Multi-parameter coupling optimization system and method for spinning bath

Abstract

The invention relates to the technical field of data processing, in particular to a multi-parameter coupling optimization system and method for a spinning bath, and the system comprises a data analysis module which obtains a multi-parameter combination in historical data, divides a process window boundary, calculates a robustness index, divides parameter space sub-regions according to the process window boundary and the robustness index, and outputs the parameter space sub-regions; and calculating the influence sensitivity of the stabilizable index to obtain a parameter importance measurement value. The numerical solution module is used for establishing a technological process mathematical equation set, microstructure characteristic parameters and defect precursor information are obtained according to the multi-parameter combined solution equation set, and the parameter evaluation module is used for obtaining a stabilization index and a defect probability. And the iterative operation module is used for generating a to-be-evaluated multi-parameter combination set according to the parameter space sub-region and the parameter importance measurement value and a differentiation mechanism, establishing a function group according to a stabilization index and a defect probability, and performing evaluation and iterative operation on the to-be-evaluated multi-parameter combination set to obtain an optimal multi-parameter combination.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of data processing, in particular to a spinning bath multi-parameter coupling optimization system and method. BACKGROUND

[0002] In modern manufacturing industry, industrial information and data processing technology has become the key support to improve production efficiency and product quality. With the enrichment of data acquisition means and the improvement of computing power, digital computing methods show great potential in the field of multi-parameter optimization. These methods process high-dimensional data by establishing complex mathematical models, solve partial differential equations using numerical analysis techniques, simplify complex calculations using matrix operations, and quantify parameter effects using function evaluation tools. Currently, statistical analysis is used to find patterns in industrial data, multi-objective optimization techniques are used to find the best balance point, and data preprocessing methods are used to ensure the reliability of the calculation basis. This mathematical-based computing method has been widely used in parameter optimization problems in the fields of chemical industry, materials, energy, etc., and has shown significant advantages in dealing with nonlinear relationships and multi-parameter coupling. Industrial information and data processing systems combine these mathematical tools with actual process knowledge, through logical operations and structured analysis, to ensure process quality while improving production flexibility.

[0003] However, in the preparation process of polyacrylonitrile fibers, existing digital computing methods face special challenges. The high-dimensional nonlinear relationship between multiple parameters is difficult to accurately express through conventional mathematical functions; the contribution of parameters to the calculation results varies significantly, but there is a lack of effective mathematical tools for evaluation; different regions within the parameter space exhibit different mathematical characteristics, and existing algorithms use a unified calculation strategy that cannot be adjusted adaptively; for potential data outliers and deviation trends, the analysis method lacks an effective prediction mechanism. These problems constrain the application effect and calculation efficiency of related numerical optimization methods.

[0004] Therefore, a spinning bath multi-parameter coupling optimization system and method are proposed. SUMMARY

[0005] The application aims to provide a spinning bath multi-parameter coupling optimization system and method, which forms a spinning bath multi-parameter coupling optimization system through a data analysis module, a numerical solution module, a parameter evaluation module and an iterative operation module.

[0006] To achieve the above-mentioned purpose, the application provides the following technical scheme. A spinning bath multi-parameter coupling optimization system comprises: A data analysis module divides a process window boundary according to a multi-parameter combination in historical data, calculates a robustness index according to the process window boundary, divides a parameter space sub-region according to the process window boundary and the robustness index, and calculates an influence sensitivity of a stabilizability index of each spinning bath parameter to obtain a parameter importance metric value. A numerical solution module establishes a process mathematical equation set, and solves the equation set according to a multi-parameter combination to obtain microstructure characteristic parameters and defect precursor information. A parameter evaluation module calculates a stabilizability index according to the microstructure characteristic parameters, and converts the defect precursor information into a defect probability. An iterative operation module generates a multi-parameter combination set to be evaluated according to the parameter space sub-region and the parameter importance metric value using a differentiation mechanism, establishes a function group according to the stabilizability index and the defect probability, evaluates the multi-parameter combination set to be evaluated using the function group, and obtains an optimal multi-parameter combination through iterative operation.

[0007] Preferably, the process window boundary includes a feasible region process window boundary, a transition region process window boundary and a failure region process window boundary, the step of dividing the process window boundary is: obtaining a plurality of parameter combinations in historical data and corresponding spinning result effectiveness, using a support vector classification method to construct a discriminant function, determining the boundary line between the feasible region and the failure region in the overall parameter space based on the discriminant function; setting the boundary range of the transition region between the feasible region and the failure region; calculating the gradient distribution of the discriminant function in the overall parameter space; correcting the boundary shape according to the gradient size and direction; outputting the determined feasible region process window boundary, transition region process window boundary and failure region process window boundary.

[0008] Preferably, a robustness index is calculated according to the parameter space distance of each multi-parameter combination from the feasible region process window boundary and the stability score of the multi-parameter combination; based on the process window boundary and the robustness index, the overall parameter space is divided into a plurality of sub-regions by combining a density clustering method and a Markov random field, the geometric characteristics and the distribution characteristics of the robustness index of each sub-region are analyzed to obtain a parameter space sub-region; the overall parameter space is the value range of each spinning bath parameter.

