achieving a stable and efficient mathematical foundation. By correcting the symmetry and invertibility of the inner product matrix, matrix ill-conditioning caused by numerical integration errors was overcome, improving the reliability of inversion. By constructing an augmented matrix and pre-calculating spectral coefficients offline, the online computational burden was significantly reduced, meeting real-time control requirements. Through temperature field reconstruction and power calculation, uniform heating and precise power distribution across the entire field were achieved, improving heating efficiency. Through real-time iteration and batch feedback correction, the problems of performance fluctuations and poor process consistency between production batches were solved, enhancing product quality stability and production efficiency.

[0028] The process involves obtaining the total length of the reactor and the number of temperature sensors to be deployed, determining the number of spatial division intervals and step size based on these two factors, establishing a one-dimensional coordinate system with the feed end as the origin, numbering each node and measuring its physical location, mapping each node to its corresponding sensor number, and dividing the reactor into several heating sections according to the nodes and numbering each section's nodes. Specifically, this includes: Obtain the total length of the reactor, denoted as . ; Obtain the total number of temperature sensors that need to be deployed, and denote it as... ,and ; Obtain the process template and sensor configuration to provide basic parameters for subsequent spatial discretization; Calculate the number of spatial intervals and the step size: , ;in, It represents the number of spatial intervals; The spatial step size; Establish a one-dimensional linear coordinate system with the reactor feed end as the origin and the axis pointing towards the discharge outlet as the positive direction. ; Calculate the physical coordinates of each node index: ;in, For node indexing; For the first Node locations; generate the physical location of each monitoring node. ; Each node on site Establish a mapping table with the corresponding temperature sensor number; ensure that the virtual model nodes correspond one-to-one with the field devices to facilitate data reading and command issuance; Set the total number of heating sections to ; For each index segment, let the index of the corresponding node segment be... ;in, Index for heating section; For the first The node index corresponding to the segment is determined; the correspondence between the heater segment and the temperature node is clarified to locate the position for subsequent power calculation.

[0029] Installation Precautions: The sensor must have good thermal contact with the reactor wall to ensure that the measured value reflects the true temperature of the fluid or catalyst bed; the length and position of the sensor probe must ensure that the probe tip is exactly located along the calibrated axis. Location; each sensor number should be mapped one-to-one with a control system node, and temperature and power commands should be read and issued according to the mapping table.

[0030] The spatial discretization and sensor mapping process in the first step was further refined. It determines the spatial segmentation intervals and step sizes by obtaining the total length of the reactor and the number of temperature sensors to be deployed. A one-dimensional coordinate system is established at the reactor feed end, each monitoring node is numbered and its physical location is measured. Each node is then mapped one-to-one with its sensor number, and finally, heating sections are divided according to the nodes, with each section independently numbered. This step tightly integrates the physical space and control unit, laying the foundation for subsequent segmented heating and power calculation. By accurately obtaining the reactor geometry and sensor layout on-site and converting them into a unified model node-equipment mapping, the measurement error problem caused by the mismatch between model nodes and on-site devices in traditional methods is solved, ensuring accurate correspondence between temperature acquisition and control commands. By flexibly setting the heating section and node numbering steps, the problems of difficult heating power positioning and inaccurate segmented control are solved, allowing each heater segment to be independently adjusted based on its own regional node data, thereby improving the temperature uniformity of the heating area and the accuracy of segmented control. The standardized modeling process simplifies the matching process between on-site installation and the control system, reduces commissioning costs, and improves engineering implementation efficiency and on-site maintenance convenience.

