A development method of a preparation process of high-strength high-modulus polyacrylonitrile-based carbon fiber

Abstract

This invention provides a method for developing a preparation process for high-strength, high-modulus polyacrylonitrile-based carbon fibers, comprising the following steps: Based on a process dataset, a prediction model is constructed, including a total model and multiple sub-models. The total model is a machine learning algorithm model that takes all process parameters during the preparation of polyacrylonitrile-based carbon fibers as input and outputs performance data and / or structural data of the polyacrylonitrile-based carbon fibers. Sub-models are machine learning algorithm models that take local process parameters during the preparation of polyacrylonitrile-based carbon fibers as input and output fiber performance data and / or structural data after local process treatment. A set of all process parameters is randomly generated, and the prediction model is used for prediction. Specifically, the sub-models are run first, and the total model is run only when the outputs of all sub-models meet the corresponding output criteria. This invention aims to reduce the time, energy consumption, and financial investment in the development process of preparing high-strength, high-modulus polyacrylonitrile-based carbon fibers.

Description

Technical Field

[0001] This invention relates to the field of carbon fiber technology, and in particular to a method for developing a preparation process for high-strength, high-modulus polyacrylonitrile-based carbon fibers. Background Technology

[0002] Polyacrylonitrile (PAN)-based carbon fiber is a high-tech integrated, high-value-added product with properties such as lightweight, high strength, high temperature resistance, and corrosion resistance. It has been widely used in aerospace, aviation, and industrial products and other technical fields.

[0003] High-performance polyacrylonitrile-based carbon fibers have numerous and demanding performance indicators. Their preparation process is exceptionally long (with many coupling and conflicting aspects in the design of upstream and downstream processes), involves numerous control points (≥3000 control points), interweaves multiple technical disciplines, and spans a wide temperature range (from room temperature to 3000℃). Therefore, the research and development of high-performance carbon fibers is a complex systems engineering problem.

[0004] A key direction in the current research and development of next-generation carbon fibers is the development of high-strength, high-model carbon fibers. Continuing with traditional experimental research would inevitably involve numerous trial-and-error choices, requiring significant time, energy, and financial investment. In recent years, machine learning has made breakthroughs in areas such as modeling, performance prediction, and optimization in materials research, demonstrating significant advantages over traditional research methods. Figure 1 A new paradigm for the research and development of carbon fiber materials (see...) Figure 1 Figure a in the text (characterized by data-driven artificial intelligence) and traditional research paradigms (see Figure a) Figure 1 Comparison of Figure b in the text.

[0005] Machine learning technology is currently permeating the field of carbon fiber production technology, forming a cross-disciplinary and integrated development trend. Existing researchers have applied specific machine learning algorithms to achieve research results with practical value and guiding significance for carbon fiber research and production in areas such as carbon fiber production process control, local process optimization, and energy consumption optimization.

[0006] In summary, there is an urgent need to introduce machine learning algorithms into the research process of the entire process and local processes of high-performance polyacrylonitrile-based carbon fiber preparation, and to innovate a method for developing a preparation process of high-strength and high-modulus polyacrylonitrile-based carbon fiber, so as to significantly reduce the research and development debugging time, energy consumption and capital investment, and thus successfully prepare high-strength and high-modulus polyacrylonitrile-based carbon fiber. Summary of the Invention

[0007] In view of this, the present invention provides a method for developing a preparation process for high-strength, high-modulus polyacrylonitrile-based carbon fibers, the main purpose of which is to reduce the time, energy consumption and capital investment in the development process of the preparation process for high-strength, high-modulus polyacrylonitrile-based carbon fibers.

[0008] To achieve the above objectives, the present invention mainly provides the following technical solutions:

[0009] On one hand, embodiments of the present invention provide a method for developing a preparation process for high-strength, high-modulus polyacrylonitrile-based carbon fibers, wherein the development method includes the following steps:

[0010] The steps for constructing the prediction model are as follows: Based on the process dataset of polyacrylonitrile-based carbon fiber, a prediction model including a total model and multiple sub-models is constructed. The total model is a machine learning algorithm model that takes the entire process parameters of polyacrylonitrile-based carbon fiber preparation as input and outputs the performance data and / or structural data of polyacrylonitrile-based carbon fiber. The sub-models are machine learning algorithm models that take the local process parameters of polyacrylonitrile-based carbon fiber preparation as input and output the fiber performance data and / or structural data after local process treatment.

[0011] The steps for setting the criteria are as follows: Based on the target performance of the required polyacrylonitrile-based carbon fiber, set the parameter range for each process point in the overall process parameters, set the output criteria for each sub-model, and set the output criteria for the overall model.

[0012] Process search steps: The parameter range of each process point in the entire process parameters is gridded and iteratively searched for the entire process parameters. The prediction model is used for prediction in each iteration.

[0013] The steps of using the prediction model for prediction include: first running the sub-models, and then running the overall model when the outputs of all the sub-models meet the corresponding output criteria conditions.

[0014] When the data of at least one sub-model does not meet the corresponding output criterion, a set of full-line process parameters is generated through a new loop iteration, and the prediction model is used again for prediction.

[0015] Among them, after running the overall model, if the output of the overall model does not meet the output criterion conditions, a set of process parameters for the entire line will be generated through a new loop iteration, and the prediction model will be used again for prediction.

[0016] In this process, after running the overall model, when the output of the overall model meets the output criterion conditions, the corresponding full-line process parameters are recorded as the selected full-line process parameters for use in preparing polyacrylonitrile-based carbon fibers with the required properties.

[0017] The process search step, as described above, further includes:

[0018] Experimental verification steps: Conduct experimental verification of the selected process parameters for the entire production line;

[0019] If the performance of the prepared polyacrylonitrile-based carbon fiber reaches the target value, then the process development is complete.

[0020] If the performance of the prepared polyacrylonitrile carbon fiber does not meet the target value, the experimental process data will be added to the process dataset of polyacrylonitrile-based carbon fiber, the overall model and sub-models will be retrained, and the process search and experimental verification steps will be repeated.

[0021] Preferably, the process dataset for the polyacrylonitrile-based carbon fiber includes: full-line process parameter data, fiber structure data, and fiber performance data during the preparation of the polyacrylonitrile-based carbon fiber; preferably, the full-line process parameter data includes: polymerization reaction process parameters, spinning process parameters, pre-oxidation process parameters, low-temperature carbonization process parameters, and high-temperature carbonization process parameters; more preferably, the polymerization reaction process parameters include: polymerization reaction formulation, temperature, and time; more preferably, the spinning process parameters include: solidification molding process parameters, water washing process parameters, hot water drawing process parameters, oiling concentration, drying and densification process parameters, and steam... The process parameters include drawing process parameters and shrinkage heat setting process; more preferably, the pre-oxidation process parameters include pre-oxidation temperature, pre-oxidation time, and draw ratio; more preferably, the low-temperature carbonization process parameters include low-temperature carbonization temperature and low-temperature carbonization time; more preferably, the high-temperature carbonization process parameters include high-temperature carbonization temperature and high-temperature carbonization time; preferably, the fiber performance data includes performance data of polyacrylonitrile precursor filament and performance data of polyacrylonitrile-based carbon fiber; preferably, the fiber structural data includes the sheath thickness of the fiber core-sheath structure and the roundness of the fiber cross-section; more preferably, the fiber structural data also includes the fiber orientation degree.

[0022] Preferably, in the step of constructing the prediction model: the machine learning algorithm model is a BP neural network model.

[0023] Preferably, in the step of constructing the prediction model: based on the process dataset of polyacrylonitrile-based carbon fiber, a machine learning algorithm model is constructed with the process parameters of the entire process of polyacrylonitrile-based carbon fiber preparation as input and the performance data and / or structural data of polyacrylonitrile-based carbon fiber as output, and trained to obtain the overall model.