[0009] Preferably, the numerical solving module specifically includes: establishing a process mathematical equation set of flow field equation, temperature field equation, concentration field equation and phase separation equation; determining the calculation grid and boundary conditions, and discretely solving the process mathematical equation set along the time axis by using the finite element method to obtain the flow field, temperature field, concentration field and phase separation field distribution data at each time point; extracting microstructure characteristic parameters and defect precursor information based on the distribution data; wherein the microstructure characteristic parameters include average pore size, pore size distribution standard deviation, skin layer density, skin layer thickness, core layer uniformity parameter and molecular chain orientation degree; the defect precursor information includes defect precursor density and defect precursor average size.

[0010] Preferably, the parameter evaluation module specifically includes: a nonlinear mapping model based on a deep neural network, used to establish a mapping relationship between the microstructure characteristic parameters and the stabilizability index; a statistical regression model, used to convert the defect precursor density and the defect precursor average size into a defect probability value.

[0011] Preferably, the parameter space sub-region includes a first space sub-region, a second sub-space sub-region and a third space sub-region; the differentiation mechanism is: using a first optimization algorithm for the first space sub-region, using a second optimization algorithm for the second sub-space sub-region, and using a third optimization algorithm for the third space sub-region, adjusting the search rule of the spinning bath parameters according to the parameter importance measurement value, and generating the set of multi-parameter combinations to be evaluated.

[0012] Preferably, the function group is used to evaluate the to-be-evaluated multi-parameter combination set, and the optimal multi-parameter combination is obtained through iterative operation, and the optimal multi-parameter combination specifically comprises: constructing a multi-objective function group comprising maximized stabilizable index and minimized defect probability; using a Pareto sorting method to sort the to-be-evaluated multi-parameter combination to form different Pareto levels; selecting the multi-parameter combination of the first Pareto level and the multi-parameter combination of other levels with a comprehensive score higher than a preset threshold to be reserved to the next round of iteration; updating the process window boundary and the robustness index, redividing the parameter space sub-region according to the updated process window boundary and the robustness index and recalculating the parameter importance measure value; generating a new to-be-evaluated multi-parameter combination based on the updated parameter space sub-region and the parameter importance measure value; judging whether an iteration termination condition is met; if the termination condition is met, outputting the optimal multi-parameter combination, and if the termination condition is not met, continuing the iterative operation.

[0013] The application further provides a spinning bath multi-parameter coupling optimization method for executing a spinning bath multi-parameter coupling optimization system, comprising: Dividing a process window boundary according to a multi-parameter combination in historical data, calculating a robustness index according to the process window boundary, dividing a parameter space sub-region according to the process window boundary and the robustness index, and calculating the influence sensitivity of each spinning bath parameter on a stabilizable index to obtain a parameter importance measure value; Establishing a process mathematical equation group, solving the equation group to obtain microstructure characteristic parameters and defect precursor information according to a multi-parameter combination; Calculating a stabilizable index according to the microstructure characteristic parameters and converting the defect precursor information into a defect probability; Generating a to-be-evaluated multi-parameter combination set using a differentiated mechanism according to the parameter space sub-region and the parameter importance measure value, establishing a function group according to the stabilizable index and the defect probability, using the function group to evaluate the to-be-evaluated multi-parameter combination set, and obtaining an optimal multi-parameter combination through iterative operation.

[0014] Compared with the prior art, the application has the following beneficial effects: 1、The present application accurately divides the feasible region, transition region and failure region three types of regions in the overall parameter space, and introduces the concept of robustness index, calculates the parameter space distance of each parameter combination and the boundary of the feasible region and its stability score, and quantitatively evaluates the process stability of the parameter combination. This quantitative method can intuitively judge the stability and anti-disturbance ability of multi-parameter combination in actual production. Further, according to the process window boundary and the robustness index distribution, the overall parameter space is divided into sub-regions with different characteristics, which lays a foundation for the subsequent differentiated optimization strategy. The present application can dynamically identify the change of the process window, accurately describe the complex characteristics of the parameter space, effectively avoid the risk of optimization process falling into unstable region, provide reliable parameter space cognition ability, significantly improve the accuracy and stability of the multi-parameter coupling optimization of the spinning bath, and effectively solve the accurate control problem of microstructure and stabilization characteristics.

[0015] 2、The present application accurately quantifies the importance of each parameter by calculating the influence sensitivity of each spinning bath parameter on the stabilizability index, forms a parameter importance measure value, realizes intelligent adjustment of the parameter search rule, and improves the efficiency of parameter optimization. Combined with the parameter space sub-region division result, the present application proposes a differentiated optimization mechanism, which selects the most suitable optimization algorithm for parameter space sub-regions with different characteristics, significantly improving the search efficiency of the algorithm in various regions. In the iterative optimization process, the process window boundary and the robustness index are updated, and the parameter importance measure value and the sub-region division are dynamically adjusted, so that the optimization process can adaptively track the change characteristics of the parameter space. The parameter importance measure and the differentiated optimization strategy provide efficient and intelligent search capability for the multi-parameter coupling optimization system of the spinning bath, accelerating the discovery process of the optimal parameter combination.