[0031] The process of mapping node physical coordinates to a normalized interval of 0 to 1, setting the highest order of the orthogonal basis, and constructing Chebyshev basis functions of various orders sequentially based on the normalized coordinates to generate a basis function value matrix specifically includes: Mapping physical coordinates to a dimensionless interval: ;in, For the first Normalized coordinates of each node; standardize node positions to... This facilitates orthogonal polynomial expansion; Let the highest order of the orthogonal basis be... and satisfy , ;in, It is a positive integer; it determines the order of the polynomial expansion, balancing precision and computational cost. Construct the first The Chebyshev basis functions are as follows: ;in, For polynomial order index; Dimensionless spatial coordinates; For the first Chebyshev polynomial functions of order; construct a set of functions in order. Orthogonal basis functions within the interval are used to represent the temperature field; Constructing the basis function value matrix For each pair of indexes calculate ;in, ; ; Let be the element of the basis function matrix, representing the th At the node The values ​​of the first-order basis functions; Let be a matrix space over the real number field with dimension . OK, The column stores the values ​​of the basis functions at each node in discrete form for use in spectral expansion.

[0032] By mapping the coordinates of monitoring nodes to a standardized interval and constructing a set of orthogonal functions, the problem of poor numerical conditions and ill-conditioned matrices in polynomial expansion caused by inconsistent physical coordinate ranges is solved, ensuring good orthogonality of basis functions within the standardized interval. The step of generating the basis function value matrix provides a high-quality discrete representation foundation for subsequent inner product calculations and temperature field reconstruction, improving the accuracy and numerical stability of spectral expansion. Overall, this mechanism makes the temperature field more representative, enabling more accurate capture of the details of the internal temperature distribution of the reactor, and providing a reliable mathematical basis for precise control and power allocation.

[0033] The method of using the composite trapezoidal integral to calculate the inner product matrix and the projection vector of the constant term of the basis function matrix, and then performing symmetry correction and invertibility correction on the resulting inner product matrix in sequence, specifically includes: Calculate the inner product matrix using the composite trapezoidal rule. and projection vector Specifically: ; ; in, For polynomial order index; For the basis function Rank and first The inner product of the order; Let be a square matrix space over the real number field, representing all A set of real matrices of dimension 1; For the basis function Order pair constant The projection; Let be a column vector space over the real number field, representing all vectors with dimension 1. The set of real vectors; numericalize the inner product of basis functions and the projection of basis functions onto constants to form a matrix form of linear equations for solving spectral coefficients; Perform steps S310 to S320 to perform the symmetry correction branch: S310, Setting Tolerances ;like Taking values ​​that are too large will ignore significant asymmetric numerical errors, leading to... Slight asymmetry may affect accuracy during inversion; excessively small values ​​can trigger frequent corrections for otherwise sufficiently symmetric matrices, increasing unnecessary computations; based on numerical stability and computational accuracy tests, a value of [value missing] can be used initially. If the bit width or noise is greater, then appropriately enlarge it. ; S320, Traverse all : S321, if It remains unchanged; S322, Otherwise, perform forced symmetry: And record the symmetry correction mark; Correct the asymmetric error caused by numerical integration to ensure that the matrix is ​​strictly symmetric; Perform steps S330 to S360 to perform the reversibility correction branch: S330, Calculate the determinant ; S340, Set the reversibility threshold ;like If the value is too large, the matrix will be prematurely judged as "ill-conditioned," leading to frequent regularization and affecting the physical accuracy of the solution; if the value is too small, the actual ill-conditioned matrix may be missed, resulting in inversion failure or loss of precision; a value slightly higher than the machine's double-precision range is recommended. The magnitude, such as It can be fine-tuned according to the actual matrix condition number distribution; S350, if ,jump over; S360, Otherwise execute Regularization, specifically: , ;in, For regularization parameters; For matrix traces; for identity matrix; To avoid matrix singularities or numerical instability, ensure the feasibility of subsequent reversible operations.