[0024] Preferably, the plurality of sub-models includes: a first sub-model, a second sub-model, a third sub-model, and a fourth sub-model; wherein,

[0025] Based on the process dataset of polyacrylonitrile-based carbon fiber, a machine learning algorithm model was constructed with polymerization reaction process parameters and solidification molding process parameters as inputs and the cross-sectional roundness of the nascent fiber as the output. The model was then trained to obtain the first sub-model.

[0026] Based on the process dataset of polyacrylonitrile-based carbon fiber, a machine learning algorithm model is constructed with polymerization reaction process parameters and spinning process parameters as inputs and performance data of polyacrylonitrile precursor fibers as outputs, and trained to obtain a second sub-model; preferably, the performance data of polyacrylonitrile precursor fibers includes tensile strength data and tensile modulus data of polyacrylonitrile precursor fibers.

[0027] Based on the process dataset of polyacrylonitrile-based carbon fiber, a machine learning algorithm model is constructed with pre-oxidation process parameters as input and pre-oxidized fiber performance data as output, and trained to obtain a third sub-model; preferably, the performance data of the pre-oxidized fiber includes the tensile strength data and tensile modulus data of the pre-oxidized fiber.

[0028] Based on the process dataset of polyacrylonitrile-based carbon fiber, a machine learning algorithm model was constructed with spinning process parameters as input and the sheath thickness data of the core-sheath structure of polyacrylonitrile precursor fiber as output. The model was then trained to obtain the fourth sub-model.

[0029] Preferably, in the step of setting the criterion conditions:

[0030] The criterion for the first sub-model is set as follows: the roundness of the cross-section of the nascent fiber is within a set range, preferably 0.75-0.95; wherein, the roundness of the cross-section of the nascent fiber is the ratio of the minor axis to the major axis of the cross-section of the nascent fiber; and / or

[0031] The criteria for the second sub-model are set as follows: the strength of the polyacrylonitrile precursor fiber is within a set strength range, preferably 450-900 MPa; the modulus of the polyacrylonitrile precursor fiber is within a set modulus range, preferably 7-15 GPa; and / or

[0032] The criteria for the third sub-model are set as follows: the tensile strength of the pre-oxidized fiber is within a set tensile strength range, preferably 350-940 MPa; the tensile modulus of the pre-oxidized fiber is within a set tensile modulus range, preferably 7-14 GPa; and / or

[0033] The criterion condition for the fourth sub-model is set as follows: the sheath thickness of the polyacrylonitrile precursor fiber is a set thickness, preferably 40-600 nm.

[0034] Preferably, in the step of setting the criterion conditions: the criterion conditions for the total model are: the tensile strength of the polyacrylonitrile-based carbon fiber is within a set tensile strength range, preferably 4.6-6.0 GPa; the tensile modulus of the polyacrylonitrile-based carbon fiber is within a set tensile modulus range, preferably 340-480 GPa.

[0035] Preferably, in the process search step, each set of full-line process parameters generated by iterative iteration is generated based on roulette wheel betting and a basic genetic algorithm.

[0036] Preferably, in the process search step: when the output criterion condition is not met, a set of full-line process parameters is generated again through iterative iteration based on the genetic algorithm.

[0037] Preferably, in the process search step: when the number of selected process parameters for the entire line is less than the set number of groups, the process search step is repeated iteratively until the number of selected process parameters for the entire line reaches the set number of groups; preferably, the set number of groups is 20 groups.

[0038] Preferably, in the experimental verification step:

[0039] Experiments are conducted to verify one or more sets of selected process parameters. If the performance of polyacrylonitrile-based carbon fibers prepared using one or more of the selected process parameters meets the target values, then the process development is complete; and / or

[0040] If the performance of the prepared polyacrylonitrile-based carbon fiber fails to meet the target value, the experimental verification data should be added to the process dataset of polyacrylonitrile-based carbon fiber, and the overall model and sub-models should be retrained and corrected. Then, the process search and experimental verification steps should be repeated.

[0041] In another aspect, embodiments of the present invention provide a high-strength, high-modulus polyacrylonitrile-based carbon fiber, wherein the high-strength, high-modulus polyacrylonitrile-based carbon fiber is prepared by a process developed using the development method of the preparation process of high-strength, high-modulus polyacrylonitrile-based carbon fiber described in any one of the above claims.

[0042] Preferably, the high-strength, high-modulus polyacrylonitrile-based carbon fiber has a tensile strength of 4.6-6.0 GPa, a tensile modulus of 340-480 GPa, a diameter of 4.2-5.7 μm, and a bulk density of 1.71-1.88 g / cm³. 3 ;

[0043] Preferably, the high-strength, high-modulus polyacrylonitrile-based carbon fiber is characterized by small-angle X-ray scattering, when the scattering vector... When, the scattering intensity I(q)∝q -α α takes values ​​of 2.85-3.3; when the scattering vector When, the scattering intensity I(q)∝q -α The value of α ranges from 0.75 to 1.1. Here, the combination of these two ranges is a significant characteristic that distinguishes the high-strength, high-modulus carbon fiber of this invention from high-strength and high-modulus fibers. The value of α reflects the surface fractal and mass fractal structural characteristics of the fiber's microscopic scattering body.

[0044] Compared with the prior art, the development method of the preparation process of high-strength and high-modulus polyacrylonitrile-based carbon fiber of the present invention has at least the following beneficial effects:

[0045] This invention provides a method for developing a preparation process for high-strength, high-modulus polyacrylonitrile-based carbon fibers, mainly including the following steps: A prediction model construction step: Based on the process dataset of polyacrylonitrile-based carbon fibers, a prediction model including a total model and multiple sub-models is constructed; wherein, the total model is a machine learning algorithm model that takes the entire process parameters in the preparation process of polyacrylonitrile-based carbon fibers as input and outputs the performance data and / or structural data of polyacrylonitrile-based carbon fibers; the sub-models are machine learning algorithm models that take the local process parameters in the preparation process of polyacrylonitrile-based carbon fibers as input and output the fiber performance data and / or structural data after local process treatment; A criterion setting step: According to the target performance of the required polyacrylonitrile-based carbon fibers, the parameter range of each process point in the entire process parameters is set, the output criterion conditions of each sub-model are set, and the output criterion conditions of the total model are set; A process search step: The entire process parameters are searched. The parameter range of each process point within the process parameter range is gridded and coded. The entire process parameter range is iteratively searched, with the prediction model used for prediction in each iteration. The steps for using the prediction model include: first, running sub-models; when the outputs of all sub-models meet the corresponding output criteria, then running the overall model; if at least one sub-model's data does not meet the corresponding output criteria, a new set of entire process parameters is generated through iterative iteration, and the prediction model is used again for prediction; after running the overall model, if the overall model's output does not meet the output criteria, a new set of entire process parameters is generated through iterative iteration, and the prediction model is used again for prediction; after running the overall model, if the overall model's output meets the output criteria, the corresponding entire process parameters are recorded as the selected entire process parameters for preparing polyacrylonitrile-based carbon fibers with the desired properties. Here, the present invention, through the above-described scheme, utilizes a process dataset of polyacrylonitrile-based carbon fibers (collected and organized from historical process-performance-structural data) and employs specific machine learning algorithms to construct a predictive model between the entire process and carbon fiber performance. This model incorporates core process points and process fiber performance or structural parameter model criteria. Using this predictive model, one or more sets of process parameters for preparing the desired performance (high strength and high modulus) are successfully selected. It is evident that, compared to traditional methods, the scheme of the present invention can significantly reduce the time, energy consumption, and capital investment in the development process of high-strength, high-modulus polyacrylonitrile-based carbon fiber preparation.