[0016] 3、The present application establishes a coupled mathematical equation group of flow field equation, temperature field equation, concentration field equation and phase separation equation, and uses finite element method for accurate solution to obtain microstructure characteristic parameters and defect precursor information in the fiber structure evolution process. On this basis, the mapping relationship between the stabilizability indexes is introduced to realize the quantitative evaluation of the internal potential of the uniform and complete trapezoidal structure of the raw silk in the pre-oxidation process. At the same time, the defect precursor information is converted into a defect probability value to realize the early warning of potential quality problems. Based on the two key indicators, a multi-objective function group is constructed, and the parameter combination is evaluated by using the Pareto sorting method. The stabilizability index evaluation and multi-objective optimization iteration framework establish an accurate mathematical correlation between the spinning process parameters and the raw silk quality characteristics, realize comprehensive prediction and optimization from microstructure to carbonization potential. BRIEF DESCRIPTION OF DRAWINGS

[0017] Fig. 1 It is a spinning bath multi-parameter coupling optimization system structure schematic diagram of the embodiment of the present application. Fig. 2 An iterative operation module flowchart of an embodiment of the present application; Fig. 3 An spinning bath multi-parameter coupling optimization method flowchart of an embodiment of the present application. DETAILED DESCRIPTION

[0018] The technical solutions in the embodiments of the present application will be described clearly and completely below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor fall within the scope of protection of the present application.

[0019] Please refer to Figs. 1 to 3 The present application provides a spinning bath multi-parameter coupling optimization system and method, and the technical solutions are as follows: Embodiment one: This embodiment is used for T800 grade carbon fiber precursor production in a PAN-based carbon fiber production line, and a wet spinning process using DMSO as a solvent. To solve the problems of unstable microstructure of PAN precursor, difficulty in controlling the skin-core structure gradient, and easy formation of broken filaments during the pre-oxidation process, a spinning bath multi-parameter coupling optimization system is applied to intelligently optimize and control the quality of the spinning process, specifically including: A data analysis module obtains a multi-parameter combination to divide a process window boundary, calculates a robustness index according to the process window boundary, divides a parameter space sub-region according to the process window boundary and the robustness index, calculates the influence sensitivity of each spinning bath parameter on the stabilizability index, and obtains a parameter importance measure value; A numerical solution module establishes a process mathematical equation set, and obtains microstructure characteristic parameters and defect precursor information by solving the equation set according to a multi-parameter combination; A parameter evaluation module calculates a stabilizability index according to the microstructure characteristic parameters, and converts the defect precursor information into a defect probability; An iterative operation module generates a set of multi-parameter combinations to be evaluated using a differentiated mechanism according to the parameter space sub-region and the parameter importance measure value, establishes a function group according to the stabilizability index and the defect probability, evaluates the set of multi-parameter combinations to be evaluated using the function group, and obtains an optimal multi-parameter combination through iterative operation.

[0020] Further, the process window boundary includes a feasible region process window boundary, a transition region process window boundary and a failure region process window boundary, and the step of dividing the process window boundary includes: obtaining a plurality of parameter combinations and corresponding spinning result effectiveness in historical data, using a support vector classification method to construct a discriminant function, and determining a boundary line between the feasible region and the failure region in the overall parameter space based on the discriminant function; setting a boundary range of the transition region between the feasible region and the failure region; calculating a gradient distribution of the discriminant function in the overall parameter space; correcting the boundary shape according to the gradient size and direction; and outputting the determined feasible region process window boundary, transition region process window boundary and failure region process window boundary.

[0021] In the embodiment, first, a data set containing spinning bath parameter combinations and corresponding spinning result effectiveness is extracted from a production line historical database. The extracted parameters include: coagulation bath temperature, desolventizing bath temperature, coagulant concentration, second bath concentration, main flow speed, side flow speed, immersion time, out-bath draft ratio and heat draft ratio. The extracted spinning result effectiveness labels are divided into two categories: "success" and "failure", indicating whether the parameter combination produces qualified raw yarn. The data is preprocessed, including: removing outliers (data points outside the reasonable range of parameters), handling missing values (filling with the average value of similar parameter combinations), and normalizing parameters (normalizing each parameter to the 0-1 interval). After preprocessing, 842 samples are obtained, including 625 success samples and 217 failure samples.

[0022] During the support vector classifier training process, higher weight is given to samples near the boundary. According to the discriminant function value, a threshold is set to divide three regions: feasible region: discriminant function value greater than 0.5; transition region: discriminant function value between -0.5 and 0.5; failure region: discriminant function value less than -0.5; the threshold is selected based on ROC curve analysis, ensuring that more than 90% of the success samples are correctly classified while minimizing misclassification; the gradient distribution of the discriminant function in the overall parameter space is calculated, and the regions with sharp gradient changes are identified as high sensitivity areas, and sampling points are added in these areas, the classifier is retrained, and after 3 iterations, the stable process window boundary is obtained.

[0023] The division of the process window boundary considers the correspondence between the multiple parameter combinations and the spinning results in the historical data, and through gradient distribution analysis and boundary correction, the accuracy of the process window boundary is ensured, providing a reliable basis for subsequent robustness evaluation and optimization algorithm selection, effectively improving the system's understanding of the complexity of the parameter space and reducing the risk of falling into unstable regions during optimization.

[0024] Table 1 shows the results of process window boundary division test on 5 typical spinning bath parameter combinations, which are highly consistent with the actual experimental verification results.