[0034] This paper describes the steps for calculating the inner product matrix and projection vector using numerical integration methods, and performing symmetry and invertibility corrections on the matrices. The process employs the composite trapezoidal rule to numerically calculate the cross-products between basis function sets and the projections onto constant functions. Asymmetry biases are then eliminated through error detection and forced symmetry correction. Subsequently, condition numbers or determinants are used to detect and regularize ill-conditioned matrices. The numerical calculation steps using the composite trapezoidal rule for inner products and projections address the insufficient accuracy of traditional discrete integration, ensuring the accuracy and reliability of matrix element values. The symmetry correction step eliminates minor asymmetry errors generated by numerical integration, guaranteeing strict matrix symmetry and providing a correct symmetric structure for subsequent solutions. The invertibility correction (regularization) step effectively avoids the risk of inversion failure or unstable results due to occasional ill-conditioned or singular matrices, ensuring the reliability of matrix operations and the success rate of inversion. Overall, this step improves the stability and reliability of numerical calculations, providing solid support for spectral coefficient pre-calculation and thus ensuring the numerical robustness of the entire control algorithm.

[0035] The process of constructing an augmented matrix based on the corrected inner product matrix, obtaining the inverse matrix using the Gauss-Jordan elimination method, and extracting the pre-calculated coefficient vector for spectral expansion from the inverse matrix specifically includes: Constructing augmented matrices ;in, For augmented matrix elements; The row number has a range of values. ; The column number has a range of values. Merge the target matrix and the identity matrix to prepare for the Gauss-Jordan inversion; Perform Gauss-Jordan elimination to find the inverse, specifically: For each primary row index : S401, Row Exchange: If Then in line Select the first one Exchange lines With Action ; S402, Principal Unity: ; S403, Column Elimination: For each row... ,make , Again ; Transform the augmented matrix into the following form using elementary row operations: The form is used to obtain the inverse matrix; Extracting the inverse matrix ; Read from the right half of the augmented matrix ; Calculate the pre-calculated coefficient vector : ;in, The first in the spectrum unfolding Pre-calculated values ​​of the product of the order coefficients and the projection vector; pre-calculated product of the spectral coefficients and the projection vector to accelerate online computation.

[0036] The focus is on constructing an augmented matrix based on a corrected inner product matrix and inverting it using an efficient elimination algorithm, as well as extracting pre-calculated coefficient vectors for spectral expansion from the inverse matrix. This step completes the core matrix operations offline and reduces the burden of subsequent online computation through a one-time inversion and coefficient extraction method. By constructing the augmented matrix and using the elimination method for one-time inversion, the problem of high computational load and slow response speed when solving large-scale linear equation systems online is solved, enabling complex matrix operations to be completed offline and improving the overall system response efficiency. The step of pre-calculating spectral expansion coefficients avoids repeated matrix operations online, requiring only simple vector multiplication to obtain the spectral coefficients, meeting the real-time requirements of high-frequency online control. This mechanism reduces the consumption of online computing resources, enabling high-precision spectral expansion and temperature reconstruction in embedded or real-time control environments, providing feasibility for industrial applications of the system.

[0037] The process of discretizing the production process according to the controller sampling interval to obtain the target temperature setpoint at each sampling moment, and calculating the spectral expansion coefficients online based on the target temperature value and the pre-calculated coefficient vector, specifically includes: The total process design cycle time is obtained and denoted as... , and ensure ;in, This represents the total number of time steps. Define the sampling time interval for the controller; define a time discrete grid to ensure that the step size is divisible by the period; Constructing the time series: , ;in, For the first Each sampling time; Number the sampling time points; specify the time for each sampling point; Obtain the discrete values ​​of the target temperature curve, denoted as ;in, For the first Set temperature at any time; provide target temperature signal for online control; Calculate the spectral expansion coefficients: ;in, For the first Time of the first Multinomial coefficients of various orders; quickly generate polynomial coefficients for temperature reconstruction.