[0046] Furthermore, this embodiment of the invention provides a method for developing a preparation process for high-strength, high-modulus polyacrylonitrile-based carbon fibers. After the process search step, it further includes an experimental verification step: the selected process parameters are experimentally verified. If the performance of the prepared polyacrylonitrile-based carbon fibers reaches the target value, the process development is completed. If the performance of the prepared polyacrylonitrile carbon fibers does not reach the target value, the overall model and sub-models are retrained, and the process search step and experimental verification step are repeated. This invention combines machine learning models for fiber performance prediction and process search models with experimental verification. This effectively reduces the probability of trial and error in the development of long and complex processes, thereby reducing the development cost of high-strength and high-modulus fibers. It effectively leverages the advantages of both models and experimental verification, compensating for their respective shortcomings. The prediction and process search models can continuously output system process parameters that meet the requirements, providing exploration directions or experimental content for experiments, greatly avoiding the risk of blind trial and error. Experimental verification continuously corrects the prediction model through experimental data feedback, timely avoiding the shortcomings of model overfitting or underfitting, improving prediction accuracy, and enhancing the quality of process groups output by iterative process search. It provides a research paradigm for the development of complex carbon fiber process systems. By using machine learning models to explore the non-explicit relationships between process parameters and performance or structural parameters in past research datasets, the model exhaustively explores various combinations of process parameters within the effective range, providing preliminary screening and prediction for experimental verification. The model is continuously corrected using real experimental data. Compared with the traditional development method of theoretical modeling + experimental verification, this approach is more realistic, more operational, more effective, and more in line with the characteristics of the current data-driven and intelligent era.

[0047] Furthermore, this embodiment of the invention provides a method for developing a preparation process for high-strength, high-modulus polyacrylonitrile-based carbon fibers. This method utilizes multiple sub-models, including a first sub-model, a second sub-model, a third sub-model, and a fourth sub-model. Specifically, based on the polyacrylonitrile-based carbon fiber process dataset, a machine learning algorithm model is constructed with polymerization reaction and spinning solidification process parameters as input and the cross-sectional roundness of the nascent fiber as output, and then trained to obtain the first sub-model. Based on the polyacrylonitrile-based carbon fiber process dataset, a machine learning algorithm model is constructed with polymerization reaction and spinning process parameters as input and the performance data of the polyacrylonitrile precursor fiber as output, and then trained to obtain the second sub-model. Based on the polyacrylonitrile-based carbon fiber process dataset, a machine learning algorithm model is constructed with pre-oxidation process parameters as input and the performance data of the pre-oxidized fiber as output, and then trained to obtain the third sub-model. Based on the polyacrylonitrile-based carbon fiber process dataset, a machine learning algorithm model is constructed with spinning process parameters as input and the sheath thickness data of the core-sheath structure of the polyacrylonitrile precursor fiber as output, and then trained to obtain the fourth sub-model. Here, the present invention constructs a prediction model by setting the sub-models as the four sub-models mentioned above and nesting the sub-models with the overall model for the following reasons: 1. The sub-models reflect the implicit relationships between process parameters and process fiber properties or structural parameters in local processes; the overall model reflects the global picture. The effective amount of data for local training is much larger than the amount of global data, and the prediction accuracy of the sub-models is much higher than that of the overall model. Using sub-models or multiple sub-models effectively compensates for the low prediction accuracy of the overall model and improves the accuracy of process selection; 2. The global process flow is too long and there are too many process parameters. If only one overall model is used, the search volume of the iterative process search process is too huge and consumes too much computing resources. By using sub-models with higher prediction accuracy, a large number of process parameter groups can be filtered out in advance, improving the efficiency of process search; 3. The entire process flow is long and requires multiple core process controls along the process flow. Multiple sub-models are set up to pre-screen process parameter sets around multiple process control points. While ensuring the screening of core process control points, the entire process parameters are searched iteratively, thereby improving iterative efficiency while maintaining high accuracy. 4. Specifically, four sub-models are selected along the core control points of the process flow. The output parameters of these four sub-models are highly correlated with the performance of high-strength and high-modulus carbon fiber based on past data analysis. 5. Specifically, if only one, two, or three sub-models are selected, the process parameter sets output during the iterative process search are prone to deviating from current principles, resulting in perceived lower quality and increasing the burden of later experimental verification. 6. Theoretically, more sub-models could be selected, but too many sub-models would reduce the effective space for process screening, hindering the exploration of new process directions. Therefore, considering all factors, four sub-models are selected along the process flow.

[0048] The above description is merely an overview of the technical solution of the present invention. In order to better understand the technical means of the present invention and to implement it in accordance with the contents of the specification, the preferred embodiments of the present invention are described in detail below with reference to the accompanying drawings. Attached Figure Description

[0049] Figure 1 This represents a new paradigm for the research and development of carbon fiber materials (see...). Figure 1 Figure a in the text (characterized by data-driven artificial intelligence) and traditional research paradigms (see Figure a) Figure 1 Comparison with Figure b in the text;

[0050] Figure 2 This is a flowchart illustrating the development method of a preparation process for high-strength, high-modulus polyacrylonitrile-based carbon fibers provided by an embodiment of the present invention. Detailed Implementation

[0051] To further illustrate the technical means and effects adopted by the present invention to achieve the intended purpose, the specific embodiments, structures, features, and effects according to the present invention will be described in detail below with reference to the accompanying drawings and preferred embodiments. In the following description, different "an embodiment" or "an embodiment" do not necessarily refer to the same embodiment. Furthermore, specific features, structures, or characteristics in one or more embodiments can be combined in any suitable form.

[0052] This invention introduces machine learning into the research of the entire process / partial process of high-performance polyacrylonitrile-based carbon fiber preparation. It constructs a predictive model between the entire process and carbon fiber performance, which is nested with core process points and process fiber performance or structural parameter model criteria. With the help of the predictive model, it successfully guides process experiments (significantly reducing R&D debugging time, energy consumption, and capital investment), and develops a preparation process of high-strength and high-modulus polyacrylonitrile-based carbon fiber.

[0053] It should be noted that the preparation process of polyacrylonitrile-based carbon fiber mainly includes polymerization reaction process (i.e., preparation of spinning solution), spinning process (coagulation molding, water washing, hot water drawing, oiling, drying densification, steam drawing and shrinkage heat setting process), and carbonization process (pre-oxidation, low temperature carbonization, high temperature carbonization).

[0054] This invention provides a method for developing a preparation process for high-strength, high-modulus polyacrylonitrile-based carbon fibers, such as... Figure 2 As shown, it includes the following steps:

[0055] The steps for constructing the process dataset for the polyacrylonitrile-based carbon fiber are as follows:

[0056] The constructed polyacrylonitrile-based carbon fiber process dataset is a standardized polyacrylonitrile-based carbon fiber preparation system dataset, which includes: polymerization-spinning-carbonization process data, fiber performance data corresponding to the process, and corresponding fiber structure data.

[0057] The polymerization-spinning-carbonization process data includes: the polymerization reaction formulation, polymerization temperature and time; coagulation and molding process parameters (temperature, concentration, time, draw ratio), washing process parameters (temperature, time), hot water drawing process parameters (temperature, draw ratio), oiling concentration, drying and densification process parameters (temperature, time), steam drawing process parameters (temperature, draw ratio), shrinkage heat setting process; pre-oxidation process parameters (temperature, time, draw ratio), low-temperature carbonization process parameters (temperature, time, preferably including draw ratio), and high-temperature carbonization process parameters (temperature, time, preferably also including draw ratio, etc.). It should be noted that: after polymerization, the reaction raw materials yield a spinning solution; the spinning solution undergoes coagulation and molding treatment to obtain nascent fibers; the nascent fibers undergo washing, hot water drawing, oiling, drying and densification, steam drawing, and shrinkage heat setting treatment to obtain polyacrylonitrile precursor fibers; the polyacrylonitrile precursor fibers undergo pre-oxidation, low-temperature carbonization, and high-temperature carbonization treatments to obtain polyacrylonitrile-based carbon fibers.