[0025] Table 1 Process window boundary division results Further, a robustness index is calculated according to the parameter space distance of each multi-parameter combination to the process window boundary of the feasible region, and the stability score of the multi-parameter combination; based on the process window boundary and the robustness index, the overall parameter space is divided into multiple sub-regions by combining the density clustering method and the Markov random field; the geometric characteristics and the distribution characteristics of the robustness index of each sub-region are analyzed to obtain the parameter space sub-region; the overall parameter space is the value range of each spinning bath parameter.

[0026] The specific implementation steps of the robustness index calculation are as follows: The minimum parameter space distance of the multi-parameter combination to the process window boundary of the feasible region is calculated, and the calculation method of the parameter space distance includes: for a given multi-parameter combination point, the point on the process window boundary closest to it is determined, and the distance between the two points (Euclidean distance or Mahalanobis distance can be used) is calculated; for a parameter combination point located inside the feasible region, the distance is kept positive, indicating the safety margin of the parameter point from the process failure boundary; for a parameter combination point located outside the feasible region (i.e. in the transition region or failure region), the distance is assigned a negative value, and the size is still the actual distance, but the sign indicates that the point has exceeded the boundary of the feasible region; the partial derivatives of the stabilizability index with respect to each parameter in the historical data are calculated, indicating the degree of influence of small changes in each parameter on the stabilizability index, and the square sum of these partial derivatives is multiplied by a negative scaling factor to obtain a stability score ranging from 0 to 1; the final robustness index is equal to the product of the minimum parameter space distance and the stability score.

[0027] The density clustering method and the Markov random field are combined to achieve the division of the parameter space. In specific implementation, first, uniform distribution grid points are generated in the overall parameter space, and the discriminant function value (based on the discriminant function value, the grid points are preliminarily marked as the feasible region, the transition region or the failure region, as the prior information for density clustering) and the robustness index of each point are calculated; the DBSCAN density clustering algorithm is used for preliminary clustering to identify high-density regions, transition regions and noise points, and then a Markov random field model is constructed based on the clustering results; an energy function is defined including data items and smoothing items, wherein the data items reflect the relationship between the points and the process window boundary, and the graph cut algorithm is used to solve the energy minimization problem to obtain the preliminary sub-region division; geometric feature analysis (including volume, shape complexity, convexity, boundary curvature, etc.) and robustness index distribution characteristic analysis (including statistics, distribution pattern, spatial gradient, etc.) are performed on each sub-region, and based on these characteristics, the sub-region optimization is performed, including merging similar small regions and splitting heterogeneous large regions, and finally the sub-regions are classified into three main types.

[0028] The robustness index combined with parameter space division not only quantifies the process stability of parameter combinations, but also considers the local characteristics of the parameter space, providing a scientific basis for differentiated optimization strategies. By analyzing the geometric characteristics and robustness index distribution characteristics of each sub-region, the system can identify stable regions, sensitive regions and transition regions in the parameter space, improving the accuracy of parameter space cognition and the adaptability of the system. This allows the optimization algorithm to adaptively adjust according to the characteristics of different regions, thereby improving optimization efficiency and result reliability.

[0029] Further, the numerical solving module specifically comprises: establishing a process mathematical equation set of flow field equations, temperature field equations, concentration field equations and phase separation equations; determining the calculation grid and boundary conditions, and discretely solving the process mathematical equation set along the time axis using the finite element method to obtain flow field, temperature field, concentration field and phase separation field distribution data at each time point; extracting microstructure characteristic parameters and defect precursor information based on the distribution data; wherein the microstructure characteristic parameters include average pore size, pore size distribution standard deviation, skin layer density, skin layer thickness, core layer uniformity parameter and molecular chain orientation degree; and the defect precursor information includes defect precursor density and defect precursor average size.

[0030] In this embodiment, COMSOL Multiphysics software platform is used to realize multi-physical field coupling simulation, and the specific steps include: establishing a three-dimensional geometric model including a spinneret hole, a coagulation bath, a desolventizing bath and a hot water bath; to improve calculation efficiency, only 1 / 4 part is modeled and symmetric boundary conditions are set by using system symmetry; the k-epsilon turbulence model is selected for the flow field equation, and the inlet flow rate, wall no-slip condition and outlet pressure condition are set; the heat conduction equation is used for the temperature field equation, and the inlet temperature, wall convective heat transfer condition and environment temperature are set; the component diffusion equation is used for the concentration field equation, and the initial concentration distribution and component diffusion coefficient are set; the Flory-Huggins thermodynamic model is used for the phase separation field process, and the interaction parameter and mobility are set; the physical properties of PAN solution, coagulation bath liquid and fiber materials are defined, including density, viscosity, heat capacity, thermal conductivity, diffusion coefficient, etc., which are obtained through experimental measurement or literature data; unstructured grid is used, and grid is encrypted in key areas such as fiber surface; the full-coupling solver is selected, the time step is set to 0.01s, and the total simulation time is the actual residence time of the fiber in the coagulation bath (usually 25 to 55s); the convergence criterion is set to relative error less than (1e-4) or maximum iteration number 30 times.