[0038] This paper describes the steps involved dividing the production process into multiple controller sampling times and calculating the corresponding spectral expansion coefficients online based on offline pre-calculated coefficients and the target temperature setpoint for each time point. This design tightly integrates time discretization with pre-calculation results, minimizing the online computational burden. By discretizing the production cycle according to sampling intervals and obtaining the target temperature setpoint, the delay problem caused by the asynchrony between the target signal and the model input in online control is solved, ensuring accurate matching between control commands and target temperatures. By generating spectral coefficients online based on the target temperature and pre-calculated coefficients, the complexity of online computation is reduced, enabling rapid spectral coefficient updates even at high sampling frequencies. This step effectively balances computational efficiency and control accuracy, providing technical support for high-frequency online temperature control and ensuring rapid system response under various operating conditions.

[0039] Based on the obtained spectral expansion coefficients, the nodal temperature field is reconstructed at each sampling time, and the corresponding heating power is calculated by combining the material mass, specific heat capacity, and heat exchange parameters of each heating section. Specifically, this includes: Reconstructing the temperature field: , , ;in, For the node ,time The reconstructed temperature is obtained; the temperature of each node is reconstructed based on the spectral expansion coefficients and basis functions; Input parameters for each segment ;in, For the first Section material quality; For the first Specific heat capacity of the section; For the first Section heat exchange coefficient; Provides ambient temperature; provides kinetic and thermodynamic parameters for power calculation; Calculate the power at the initial moment : ;in, For the first The heating power of the segment at the first moment; the first power command is estimated using the temperature change in the first step; right Step-by-step calculation: S601, Temperature-Time Derivative: ; S602, Heating Power: ; The real-time heating power is calculated based on the gradient of the reconstructed temperature and the heat loss.

[0040] This paper describes the steps involved in reconstructing the temperature field at each sampling time using spectral coefficients calculated online, and then calculating the corresponding heating power based on the thermal parameters of the materials in each heating section. The reconstruction results directly determine the power allocation, achieving precise regional heating. By combining spectral coefficients with orthogonal basis functions to reconstruct the temperature field, the paper solves the problem that traditional point-based control cannot reflect the overall temperature distribution, achieving accurate restoration of the temperature inside the entire reactor. By combining material mass, specific heat capacity, and heat exchange characteristics to calculate the heating power, the paper addresses the problem of ignoring process differences during power allocation, which can lead to uneven heating or overheating risks, enabling differentiated power output for different regions and material states. This process improves the accuracy and efficiency of temperature control, reduces energy consumption, and ensures product consistency across different regions, thereby improving the quality stability of the final product.

[0041] The process of sending the calculated heating power as a control command to the field heater and continuously adjusting the control parameters based on real-time temperature feedback specifically includes: right ,Will The command is sent to the field heaters and executed; the heating power of each section is controlled in real time to track the target temperature. like Then let If the time step ends, return to step S5; otherwise, proceed to step S8; thus forming a closed-loop iteration from the time step to the end of the batch.

[0042] This paper elaborates on the control command issuance and real-time iteration process. It sends the calculated heating power command to the field heater in real time and continuously adjusts the control parameters based on the temperature data fed back from the field, achieving closed-loop iterative control. By issuing and executing the calculated power command in real time, the problem of temperature overshoot or lag caused by command delay and asynchronous execution is solved, ensuring immediate matching between the heating command and the field response. The step of iteratively adjusting the spectral coefficient and control parameters based on the feedback temperature data enables the control system to adaptively correct for environmental disturbances and sensor errors, enhancing the system's robustness. This closed-loop iterative mechanism improves the stability and accuracy of temperature control, reduces manual intervention, and lowers the risk of fluctuations in the production process.