[0058] The fiber performance data corresponding to the process include tensile strength, tensile modulus, and preferably also elongation.

[0059] The corresponding fiber structure data include: core-sheath structure skin thickness, fiber roundness, and preferably also fiber orientation.

[0060] Specifically, data from laboratory to pilot-scale R&D, engineering R&D, and industrial production of different grades of carbon fiber (T300, T700, T800, T1000, T1100, M40J, M46J, M55J, M60J, M65J, etc.) will be standardized and normalized. Data from different R&D stages will be compiled, screened / supplemented, and standardized and normalized. Data from past and recent systematic research findings worldwide (articles / patents containing systematic processes, fiber properties, or fiber structure information) will be collected, screened, and normalized.

[0061] Among them, data unification and standardization: process parameter data are organized according to a unified process flow sequence. The process flow settings may be different at different R&D stages. When the actual data is organized according to the standard process, the blank spaces are filled with unified data. For example, the standard hot water stretching is set to four levels, while the actual pilot test data may be set to three levels. When standardizing, the fourth level temperature is set to 20℃ and the stretching ratio is set to 1.

[0062] Among them, data normalization is performed by converting the specific data corresponding to each parameter into numbers between [0,1]. The purpose is to eliminate the difference in magnitude between the data of each dimension and avoid large network prediction errors caused by large differences in magnitude. Commonly used methods include the maximum-minimum method and the mean-variance method. The mapminmax function built into MATLAB can be used for processing.

[0063] The steps for constructing the prediction model are as follows: Based on the process dataset of polyacrylonitrile-based carbon fiber, a prediction model including a total model and multiple sub-models is constructed. The total model is a machine learning algorithm model that takes the entire process parameters of polyacrylonitrile-based carbon fiber preparation as input and outputs the performance data and / or structural data of polyacrylonitrile-based carbon fiber. The sub-models are machine learning algorithm models that take the local process parameters of polyacrylonitrile-based carbon fiber preparation as input and output the fiber performance data and / or structural data after local process treatment. This step is detailed below:

[0064] Overall Model: A backpropagation (BP) neural network model is established to represent the entire process parameters and carbon fiber properties. Specifically, the entire process parameter data during the preparation of polyacrylonitrile-based carbon fiber is used as input, and the performance data (strength and modulus) of polyacrylonitrile-based carbon fiber is used as output. The initial number of hidden layers and the initial number of nodes in the hidden layer are set.

[0065] Before training the BP neural network model, the standardized and normalized dataset can be randomly divided into a training set (45-90%), a validation set (5-20%), and a test set (5-35%), with the sum of the proportions of the training set, validation set, and test set always being 1. The training set is used to train the neural network model, and the validation set is used to prevent overfitting of the neural network model. The training set and validation set are used during the training of the BP neural network model. The test set is used to test the accuracy of the neural network model and is used after the model training is completed.

[0066] Training the overall model: Select the transfer function and training method, and train the overall model; the overall model's comprehensive accuracy (training, validation, and testing) feature should not be less than 0.84; accuracy is the linear regression correlation between the predicted value obtained by using the dataset as input and the actual value. Training methods mainly include steepest descent, additional momentum, adaptive learning rate, momentum-adaptive learning rate adjustment, and quantized conjugate gradient algorithm. The following example uses the steepest descent algorithm, but other algorithms can also be selected.

[0067] First sub-model: Establishing a BP neural network model for the roundness of the cross-section of the nascent coagulated fiber by polymerization and spinning coagulation process parameters. Specifically, using polymerization and spinning coagulation process data as input and the roundness of the nascent coagulated fiber as output, the initial number of hidden layers and the initial number of hidden layer nodes of the neural network are set to establish a BP neural network model that predicts the roundness of the nascent coagulated fiber from the polymerization reaction process and the coagulation process.

[0068] Before model training, the standardized and normalized dataset can be randomly divided into a training set (45-85%), a validation set (5-20%), and a test set (10-35%). The sum of the proportions of the training, validation, and test sets is always 1. The training set is used to train the neural network model, the validation set is used to prevent overfitting, and both are used during the training of the BP neural network model. The test set is used to test the accuracy of the neural network model and is used after model training is complete.

[0069] Training the first sub-model: Select the transfer function and training method and train the first sub-model; the overall accuracy (training, validation and testing) feature of the first sub-model is not less than 0.92; accuracy is the linear regression correlation between the predicted value and the actual value obtained by taking the dataset as input and the trained model.

[0070] Training the second sub-model: Construct a BP neural network model (second sub-model) based on the dataset to determine the polymerization reaction process parameters and spinning process parameters → raw filament properties (strength and modulus), and train it (the construction process is the same as the first sub-model);

[0071] Training the third sub-model: Construct a BP neural network model (third sub-model) based on the dataset to determine the performance (strength and modulus) of pre-oxidized fiber (pre-oxidized filament) by pre-oxidation process parameters, and train it (construction process is the same as the first sub-model);

[0072] Training the fourth sub-model: Construct a BP neural network model (fourth sub-model) based on the dataset to determine the spinning process parameters → core-sheath structure thickness of the raw filament, and train it (the construction process is the same as the first sub-model).

[0073] To achieve the performance target of high-strength and high-modulus carbon fibers, a full-line performance prediction model is constructed, using the outputs of the first, second, third, and fourth sub-models as criteria. The overall model is only run after the outputs of all sub-models meet the conditions.

[0074] Steps for setting criteria conditions:

[0075] Based on the target performance of the required polyacrylonitrile-based carbon fiber, the parameter range of each process point in the overall process parameters is set, the output criterion conditions of each sub-model are set, and the output criterion conditions of the overall model are set.

[0076] Process search steps

[0077] Based on the entire range of process parameters, a roulette wheel algorithm combined with a genetic algorithm is used to iteratively search for all process parameters. In each iteration, the aforementioned prediction model is used for prediction. The first sub-model predicts the roundness of the nascent fibers, the second sub-model predicts the properties of the polyacrylonitrile precursor fibers, the third sub-model predicts the properties of the pre-oxidized fibers, and the fourth sub-model predicts the thickness of the sheath layer in the core-sheath structure of the polyacrylonitrile precursor fibers. If the prediction results of all four sub-models meet the criteria, the overall model is run to predict the carbon fiber properties; otherwise, the iteration restarts to generate a new set of process parameters for the entire line. If the overall model prediction results meet the set indicators, the set of process parameters is recorded; otherwise, the iteration restarts to generate a new set of process parameters. When the number of qualified process parameter sets is less than 20, the process search iterates again until 20 sets are reached, at which point the process search stops.

[0078] Experimental verification steps: Conduct process experiments and measure fiber properties. If the fiber properties meet the target, the development is terminated. If the target is not met, the measured data is added to the polyacrylonitrile-based carbon fiber process dataset, and the overall model, first sub-model, second sub-model, third sub-model, and fourth sub-model are retrained and corrected. After the model correction and training are completed, the process search is repeated until the experimental fiber properties meet the requirements.

[0079] The present invention will be further illustrated below with specific embodiments:

[0080] Example 1

[0081] This embodiment proposes a method for developing a preparation process for high-strength, high-modulus polyacrylonitrile-based carbon fibers, as detailed below:

[0082] (I) Constructing the prediction model (overall model and sub-models)

[0083] 1) Construct a standardized carbon fiber preparation system dataset in Excel or MATLAB software, including: polymerization-spinning-carbonization process data, fiber performance data corresponding to the process, and corresponding fiber structure data.

[0084] The laboratory, pilot-scale R&D data, engineering R&D data, and industrial production data of different grades of carbon fiber (T300, T700, T800, T1000, T1100, M40J, M46J, M55J, M60J, M65J, etc.) are unified, standardized, and normalized.