[0031] The specific steps of microstructure characteristic parameter extraction include: exporting the phase separation field data of the fiber cross section and longitudinal section in the COMSOL post-processing module, and using MATLAB to realize the extraction of microstructure characteristic parameters: Average pore size: the average value of equivalent diameter is calculated by binarizing cross-sectional phase separation field data (threshold value is set to 0.5), using region connectivity analysis to identify pores; Pore size distribution standard deviation: the standard deviation of the equivalent diameters of all pores is calculated; Skin density: the average PAN phase volume fraction within a 2 μm thickness range from the fiber surface is calculated; Skin thickness: the radial profile is made from the fiber surface to the center, and the distance when the PAN phase volume fraction drops to 85% of the surface value is defined as the skin thickness; Core uniformity parameter: the coefficient of variation of the PAN phase volume fraction in the central region of the fiber (radius < 50% of the fiber radius) is calculated; Molecular chain orientation degree: the orientation tensor is calculated based on the velocity gradient field data, and the maximum eigenvalue is taken as the orientation degree.

[0032] The specific steps of defect precursor information extraction include: identifying various types of defect precursors: when the local solvent concentration exceeds 25%, it is marked as a solvent-rich area; when the local pore diameter exceeds 100 nm, it is marked as a large-sized hole; when the local phase separation gradient exceeds 0.2 / μm, it is marked as a phase separation non-uniform area; when the local stress exceeds 5 MPa, it is marked as a stress concentration area; the number of each type of defect precursor per unit volume is calculated as the defect precursor density, and the arithmetic mean of the size of all defect precursors is calculated as the average size of the defect precursors.

[0033] The multi-physics field coupling simulation method of the process mathematical equation set reveals the internal relationship between the spinning bath parameters and the formation of the fiber microstructure, greatly improving the prediction accuracy. Considering the interaction and dynamic evolution process of multiple physical fields, key phenomena in the fiber formation process can be accurately captured, providing reliable microstructure data for subsequent stablezation index evaluation, thereby realizing accurate mapping from process parameters to microstructure.

[0034] Further, the parameter evaluation module specifically includes: a nonlinear mapping model based on a deep neural network, for establishing a mapping relationship between microstructure feature parameters and a stablezation index; a statistical regression model based on a statistical regression model, for converting defect precursor density and defect precursor average size into defect probability values.

[0035] In this embodiment, a nonlinear mapping model is constructed by a fully connected network and an activation function, and a negative binomial regression model is selected to convert the defect probability value. The independent variables of the negative binomial regression model are the defect precursor density and the average size of the defect precursor, and the dependent variable is the actual observed number of defects. The nonlinear mapping model is trained by historical data, which includes microstructure feature parameters, pre-oxidation process parameters and pre-oxidation results (cyclization index, DSC curve characteristics, etc.). Based on the pre-oxidation results in the historical data, the real value of the stabilizability index of each group of historical samples is calculated. The stabilizability index is defined as a comprehensive evaluation index reflecting the potential of the original wire to form a uniform and complete ladder structure under standard pre-oxidation conditions. The specific calculation method is as follows: First, calculate the pre-oxidation efficiency coefficient = (actual cyclization index / theoretical maximum cyclization index) x (1-cyclization degree variation coefficient); Calculate the structure uniformity coefficient = (1-radial cyclization gradient) x (1-DSC exothermic peak width / standard width); Calculate the defect sensitivity coefficient = 1-(actual defect number / standard sample average defect number); The final stabilizability index is the weighted average of the above three coefficients.

[0036] Through the nonlinear mapping model based on deep neural network and the statistical regression model, the conversion relationship between the microstructure feature parameters and the stabilizability index, and the conversion relationship between the defect precursor information and the defect probability are established respectively. The deep neural network is trained by a large amount of historical data, which can capture the complex nonlinear relationship between the microstructure and the pre-oxidation performance, and the statistical regression model provides a reliable evaluation of the potential defect risk. This evaluation method establishes a quantitative correlation between the fiber microstructure and the subsequent processing performance, so that the system can accurately predict the carbonization potential of the original wire at the spinning stage, and provides a scientific basis for multi-objective optimization.

[0037] Further, the parameter space sub-region includes a first space sub-region, a second sub-space region and a third space sub-region; the difference mechanism is that a first optimization algorithm is used for the first space sub-region, a second optimization algorithm is used for the second sub-space region, and a third optimization algorithm is used for the third space sub-region; the search rule of the spinning bath parameters is adjusted according to the parameter importance measurement value, and the to-be-evaluated multi-parameter combination set is generated.

[0038] The first parameter space sub-region feature is: high shape complexity, large boundary curvature, local non-convex, usually accounting for 20-30% of the overall parameter space; narrow range of robustness index (usually between 0.5-2.5), moderate mean, large standard deviation, skewed distribution, non-uniform spatial gradient; located in the feasible region but close to the transition region boundary, high parameter sensitivity, this type of region is suitable for using simulated annealing and tabu search hybrid algorithm, making full use of its global search ability to deal with complex parameter space.

[0039] The second parameter space sub-region feature is: shape close to convex, small boundary curvature, smooth boundary, usually accounting for 15-25% of the overall parameter space; wide range of robustness index (usually between 2.5-5.5), high mean, small standard deviation, approximately normal distribution, uniform and small spatial gradient; located in the core region of the feasible region, high process stability, weak parameter coupling effect. This type of region presents a smooth response surface characteristic, suitable for using an efficient local optimization algorithm combining response surface method and gradient descent.