[0043] The process of measuring the number-average molecular weight of each production batch after completion, calculating its deviation from the preset target value, performing adaptive correction based on the deviation and sensitivity coefficient, and updating the temperature profile setting for the next batch specifically includes: After the batch ends, at each Sampling and determination of number-average molecular weight The target number-average molecular weight is given by the process. Compare the product's core performance with the target specifications; Calculate the deviation Quantify the degree to which batch performance deviates from the target; Perform steps S801 to S803 to conduct the initial calibration: S801, Let the temperature increase be... The temperature increment uniformly added to the target temperature curve during the initial calibration; if Taking a value that is too small may result in the measured value being too small. A low signal-to-noise ratio leads to inaccurate sensitivity calculations; an excessively large value may alter the polymerization mechanism and compromise the linear approximation premise; recommended value: based on process experience, generally set to [value missing]. This ensures that the detection signal is identifiable without causing drastic changes in the response; S802, run the test batch and measure the batch-average molecular weight. ; S803, Calculate the sensitivity coefficient: ; The sensitivity of the calibrated temperature to the performance is used for subsequent adaptive testing; Adaptive update: , ,make ;in, For the updated target temperature curve at time The set value; Set the tolerance threshold as The allowable absolute deviation range of number-average molecular weight; if A small value will lead to multiple corrections and low production efficiency; a large value may fail to meet product quality requirements. Generally, a target value is chosen based on a combination of quality standards and experimental data. or absolute deviation ; For all examine: If all checkpoints are satisfied, the iteration ends; otherwise, return to step S5 and continue with the next batch of adaptive iteration. Batch closed-loop optimization is achieved by dynamically correcting the target temperature based on performance deviations.

[0044] This paper describes the steps involved in updating the temperature profile for the next batch after each production batch. These steps involve adaptive correction based on the deviation of the product's number-average molecular weight from the target value, combined with a sensitivity coefficient, thus creating a batch-level closed-loop optimization. By measuring key product performance and calculating deviations after each batch, the paper addresses the problem of traditional methods relying solely on empirical parameter adjustments failing to quantify performance differences, ensuring the objectivity and reliability of deviation data. The adaptive correction step, incorporating sensitivity analysis, overcomes the challenge of accurately improving performance deviations with single, small adjustments, achieving a scientific quantification of the impact of temperature. Applying the corrected temperature profile to the next batch creates a closed loop of action-result-correction, continuously reducing inter-batch performance fluctuations and providing a continuously optimized process loop for polymer production, effectively improving product consistency and process stability.

[0045] This embodiment also provides an intelligent agent system for an intelligent control method in polymer material production, including: Temperature sensor module: used to collect temperature signals from each node; Data acquisition and processing module: used to receive and preprocess sensor signals; Controller module: Used to perform online calculation and iterative control of spectral expansion coefficients based on preprocessing results; Heating execution module: used to receive control commands and drive the on-site heating equipment; Evaluation and calibration module: Used to evaluate the number-average molecular weight and adaptively update the target temperature profile after the batch is completed.

[0046] A smart agent system architecture was constructed to implement the aforementioned intelligent temperature field control method. The system includes a temperature sensor module, a data acquisition and processing module, a controller module, a heating execution module, and an evaluation and calibration module, covering the entire process from data acquisition, online calculation, command execution, and batch evaluation. Through the integrated multi-functional module system design, the interface adaptation and data flow problems when interfacing complex algorithms with field devices were solved, achieving seamless integration of each stage. The preprocessing of sensor signals by the data acquisition and processing module ensured the reliability of the controller input data and reduced the impact of noise interference on temperature control. The collaborative operation of the controller and execution module ensured the stability of high-frequency real-time spectral coefficient calculation and command issuance. The performance feedback mechanism of the evaluation and calibration module ensured batch-level adaptive optimization. The overall system design enables one-click deployment of the temperature control method in industrial settings, reducing engineering integration difficulty and improving the automation and intelligence level of the production process.

[0047] It should be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus.

[0048] The above description is only a preferred embodiment of the present invention. It should be noted that for those skilled in the art, several improvements and modifications can be made without departing from the technical principles of the present invention, and these improvements and modifications should also be considered within the scope of protection of the present invention.