[0085] Data from different R&D stages are collected, filtered, supplemented, and standardized and normalized.

[0086] Collect, screen, and organize systematic research data from the past and recent worldwide (articles / patents containing information on systematic processes, fiber properties, or fiber structures, etc.), and standardize and normalize the data.

[0087] 2) Construction of the overall model and sub-models (conducted in MATLAB software, which includes embedded machine learning algorithms).

[0088] Overall Model: Establish a backpropagation neural network model for the entire process parameters → carbon fiber performance. Using the entire process data as input and the carbon fiber performance data (strength, modulus) as output, set the initial number of hidden layers and nodes in the neural network.

[0089] Before training a BP neural network model, the standardized and normalized dataset can be randomly divided into a training set (45-90%), a validation set (5-20%), and a test set (5-35%). The sum of the proportions of the training set, validation set, and test set is always 1. The training set is used to train the neural network model, and the validation set is used to prevent overfitting. The training set and validation set are used during the training of the BP neural network model. The test set is used to test the accuracy of the neural network model and is used after the model training is completed.

[0090] Training the overall model: Select the transfer function and training method and train the overall model; the overall accuracy (training, validation and testing) of the overall model is not less than 0.90; accuracy is the linear regression correlation between the predicted value and the actual value obtained by taking the dataset as input and using the trained model.

[0091] Training methods mainly include steepest descent, additional momentum, adaptive learning rate, momentum-adaptive learning rate adjustment, and quantized conjugate gradient algorithm. The following examples use the steepest descent algorithm, but other algorithms can also be selected.

[0092] First sub-model: Establishing a BP neural network model for the roundness of the cross-section of the coagulated nascent fiber by polymerization and spinning coagulation process parameters (first sub-model). Using polymerization and spinning coagulation process data as input and the roundness of the nascent coagulated fiber as output, the initial number of hidden layers and the initial number of hidden layer nodes of the neural network are set to establish a BP neural network model for predicting the roundness of the nascent coagulated fiber from the polymerization and spinning process;

[0093] Before model training, the standardized and normalized dataset can be randomly divided into a training set (45-85%), a validation set (5-20%), and a test set (10-35%). The sum of the proportions of the training, validation, and test sets is always 1. The training set is used to train the neural network model, the validation set is used to prevent overfitting, and both are used during the training of the BP neural network model. The test set is used to test the accuracy of the neural network model and is used after model training is complete.

[0094] Training the first sub-model: Select the transfer function and training method and train sub-model 1; the comprehensive accuracy (training, validation and testing) feature of sub-model 1 is not less than 0.92; accuracy is the linear regression correlation between the predicted value and the actual value obtained by taking the dataset as input and the trained model.

[0095] Second sub-model: Construct a BP neural network model (second sub-model) based on the dataset to determine the polymerization and spinning process parameters → raw filament properties (strength and modulus), and train it (construction process is the same as the first sub-model);

[0096] The third sub-model: Construct a BP neural network model (third sub-model) based on the dataset to determine the pre-oxidation process parameters → pre-oxidized fiber properties (strength and modulus), and train it (the construction process is the same as the second sub-model);

[0097] Fourth sub-model: Construct a BP neural network model (fourth sub-model) based on the dataset to determine the spinning process parameters → core-sheath structure thickness of the raw filament, and train it (construction process is the same as the third sub-model).

[0098] (II) Output Criteria for Overall Model and Sub-models

[0099] The overall model predicts the strength and modulus of carbon fibers, and the output criteria (the target range of fiber performance in this embodiment) are: tensile strength of 5.65-6.0 GPa and tensile modulus of 340-380 GPa.

[0100] Sub-model 1 (first sub-model) predicts the roundness of the initial solidified filament, with the output criterion being a roundness of 0.80-0.95.

[0101] Sub-model 2 (second sub-model) predicts the output of the raw fiber strength and modulus, with the following output criteria: tensile strength of 460-810MPa and tensile modulus of 7-13GPa.

[0102] Sub-model 3 (third sub-model) predicts the output of pre-oxidized fiber strength and modulus, with the following output criteria: tensile strength of 360-860MPa and tensile modulus of 7-12GPa.

[0103] Sub-model 4 (fourth sub-model) predicts the thickness of the protofilament skin, with an output criterion of 50-180nm.

[0104] (III) Set the parameter range for each process point in the overall process parameters as follows:

[0105] 1) Preparation of spinning solution:

[0106] Acrylonitrile monomers, itaconic acid monomers, and 2-acrylamido-2-methylpropanesulfonic acid monomers are blended in a mass ratio of (90.0-99.9):(0.1-5):(0.00-5.0) and added to a reactor equipped with a stirrer. The total monomer mass accounts for 10-25% of the mass of the reaction solution (the mass of the reaction solution is the sum of the masses of monomers, initiator, and solvent). Azobisisobutyronitrile is used as the initiator, with an initiator amount of 0.1-2.0% of the total monomer weight. The solvent is selected from one of dimethylformamide, dimethylacetamide, dimethyl sulfoxide, ethylene carbonate, propylene carbonate, and γ-butyrolactone for solution polymerization. The polymerization reaction temperature is controlled within the range of 30-85℃, and the polymerization reaction time is 3-36 h to obtain a polyacrylonitrile spinning solution.

[0107] 2) Solidification and shaping:

[0108] Wet spinning is employed, with the spinning solution being ejected from the spinneret and then sequentially passed through 2-8 stages of coagulation baths for solidification. The coagulation bath consists of three parts: solvent, coagulant, and regulator. The solvent is selected from one of dimethylformamide, dimethylacetamide, dimethyl sulfoxide, ethylene carbonate, propylene carbonate, and γ-butyrolactone; the coagulant is deionized water; and the regulator is selected from one of ethanol, n-propanol, n-butanol, isobutanol, ammonium bicarbonate, and ammonia.

[0109] a) Primary solidification molding

[0110] The temperature of the coagulation bath is 35-65℃; the coagulation stretching ratio is 0.4-1.2 times; the circulation rate of the coagulation bath is 4-14 mm / s; the coagulation residence time is 0.1-3 min; the solvent mass fraction in the coagulation bath is 35-70%; and the concentration of the regulator in the coagulation bath is 0.01-0.1 mol / L.

[0111] b) Final stage solidification

[0112] The temperature of the coagulation bath is 50-95℃; the coagulation stretching ratio is 1-2 times; the circulation flow rate of the coagulation bath is 4-14 mm / s; the coagulation residence time is 0.2-1 min; the solvent mass fraction in the coagulation bath is 0-15%; and the molar concentration of the regulator in the coagulation bath is 0.01-0.03 mol / L.

[0113] 3) Water washing process

[0114] The water washing temperature is 40-80℃; the water washing time is 1-3 minutes.

[0115] 4) Hot water stretching process

[0116] The hot water stretching temperature is 70-95℃; the stretching ratio is 1-4 times.

[0117] 5) Oiling process

[0118] Use self-prepared silicone oil, with a concentration of 0.5-3%.

[0119] 6) Drying and densification process

[0120] A low-temperature, long-time gradient drying densification process is adopted.

[0121] The temperature gradient has 4-18 stages; each stage uses a different temperature, with the temperature increasing progressively. The first stage temperature is 80-100℃; the last stage temperature is 115-140℃; the temperature difference between adjacent temperature gradients is 0-8℃; the drying and densification time for each stage is 4-9 seconds.

[0122] 7) Steam drawing and shrinkage heat setting process

[0123] Steam drawing process: Using saturated steam or superheated steam as the medium, a high-ratio drawing is applied to the fiber bundle. The steam pressure is 0.15-0.4 MPa; the drawing ratio is 2-4 times; and the steam drawing residence time is 2-5 seconds.