[0040] The third parameter space sub-region feature is: extremely irregular shape, high proportion of high curvature boundary points, multiple non-convex, complex boundary, usually accounting for 10-15% of the overall parameter space; extremely narrow range of robustness index (usually between 0.0-0.8), low mean, bimodal distribution, steep spatial gradient; across the boundary of the feasible region and the transition region, high process result uncertainty, strong parameter coupling effect. This type of region represents the true boundary of the process window, suitable for using adaptive grid local encryption and boundary tracking algorithm, which needs to finely describe the boundary shape.

[0041] The calculation process of the parameter importance measure (PIM) includes: selecting 200 representative points in the parameter space, making a small perturbation (±5%) to each parameter dimension, calculating the change rate of the stabilizability index as the local sensitivity, for each parameter, calculating the square root of the sum of squares of the sensitivities of the 200 points to obtain the overall sensitivity of the parameter; multiply the overall sensitivity of the parameter by its normalized change range, and then divide by the standard deviation of the stabilizability index to obtain the parameter importance measure.

[0042] Adjusting the search rules of the spinning bath parameters according to the parameter importance measure includes: Search step adjustment: important parameters (PIM>1.0): search step=reference step×0.2; moderately important parameters (0.5≤PIM≤1.0): search step=reference step×0.5; secondary parameters (PIM<0.5): search step=reference step; wherein the reference step is set to 5% of the parameter range; Exploration frequency adjustment: important parameters: exploration frequency = reference frequency x 2.5; medium important parameters: exploration frequency = reference frequency x 1.5; secondary parameters: exploration frequency = reference frequency; wherein the reference frequency is set to 10 times per round; Convergence criterion adjustment: important parameters: convergence criterion = reference criterion x 0.25; medium important parameters: convergence criterion = reference criterion x 0.5; secondary parameters: convergence criterion = reference criterion; wherein the reference criterion is set to 1% of the parameter range.

[0043] The differentiation mechanism uses different optimization algorithms for different parameter space sub-regions, and adjusts the search rules according to the parameter importance measurement value, overcoming the limitations of traditional single algorithm in complex parameter space, improving the search efficiency, and realizing the rational allocation of computing resources. The intelligent differentiation mechanism significantly improves the exploration ability of the system in high-dimensional parameter space, speeds up the discovery process of the optimal parameter combination, and provides an efficient solution for complex process optimization.

[0044] Further, the set of multi-parameter combinations to be evaluated is evaluated using a function group, and the best multi-parameter combination is obtained by iterative operation, which specifically includes: constructing a multi-objective function group including maximization of stabilizability index and minimization of defect probability; using a Pareto sorting method to sort the multi-parameter combinations to be evaluated, forming different Pareto levels; selecting the multi-parameter combinations of the first Pareto level and the multi-parameter combinations of other levels with a comprehensive score higher than a preset threshold to be retained to the next iteration; updating the process window boundary and the robustness index, and redividing the parameter space sub-regions and recalculating the parameter importance measurement value based on the updated process window boundary and the robustness index; generating new multi-parameter combinations to be evaluated based on the updated parameter space sub-regions and the parameter importance measurement value; determining whether the iteration termination condition is met; if the termination condition is met, outputting the best multi-parameter combination, and if the termination condition is not met, continuing the iterative operation.

[0045] The flowchart of the iterative operation module is shown in Fig. 2 The Pareto sorting step is: a, find all non-dominated solutions in the current set, and mark them as the first level; b, remove these solutions from the set, and repeat step a in the remaining solutions, and mark them as the second level; c, repeat the above process until all solutions are classified; the comprehensive score is the weighted average of the stabilizability index and the defect probability.

[0046] The specific settings of the iteration termination condition include: maximum iteration times of 30 rounds; in the last 3 rounds of iteration, the relative improvement amplitude of the optimal stabilizability index is less than 0.1%, and the relative improvement amplitude of the optimal defect probability is less than 0.5%; the total calculation time exceeds 48 hours; in actual application, the system will check these three conditions at the same time, and any condition met will terminate the iteration.

[0047] A multi-objective function group including maximized stabilizability index and minimized defect probability is constructed to realize the comprehensive optimization of PAN precursor quality. The process window boundary, robustness index and parameter space division are dynamically updated after each iteration, so that the optimization process can adaptively track the changes of parameter space characteristics, and realize the continuous evolution and self-improvement of the optimization process.

[0048] Table 2 Comparison of production indicators Table 2 shows the comparison of various production indicators before and after the application of the system. After the application, the qualified rate of the precursor is increased by 6.7%, the qualified rate of the pre-oxidized yarn is increased by 8.5%, the qualified rate of the carbon fiber is increased by 8.7%, the high-strength and high-modulus rate is increased by 51.7%, the average number of broken filaments is reduced by 64.3%, and the energy consumption is reduced by 7.7%.

[0049] The system first divides the process window boundary based on historical data, calculates the robustness index, divides the parameter space sub-region, obtains the parameter importance measure value, and establishes the accurate knowledge of the parameter space; then the numerical solution module is used to establish the mathematical equation group of the process, to accurately predict the microstructure characteristics and defect precursor information under different parameter combinations; then the parameter evaluation module is used to convert the microstructure characteristics into the stabilizability index and the defect precursor information into the defect probability, to realize the quantitative correlation between the process parameters and the quality characteristics; finally, the iteration operation module uses the differentiated mechanism to generate the parameter combination, and uses the multi-objective function group for evaluation and iterative optimization, until the best multi-parameter combination is obtained. The systematic and intelligent parameter optimization method realizes the accurate control of the multi-parameters of the spinning bath, and significantly improves the structure uniformity, defect control ability and stabilizability of the PAN precursor.