Claims

1. A method for intelligent control of production of high molecular materials, characterized in that, The method comprises the following steps: S1, obtaining the total length of the reactor and the number of temperature sensors to be arranged, determining the number of space division intervals and the step length based on the two, establishing a one-dimensional coordinate system with the feed end as the origin, numbering and measuring the physical positions of the nodes, mapping each node to the corresponding sensor number, and dividing a plurality of heating sections according to the nodes and numbering the nodes in each section; S2, mapping the node physical coordinates to the normalized interval of 0 to 1, setting the highest order of the orthogonal basis, and sequentially constructing each order Chebyshev basis function based on the normalized coordinates to generate a basis function value matrix; S3, calculating the inner product matrix and constant term projection vector of the basis function matrix respectively by using the composite trapezoidal integration method, and sequentially performing symmetry correction and reversibility correction on the obtained inner product matrix; S4, constructing an augmented matrix based on the corrected inner product matrix, solving the inverse matrix by using the Gauss-Jordan elimination method, and extracting a pre-computed coefficient vector for spectral expansion from the inverse matrix; S5, time-discretizing the production process according to the controller sampling interval, obtaining the target temperature set value at each sampling time, and calculating the spectral expansion coefficient based on the target temperature value and the pre-computed coefficient vector; S6, based on the obtained spectral expansion coefficient, reconstructing the node temperature field at each sampling time, and calculating the corresponding heating power in combination with the material mass, specific heat capacity and heat exchange parameters of each heating section; S7, issuing the calculated heating power to the on-site heater as a control instruction, and iteratively adjusting the control parameters according to the real-time feedback temperature data; S8, measuring the number-average molecular weight of each production batch, calculating the deviation from the preset target value, performing adaptive correction according to the deviation and the sensitivity coefficient, and updating the temperature curve set value for the next batch. 2.The intelligent control method for high polymer material production of claim 1, wherein, The method comprises the following steps: The total length of the reactor is obtained, noted Ltot ; The total number of temperature sensors required to be arranged is obtained, denoted by and ; Computing the spatial bin number and step size: , ; wherein, is the spatial bin number; is the spatial step size; A one-dimensional straight coordinate system is established with the reactor feed end as the coordinate origin and the direction along the axis to the discharge port as the positive direction ; For each node index, compute the node physical coordinates: ; wherein, is the node index; is the th node position; At the site, a mapping table is established for each node with the corresponding temperature sensor number; The total number of heating sections is set to ; For each segment index, let the segment corresponding node index ; wherein, is the heating segment index; is the first segment corresponding node index. 3.The intelligent control method for high polymer material production of claim 2, wherein, The method comprises the following steps: mapping physical coordinates to a dimensionless interval: ; wherein, is the normalized coordinate of the th node; Let the highest order of the orthogonal basis be and satisfy , ; wherein, is a positive integer; Constructing the first Chebyshev basis functions, in particular: ; wherein, is a polynomial order index; is a dimensionless spatial coordinate; is a Chebyshev polynomial function of the th order; Constructing the basis function value matrix : for each pair of indices compute ; where ; ; is the basis function matrix element representing the value of the th basis function at the th node; is the space of matrices over the field of real numbers with rows, columns.

4. The intelligent control method for the production of high molecular materials according to claim 3, characterized in that, The method comprises the following steps: The inner product matrix and the projection vector are calculated by using the compound trapezoidal method, specifically: ； ； wherein, is a polynomial order index; is an inner product of the basis functions of order and order is a space of square matrices over the field of real numbers, representing the set of all dimensional real matrices; is a projection of the basis functions of order onto the constant is a space of column vectors over the field of real numbers, representing the set of all dimensional real vectors;​​ The method comprises the following steps: S310, set the tolerance ; S320, traverse all : S321、If , remains unchanged; S322, otherwise enforce symmetry: and record the symmetry correction flag; The method comprises the following steps: S330, calculating the determinant ; S340, setting a reversibility threshold ; S350、if , skip; S360, else perform regularization, in particular: , ; wherein, is a regularization parameter; is the trace of the matrix ; is the identity matrix.