[0124] Shrink heat setting process: using saturated steam or superheated steam as the medium. Steam temperature is 110-180℃; stretch ratio is 0.9-1; shrink heat setting residence time is 2-5s.

[0125] 8) Pre-oxidation process

[0126] A 3-8 stage hot air medium pre-oxidation process is adopted, with the pre-oxidation temperature increasing step by step. The first stage pre-oxidation temperature is 170-200℃; the last stage pre-oxidation temperature is 240-280℃; the temperature difference between adjacent temperature gradients is 5-35℃; the total pre-oxidation draw ratio is 0.8-1.4 times; and the total pre-oxidation residence time is 20-90 min.

[0127] 9) Low-temperature carbonization process

[0128] High-purity nitrogen protection, low-temperature carbonization temperature is 300-900℃; low-temperature carbonization time is 1-6min.

[0129] 10) High-temperature carbonization process

[0130] Under high-purity nitrogen protection, the carbonization process is carried out in stages 2-6. The first stage temperature is 900-1000℃; the highest carbonization temperature is 1300-2100℃; the temperature difference between adjacent temperature gradients is 50-600℃; and the high-temperature carbonization time is 0.4-5 min.

[0131] (iv) Fiber test performance and verified optimal overall process parameters

[0132] With carbon fiber strength and modulus as targets (the target range for fiber performance in this embodiment is tensile strength 5.65-6.0 GPa and tensile modulus 340-380 GPa), a process search is performed based on the prediction outputs of the overall model and sub-models (see [link]). Figure 2 After five rounds of prediction, experimentation, and correction, the model finally output 20 sets of process parameters. Through experimental verification, the fiber performance of five sets of process parameters across the entire line reached the target range.

[0133] The five groups of fibers that achieved the target test performance are listed in Table 1:

[0134] Table 1

[0135]

[0136] Among them, the carbon fiber corresponding to number 2 has the best performance, and its corresponding full-line process parameters are as follows:

[0137] 1) Preparation of spinning solution

[0138] Acrylonitrile monomers, itaconic acid monomers, and 2-acrylamido-2-methylpropanesulfonic acid monomers were added to a reactor equipped with a stirrer in a ratio of 98.75:1.24:0.01. The total monomers comprised 22% of the total reaction solution by mass, and the initiator azobisisobutyronitrile (AIBN) accounted for 1.0% of the total monomer weight. Solution polymerization was carried out using dimethyl sulfoxide (DMSO) as the solvent. During polymerization, high-purity nitrogen gas was first bubbled into the reactor for 15 minutes to replace the air in the reaction system. Then, stirring was initiated, and the temperature was raised to 68°C. Under nitrogen protection, polymerization was carried out at this constant temperature for 16 hours to obtain the ternary copolymer polyacrylonitrile-based carbon fiber spinning solution.

[0139] 2) Solidification and molding of polymer solution

[0140] Wet spinning is employed, with the spinning solution being ejected from the spinneret and then solidified through eight stages of coagulation baths. The composition of each stage of the coagulation bath is as follows: dimethyl sulfoxide as the solvent, water as the coagulant, and ammonia as the modifier.

[0141] a) First-stage solidification

[0142] Coagulation bath temperature: 55℃; coagulation stretching: 0.65 times; coagulation bath circulation rate: 14mm / s; coagulation residence time: 0.2min; solvent mass fraction in coagulation bath: 55%; modifier molar concentration: 0.01mol / L.

[0143] b) Second-stage solidification

[0144] Coagulation bath temperature: 65℃; coagulation stretching: 1.05 times; coagulation bath circulation rate: 4mm / s; coagulation residence time: 0.3min; solvent mass fraction in coagulation bath: 30%; modifier molar concentration: 0.03mol / L.

[0145] c) Three-stage solidification molding

[0146] The temperature of the coagulation bath was 68℃; the coagulation stretching ratio was 1.65 times; the circulation rate of the coagulation bath was 5 mm / s; the coagulation residence time was 2 min; the solvent mass fraction in the coagulation bath was 35% and the molar concentration of the regulator was 0.025 mol / L.

[0147] d) Four-stage solidification molding

[0148] The temperature of the coagulation bath was 72℃; the coagulation stretching ratio was 1.4 times; the circulation rate of the coagulation bath was 5 mm / s; the coagulation residence time was 1.6 min; the solvent mass fraction in the coagulation bath was 18%; and the molar concentration of the regulator was 0.015 mol / L.

[0149] e) Five-stage solidification molding

[0150] The temperature of the coagulation bath was 76℃; the coagulation stretching ratio was 1.4 times; the circulation rate of the coagulation bath was 5 mm / s; the coagulation residence time was 1.5 min; the solvent mass fraction in the coagulation bath was 15%; and the molar concentration of the regulator was 0.015 mol / L.

[0151] f) Six-stage solidification molding

[0152] The temperature of the coagulation bath was 77℃; the coagulation stretching ratio was 1.3 times; the circulation rate of the coagulation bath was 8 mm / s; the coagulation residence time was 1.4 min; the solvent mass fraction in the coagulation bath was 10%; and the molar concentration of the regulator was 0.015 mol / L.

[0153] g) Seven-stage solidification molding

[0154] The temperature of the coagulation bath was 82℃; the coagulation stretching ratio was 1.5 times; the circulation rate of the coagulation bath was 10 mm / s; the coagulation residence time was 1.1 min; the solvent mass fraction in the coagulation bath was 7%; and the molar concentration of the regulator was 0.015 mol / L.

[0155] h) Eight-stage solidification molding

[0156] The temperature of the coagulation bath was 85℃; the coagulation stretching ratio was 1.9 times; the circulation rate of the coagulation bath was 14 mm / s; the coagulation residence time was 1 min; the solvent mass fraction in the coagulation bath was 1%; and the molar concentration of the regulator was 0.01 mol / L.

[0157] 3) Water washing process

[0158] Washing temperature: 70℃; washing time: 2min.

[0159] 4) Hot water stretching process

[0160] The filament bundle is drawn in hot water at 80-95℃, with a drawing ratio of 2 times.

[0161] 5) Oiling process

[0162] We use a self-prepared silicone oil with a concentration of 0.8%.

[0163] 6) Drying and densification process

[0164] Temperature gradient levels: 6 levels, with temperature gradients of 95℃, 102℃, 108℃, 110℃, 115℃, and 120℃ respectively; drying time for each level: 4 seconds.

[0165] 7) Steam drawing and shrinkage heat setting process

[0166] Steam drawing process: Using saturated steam as the medium, the fiber bundle is subjected to high-ratio drawing. Saturated steam pressure: 0.25MPa; drawing ratio: 2 times; steam drawing residence time: 3s.

[0167] Shrink heat setting process: using superheated steam as the medium, superheated steam temperature: 150℃; stretch ratio: 0.92 times; shrink heat setting residence time: 3s.

[0168] 8) Pre-oxidation process

[0169] Under hot air medium, a 5-stage pre-oxidation process is adopted. The pre-oxidation temperatures are 198℃, 212℃, 228℃, 238℃, and 258℃ respectively; the total pre-oxidation draw ratio is 1.25 times; and the total pre-oxidation residence time is 50 min.

[0170] 9) Low-temperature carbonization process

[0171] Under high-purity nitrogen protection, the low-temperature carbonization temperature is 300-900℃; the low-temperature carbonization time is 4min.

[0172] 10) High-temperature carbonization process

[0173] Under high-purity nitrogen protection, five levels of high-temperature carbonization were performed. The carbonization temperatures were 900℃, 1100℃, 1400℃, 1750℃, and 1350℃ respectively; the high-temperature carbonization time was 2 minutes.