[0050] Example two This embodiment further illustrates a spinning bath multi-parameter coupling optimization method, as shown in Fig. 3 , which includes: obtaining the multi-parameter combination in the historical data to divide the process window boundary, calculating the robustness index according to the process window boundary, dividing the parameter space sub-region according to the process window boundary and the robustness index, calculating the sensitivity of each spinning bath parameter to the stabilizability index, and obtaining the parameter importance measure value; Furthermore, the process window boundary includes a feasible domain process window boundary, a transition domain process window boundary, and a failure domain process window boundary. The steps of dividing the process window boundary are: obtaining a multi-parameter combination in historical data and the corresponding spinning result validity, using a support vector classification method to construct a discriminant function, and determining the boundary between the feasible domain and the failure domain in the overall parameter space based on the discriminant function; setting a boundary range of the transition domain between the feasible domain and the failure domain; calculating the gradient distribution of the discriminant function in the overall parameter space; correcting the boundary shape according to the gradient size and direction; and outputting the determined feasible domain process window boundary, transition domain process window boundary, and failure domain process window boundary. Furthermore, a robustness index is calculated based on the parameter space distance between each multi-parameter combination and the boundary of the feasible process window, as well as the stability score of the multi-parameter combination. Based on the process window boundary and the robustness index, the overall parameter space is divided into multiple sub-regions in combination with the density clustering method and the Markov random field. The geometric characteristics of each sub-region and the distribution characteristics of the robustness index are analyzed to obtain the parameter space sub-regions. The overall parameter space is the value range of each spinning bath parameter. Establish a mathematical equation group for the process, and solve the equation group based on multi-parameter combination to obtain microstructure characteristic parameters and defect precursor information; Furthermore, establishing a set of mathematical equations for the process specifically includes: establishing a set of mathematical equations for the process of flow field equations, temperature field equations, concentration field equations, and phase separation equations; determining a computational grid and boundary conditions, and using a finite element method to discretely solve the set of mathematical equations for the process along a time axis to obtain flow field, temperature field, concentration field, and phase separation field distribution data at each time point; extracting microstructure characteristic parameters and defect precursor information based on the distribution data; wherein the microstructure characteristic parameters include average pore size, pore size distribution standard deviation, skin density, skin thickness, core layer uniformity parameter, and molecular chain orientation; and the defect precursor information includes defect precursor density and average size of defect precursors; Calculating a stabilization index based on the microstructural characteristic parameters and converting the defect precursor information into a defect probability; Furthermore, a nonlinear mapping model based on a deep neural network is used to establish a mapping relationship between microstructural characteristic parameters and the stabilization index; a statistical regression model is used to convert the defect precursor density and the average size of the defect precursor into a defect probability value; generating a set of multi-parameter combinations to be evaluated using a differentiation mechanism based on the parameter space sub-regions and the parameter importance metrics, establishing a function group based on the stabilization index and the defect probability, evaluating the set of multi-parameter combinations to be evaluated using the function group, and obtaining an optimal multi-parameter combination through iterative calculation; Further, the parameter space sub-region includes a first space sub-region, a second sub-space region and a third space sub-region; the differentiation mechanism is that a first optimization algorithm is used for the first space sub-region, a second optimization algorithm is used for the second sub-space region, and a third optimization algorithm is used for the third space sub-region; a search rule of the spinning bath parameters is adjusted according to the parameter importance measure value, and the to-be-evaluated multi-parameter combination set is generated.

[0051] Further, the to-be-evaluated multi-parameter combination set is evaluated using a function group, and the optimal multi-parameter combination obtained through iterative operation includes: constructing a multi-objective function group including maximization of a stabilizability index and minimization of a defect probability; using a Pareto sorting method to sort the to-be-evaluated multi-parameter combination to form different Pareto levels; selecting the multi-parameter combination of the first Pareto level and the multi-parameter combination of other levels with a comprehensive score higher than a preset threshold to be reserved to the next round of iteration; updating a process window boundary and a robustness index, re-dividing the parameter space sub-region and recalculating the parameter importance measure value according to the updated process window boundary and the robustness index; generating a new to-be-evaluated multi-parameter combination based on the updated parameter space sub-region and the parameter importance measure value; judging whether an iteration termination condition is met; if the termination condition is met, outputting the optimal multi-parameter combination, and if the termination condition is not met, continuing the iterative operation.

[0052] Although the embodiments of the present application have been shown and described, it can be understood by those skilled in the art that various changes, modifications, replacements and variations can be made to the embodiments without departing from the principles and spirits of the present application, and the scope of the present application is defined by the appended claims and their equivalents.