5. The intelligent control method for the production of high molecular materials according to claim 4, characterized in that, The method comprises the following steps: Constructing an augmented matrix ; wherein, is an augmented matrix element; is a row number, taking values in the range ; is a column number, taking values in the range ; The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the for each pivot row index : S401, row exchange: if , select the first in row , exchange row with row ; S402, main element normalization: ; S403, Column elimination: for each row , let , Again ; Extracting inverse matrix ; ; Computing a pre-computed coefficient vector : ; wherein, is a product of a th coefficient of a spectral expansion and a projection vector pre-computed value.

6. The intelligent control method for the production of high molecular materials according to claim 5, characterized in that, The production process is time-discretized according to the controller sampling interval, the target temperature set value at each sampling time is obtained, and the spectral expansion coefficient is calculated online based on the target temperature value and the pre-calculated coefficient vector, specifically including: The total length of the process design cycle is obtained, denoted as , , and ensuring ; wherein is the total number of time steps; is the sampling time interval of the controller. Sequence of construction instants: , ; wherein, is the sampling instant; is the sampling instant number; The discrete values of the target temperature curve are obtained, denoted as ; wherein, is the set temperature at the moment; Computing the spectral expansion coefficients: ; where, is the nth polynomial coefficient at the mth time instant.

7. The intelligent control method for high polymer material production according to claim 6, wherein, The node temperature field is reconstructed at each sampling time based on the obtained spectral expansion coefficient, and the corresponding heating power is calculated in combination with the material mass, specific heat capacity and heat exchange parameters of each heating section, specifically including: reconstructing the temperature field: , , ; wherein is the reconstructed temperature at the node at the time instance. Input parameters for each section ; wherein, is the mass of the material in the th section; is the specific heat capacity of the material in the th section; is the heat exchange coefficient of the material in the th section; is the ambient temperature; Calculate the power at the initial moment : ;in, For the first Heating power at the first moment of the segment; To Stepwise calculation: S601, temperature time derivative: ; S602, heating power: . 8.The intelligent control method for high polymer material production of claim 7, wherein, The calculated heating power is sent to the field heater as a control instruction, and the control parameters are continuously iteratively adjusted according to the real-time feedback temperature data, specifically including: To The is issued to the field heater and executed; If , then let , return to step S5; otherwise, go to step S8. 9.The intelligent control method for high polymer material production of claim 8, wherein, After each production batch, the number-average molecular weight is measured, the deviation from the preset target value is calculated, the adaptive correction is performed according to the deviation and the sensitivity coefficient, and the temperature curve set value of the next batch is updated, specifically including: After the end of the batch, a sample is taken every The number average molecular weight is determined on the sample and the target number average molecular weight is given by the process ; Computing bias ; Steps S801 to S803 are performed for the first calibration: S801, set the temperature increasing amount as ; S802, run the test batch, measure the batch number average molecular weight ; S803, calculate the calculation sensitivity coefficient: ; Adaptive update: , , let ; wherein, is the set value of the updated target temperature curve at time ; Setting a tolerance threshold to ; for all checks: If all checks are satisfied, end iteration; otherwise, return to step S5 and continue with the next batch of adaptation.

10. An agent system employing the intelligent control method for the production of high molecular materials according to claim 9, characterized in that, It includes: Temperature sensor module: for collecting node temperature signals; Data acquisition and processing module: for receiving and preprocessing sensor signals; Controller module: for performing spectral expansion coefficient online calculation and iterative control according to the preprocessing result; Heating execution module: for receiving control commands and driving field heating equipment; Evaluation and correction module: for evaluating the number-average molecular weight after the batch and adaptively updating the target temperature curve.

Images