[0174] Example 2

[0175] This embodiment proposes a method for developing a preparation process for high-strength, high-modulus polyacrylonitrile-based carbon fibers, as detailed below:

[0176] (I) Constructing the prediction model (overall model and sub-models)

[0177] Same as Example 1

[0178] (II) Output Criteria for Overall Model and Sub-models

[0179] The overall model predicts the strength and modulus of carbon fibers, and the output criteria (the target range of fiber performance in this embodiment) are: tensile strength of 5.1-5.7 GPa and tensile modulus of 385-430 GPa.

[0180] Sub-model 1 (first sub-model) predicts the roundness of the initial solidified filament, with the output criterion being a roundness of 0.85-0.95;

[0181] Sub-model 2 (second sub-model) predicts the output of the raw fiber strength and modulus, with the following output criteria: tensile strength of 620-890MPa and tensile modulus of 8-13GPa.

[0182] Sub-model 3 (third sub-model) predicts the strength and modulus of pre-oxidized fibers, with the following output criteria: tensile strength of 510-830 MPa and tensile modulus of 8-11 GPa.

[0183] Sub-model 4 (fourth sub-model) predicts the thickness of the original silk cortex, and its output criterion is 200-450nm.

[0184] (III) Setting the parameter range for each process point in the overall process parameters.

[0185] Same as Example 1.

[0186] (iv) Fiber test performance and experimentally verified output process parameters

[0187] With carbon fiber strength and modulus as targets (the target range for fiber performance in this embodiment is: tensile strength of 5.1-5.7 GPa and tensile modulus of 385-430 GPa), and based on the predicted outputs of the overall model and sub-models, such as... Figure 2 As shown, the process search was performed, and after three rounds of prediction-experiment-correction model, 20 sets of process parameters were finally output. After experimental verification, the fiber performance of 7 sets of process parameters across the entire line reached the target range.

[0188] Table 2 lists the seven fiber test performances that achieved the target:

[0189] Table 2

[0190]

[0191] Among them, the carbon fiber corresponding to number 1 has the best performance, and its corresponding process parameters for the entire production line are as follows:

[0192] 1) Preparation of spinning solution

[0193] Acrylonitrile, itaconic acid, and 2-acrylamido-2-methylpropanesulfonic acid monomers were blended in a ratio of 98.5:1.4:0.1 and added to a reactor equipped with a stirrer. The total monomers comprised 22% of the solution by mass, and the initiator azobisisobutyronitrile (AIBN) accounted for 1.0% of the total monomer weight. Solution polymerization was carried out using dimethyl sulfoxide (DMSO) as the solvent. During polymerization, high-purity nitrogen gas was first bubbled into the reactor for 15 minutes to displace the air in the reaction system. Then, stirring was initiated, and the temperature was raised to 62°C. Under nitrogen protection, polymerization was carried out at this constant temperature for 20 hours to obtain the ternary copolymer polyacrylonitrile-based carbon fiber spinning solution.

[0194] 2) Solidification and molding of polymer solution

[0195] The process employs wet spinning, where the polymerization solution is ejected from the spinneret and then coagulated through a six-stage coagulation bath. The coagulation bath consists of: solvent—dimethyl sulfoxide, coagulant—water, and regulator—ammonia.

[0196] a) First-stage solidification

[0197] Coagulation bath temperature: 52℃; coagulation stretching: 0.66 times; coagulation bath circulation rate: 14mm / s; coagulation residence time: 0.2min; solvent mass fraction in coagulation bath: 54%; modifier molar concentration: 0.01mol / L.

[0198] b) Second-stage solidification

[0199] Coagulation bath temperature: 65℃; coagulation stretching: 1.05 times; coagulation bath circulation rate: 4mm / s; coagulation residence time: 0.3min; solvent mass fraction in coagulation bath: 45%; modifier concentration: 0.03mol / L.

[0200] c) Three-stage solidification molding

[0201] The temperature of the coagulation bath was 68℃; the coagulation stretching ratio was 1.65 times; the circulation rate of the coagulation bath was 5 mm / s; the coagulation residence time was 2 min; the solvent mass fraction in the coagulation bath was 35%; and the molar concentration of the regulator was 0.012 mol / L.

[0202] d) Four-stage solidification molding

[0203] The temperature of the coagulation bath was 72℃; the coagulation stretching ratio was 1.4 times; the circulation rate of the coagulation bath was 5 mm / s; the coagulation residence time was 1.6 min; the solvent mass fraction in the coagulation bath was 18%; and the molar concentration of the regulator was 0.012 mol / L.

[0204] e) Five-stage solidification molding

[0205] The temperature of the coagulation bath was 76℃; the coagulation stretching ratio was 1.4 times; the circulation rate of the coagulation bath was 5 mm / s; the coagulation residence time was 1.5 min; the solvent mass fraction in the coagulation bath was 15%; and the molar concentration of the regulator was 0.012 mol / L.

[0206] f) Six-stage solidification molding

[0207] The temperature of the coagulation bath was 77℃; the coagulation stretching ratio was 1.3 times; the circulation rate of the coagulation bath was 8 mm / s; the coagulation residence time was 1.4 min; the solvent mass fraction in the coagulation bath was 5%; and the molar concentration of the regulator was 0.01 mol / L.

[0208] 3) Water washing process

[0209] Washing temperature: 60℃; washing time: 2min.

[0210] 4) Hot water stretching process

[0211] The filament bundle is drawn in hot water at 80-95℃, with a drawing ratio of 2 times.

[0212] 5) Oiling process

[0213] We use a self-prepared silicone oil with a concentration of 0.8%.

[0214] 6) Drying and densification process

[0215] Temperature gradient levels: 6 levels, with temperature gradients of 100℃, 102℃, 110℃, 112℃, 115℃, and 120℃ respectively; drying time for each level: 4 seconds.

[0216] 7) Steam drawing and shrinkage heat setting process

[0217] Steam drawing process: Using saturated steam as the medium, the fiber bundle is subjected to high-ratio drawing. Saturated steam pressure: 0.3MPa; drawing ratio: 3 times; steam drawing residence time: 3s.

[0218] Shrink heat setting process: using superheated steam as the medium, superheated steam temperature: 150℃; stretch ratio: 0.92 times; shrink heat setting residence time: 3s.

[0219] 8) Pre-oxidation process

[0220] Under hot air medium, a 5-stage pre-oxidation process is adopted. The pre-oxidation temperatures are 190℃, 212℃, 230℃, 238℃, and 260℃ respectively; the total pre-oxidation draw ratio is 1.25 times; and the total pre-oxidation residence time is 45 min.

[0221] 9) Low-temperature carbonization process

[0222] Under high-purity nitrogen protection, the low-temperature carbonization temperature is 300-900℃; the low-temperature carbonization time is 4min.

[0223] 10 High-temperature carbonization process

[0224] Under high-purity nitrogen protection, five levels of high-temperature carbonization were performed. The carbonization temperatures were 950℃, 1200℃, 1450℃, 1950℃, and 1450℃ respectively; the high-temperature carbonization time was 4 minutes.

[0225] Example 3

[0226] This embodiment proposes a method for developing a preparation process for high-strength, high-modulus polyacrylonitrile-based carbon fibers, as detailed below:

[0227] (I) Constructing the prediction model (overall model and sub-models)

[0228] Same as Example 1

[0229] (II) Output Criteria for Overall Model and Sub-models

[0230] The overall model predicts the strength and modulus of carbon fibers, and the output criteria (the target range of fiber performance in this embodiment) are: tensile strength of 4.6-5.2 GPa and tensile modulus of 420-480 GPa.

[0231] Sub-model 1 (first sub-model) predicts the roundness of the initial solidified filament, with the output criterion being a roundness of 0.75-0.90.

[0232] Sub-model 2 (second sub-model) predicts the output of the raw fiber strength and modulus, with the following output criteria: tensile strength of 720-890 MPa and tensile modulus of 8.5-14 GPa.