Claims

1. A spinning bath multi-parameter coupling optimization system, characterized in that: include: a data analysis module, which obtains multi-parameter combinations from historical data to divide process window boundaries, calculates a robustness index based on the process window boundaries, divides parameter space into sub-regions based on the process window boundaries and the robustness index, calculates the sensitivity of each spinning bath parameter to the stabilization index, and obtains a parameter importance metric; Numerical solution module, which establishes a set of mathematical equations for the process and solves the equations based on multi-parameter combinations to obtain microstructural characteristic parameters and defect precursor information; a parameter evaluation module, calculating a stabilization index based on the microstructural characteristic parameters and converting the defect precursor information into a defect probability; An iterative operation module generates a set of multi-parameter combinations to be evaluated using a differentiation mechanism based on the parameter space sub-regions and the parameter importance metrics, establishes a function group based on the stabilization index and the defect probability, evaluates the set of multi-parameter combinations to be evaluated using the function group, and obtains the optimal multi-parameter combination through iterative operation.

2. A spinning bath multi-parameter coupling optimization system according to claim 1, characterized in that: The process window boundary includes a feasible domain process window boundary, a transition domain process window boundary and a failure domain process window boundary. The steps of dividing the process window boundary are: obtaining a multi-parameter combination in historical data and the corresponding spinning result validity, using a support vector classification method to construct a discriminant function, and determining the dividing line between the feasible domain and the failure domain in the overall parameter space based on the discriminant function; setting the boundary range of the transition domain between the feasible domain and the failure domain; calculating the gradient distribution of the discriminant function in the overall parameter space; correcting the boundary shape according to the gradient size and direction; and outputting the determined feasible domain process window boundary, transition domain process window boundary and failure domain process window boundary.

3. The spinning bath multi-parameter coupling optimization system according to claim 2, characterized in that: The robustness index is calculated based on the parameter space distance between each multi-parameter combination and the boundary of the feasible domain process window, as well as the stability score of the multi-parameter combination; based on the process window boundary and the robustness index, the overall parameter space is divided into multiple sub-regions in combination with the density clustering method and Markov random field, and the geometric characteristics of each sub-region and the distribution characteristics of the robustness index are analyzed to obtain the parameter space sub-regions; the overall parameter space is the value range of each spinning bath parameter.

4. The multi-parameter coupling optimization system for a spinning bath according to claim 1, characterized in that: The numerical solution module specifically includes: establishing a process mathematical equation group of flow field equations, temperature field equations, concentration field equations and phase separation equations; determining the calculation grid and boundary conditions, and using the finite element method to discretely solve the process mathematical equation group along the time axis to obtain the flow field, temperature field, concentration field and phase separation field distribution data at each time point; extracting microstructure characteristic parameters and defect precursor information based on the distribution data; wherein the microstructure characteristic parameters include average pore size, pore size distribution standard deviation, skin density, skin thickness, core layer uniformity parameter and molecular chain orientation; the defect precursor information includes defect precursor density and defect precursor average size.

5. The spinning bath multi-parameter coupling optimization system according to claim 1, characterized in that: The parameter evaluation module specifically includes: a nonlinear mapping model based on a deep neural network, which is used to establish a mapping relationship between microstructure characteristic parameters and stabilization index; and a statistical regression model, which is used to convert the defect precursor density and the average size of the defect precursor into a defect probability value.

6. The spinning bath multi-parameter coupling optimization system according to claim 1, characterized in that: The parameter space sub-regions include a first space sub-region, a second sub-space region and a third space sub-region; the differentiation mechanism is: using a first optimization algorithm for the first space sub-region, a second optimization algorithm for the second sub-space region, and a third optimization algorithm for the third space sub-region, adjusting the search rules of the spinning bath parameters according to the parameter importance metric value, and generating the multi-parameter combination set to be evaluated.

7. The spinning bath multi-parameter coupling optimization system according to claim 1, characterized in that: The set of multi-parameter combinations to be evaluated is evaluated using a function group, and the iterative operation to obtain the optimal multi-parameter combination specifically includes: constructing a multi-objective function group including maximization of the stabilization index and minimization of the defect probability; using the Pareto sorting method to sort the multi-parameter combinations to be evaluated to form different Pareto levels; selecting the multi-parameter combination of the first Pareto level and the multi-parameter combinations of other levels with a comprehensive score higher than a preset threshold to be retained for the next round of iteration; updating the process window boundary and the robustness index, re-dividing the parameter space sub-region according to the updated process window boundary and the robustness index and recalculating the parameter importance measurement value; generating a new multi-parameter combination to be evaluated based on the updated parameter space sub-region and the parameter importance measurement value; judging whether the iteration termination condition is met; if the termination condition is met, outputting the optimal multi-parameter combination, if not, continuing the iterative operation.

8. A spinning bath multi-parameter coupling optimization method, executing a spinning bath multi-parameter coupling optimization system according to claim 1, characterized in that: include: Obtaining a multi-parameter combination in historical data to divide a process window boundary, calculating a robustness index based on the process window boundary, dividing the parameter space into sub-regions based on the process window boundary and the robustness index, calculating the sensitivity of each spinning bath parameter to the stabilization index, and obtaining a parameter importance metric; Establish a mathematical equation group for the process, and solve the equation group based on multi-parameter combination to obtain microstructure characteristic parameters and defect precursor information; Calculating a stabilization index based on the microstructural characteristic parameters and converting the defect precursor information into a defect probability; According to the parameter space sub-region and the parameter importance metric value, a differentiation mechanism is used to generate a set of multi-parameter combinations to be evaluated, a function group is established according to the stabilization index and the defect probability, the function group is used to evaluate the set of multi-parameter combinations to be evaluated, and an optimal multi-parameter combination is obtained through iterative calculation.

Images