[0233] Sub-model 3 (third sub-model) predicts the strength and modulus of pre-oxidized fiber, with the following output criteria: tensile strength of 510-930 MPa and tensile modulus of 9-13 GPa.

[0234] Sub-model 4 (fourth sub-model) predicts the thickness of the protofilament skin, with an output criterion of 300-600nm.

[0235] (III) Setting the parameter range for each process point in the overall process parameters.

[0236] Same as Example 1.

[0237] (iv) Fiber test performance and experimentally verified output process parameters

[0238] With carbon fiber strength and modulus as targets (in this embodiment, the target fiber performance range is: tensile strength of 4.6-5.2 GPa and tensile modulus of 420-480 GPa), and based on the predicted outputs of the overall model and sub-models, such as... Figure 2 As shown, the process search was performed, and after 6 rounds of prediction-experiment-correction model, 20 sets of process parameters were finally output. After experimental verification, the fiber performance of 4 sets of process parameters for the entire line reached the target range.

[0239] Table 3 lists the four fiber test performances that achieved the target:

[0240] Table 3

[0241]

[0242] Among them, the carbon fiber corresponding to number 2 has the best performance, and its corresponding full-line process parameters are as follows:

[0243] 1) Preparation of spinning solution

[0244] Acrylonitrile, itaconic acid, and 2-acrylamido-2-methylpropanesulfonic acid monomers were blended in a ratio of 98.3:1.6:0.1 and added to a reactor equipped with a stirrer. The total monomers comprised 21% of the solution by mass, and the initiator azobisisobutyronitrile (AIBN) accounted for 0.9% of the total monomer weight. Solution polymerization was carried out using dimethyl sulfoxide (DMSO) as the solvent. During polymerization, high-purity nitrogen gas was first bubbled into the reactor for 15 minutes to displace the air in the reaction system. Then, stirring was initiated, and the temperature was raised to 68°C. Under nitrogen protection, polymerization was carried out at this constant temperature for 12 hours to obtain the ternary copolymer polyacrylonitrile-based carbon fiber spinning solution.

[0245] 2) Solidification and molding of polymer solution

[0246] The process employs wet spinning, where the polymerization solution is ejected from the spinneret and then coagulated in a two-stage coagulation bath. The coagulation bath consists of: solvent—dimethyl sulfoxide, coagulant—water, and regulator—ammonia.

[0247] a) First-stage solidification

[0248] Coagulation bath temperature: 50℃; coagulation stretching: 0.62 times; coagulation bath circulation rate: 14mm / s; coagulation residence time: 0.2min; solvent mass fraction in coagulation bath: 52%; modifier molar concentration: 0.01mol / L.

[0249] b) Second-stage solidification

[0250] Coagulation bath temperature: 60℃; coagulation stretching: 1.05 times; coagulation bath circulation rate: 4mm / s; coagulation residence time: 0.3min; solvent mass fraction in coagulation bath: 14%; modifier molar concentration: 0.03mol / L.

[0251] 3) Water washing process

[0252] Washing temperature: 60℃; washing time: 2min.

[0253] 4) Hot water stretching process

[0254] The filament bundle is drawn in hot water at 80-95℃, with a drawing ratio of 2 times.

[0255] 5) Oiling process

[0256] We use a self-prepared silicone oil with a concentration of 0.8%.

[0257] 6) Drying and densification process

[0258] Temperature gradient levels: 6 levels, with temperature gradients of 100℃, 102℃, 108℃, 110℃, 115℃, and 120℃ respectively; drying time for each level: 4 seconds.

[0259] 7) Steam drawing and shrinkage heat setting process

[0260] Steam drawing process: Using saturated steam as the medium, the fiber bundle is subjected to high-ratio drawing. Saturated steam pressure: 0.3MPa; drawing ratio: 3.5 times; steam drawing residence time: 3s.

[0261] Shrink heat setting process: using superheated steam as the medium, superheated steam temperature: 150℃; stretch ratio: 0.95 times; shrink heat setting residence time: 3s.

[0262] 8) Pre-oxidation process

[0263] Under hot air medium, a 5-stage pre-oxidation process is adopted. The pre-oxidation temperatures are 190℃, 212℃, 230℃, 238℃, and 260℃ respectively; the total pre-oxidation draw ratio is 1.25 times; and the total pre-oxidation residence time is 45 min.

[0264] 9) Low-temperature carbonization process

[0265] Under high-purity nitrogen protection, the low-temperature carbonization temperature is 300-900℃; the low-temperature carbonization time is 4min.

[0266] 10) High-temperature carbonization process

[0267] Under high-purity nitrogen protection, six levels of high-temperature carbonization were performed. The carbonization temperatures were 950℃, 1100℃, 1450℃, 2050℃, 1650℃, and 1450℃ respectively; the high-temperature carbonization time was 3 minutes.

[0268] Comparative Example 1

[0269] Comparative Example 1 proposes a method for developing a preparation process for high-strength, high-modulus polyacrylonitrile-based carbon fibers, as detailed below:

[0270] (I) Constructing the prediction model (overall model and sub-models)

[0271] Compared to Example 1, sub-model 4 (fourth sub-model) has been removed, while the rest are the same as in Example 1.

[0272] (II) Output Criteria for Overall Model and Sub-models

[0273] Compared to Example 1, the output criterion of sub-model 4 (fourth sub-model) has been removed, while the rest are the same as in Example 1.

[0274] (III) Setting the parameter range for each process point in the overall process parameters.

[0275] Same as Example 1

[0276] (iv) Fiber test performance and verified optimal overall process parameters

[0277] With carbon fiber strength and modulus as targets (target range of fiber properties in Comparative Example 1: tensile strength 5.65-6.0 GPa, tensile modulus 340-380 GPa), a process search is performed based on the prediction outputs of the overall model and the three sub-models (see [link]). Figure 2 ).

[0278] After 12 rounds of prediction, experimentation, and model correction, 20 sets of process parameters were finally output. Experimental verification showed that the fiber properties of 3 sets of process parameters were close to the target range (both strength and modulus were close).

[0279] Table 4 lists the test performance of the three groups of fibers that are close to the target:

[0280] Table 4

[0281]

[0282] Compared with Example 1, Comparative Example 1 has one less sub-model. According to the analysis of past data, the output parameters of the sub-model are highly correlated with the performance of high strength and high modulus of carbon fiber. With one less sub-model, the process search space range increases, the trial and error probability of process verification increases to a certain extent, resulting in an increase in the process development cycle of the target and an increase in input consumption (the comparative example still did not meet the standard after 12 rounds of test results).

[0283] Additionally, it should be noted that there should not be too many sub-models, as too many will limit the process search process and hinder the process development process.

[0284] Comparative Example 2

[0285] Comparative Example 2 proposes a method for developing a preparation process for high-strength, high-modulus polyacrylonitrile-based carbon fibers, as detailed below:

[0286] (I) Constructing the prediction model (overall model and sub-models)

[0287] Compared to Example 1, Sub-model 4 (the fourth sub-model) is different, but the rest are the same as Example 1.

[0288] Sub-model 4 (fourth sub-model) construction: Based on the dataset, construct a BP neural network model (fourth sub-model) for spinning process parameters → raw filament orientation, and train it.

[0289] (II) Output Criteria for Overall Model and Sub-models

[0290] Compared to Example 1, the output criterion for Sub-model 4 (fourth sub-model) is 75-88%;

[0291] The rest are the same as in Example 1.

[0292] (III) Setting the parameter range for each process point in the overall process parameters.

[0293] Same as Example 1.

[0294] (iv) Fiber test performance and verified optimal overall process parameters

[0295] With carbon fiber strength and modulus as targets (the fiber performance target range for Comparative Example 2: tensile strength 5.65-6.0 GPa, tensile modulus 340-380 GPa), a process search is performed based on the prediction outputs of the overall model and four sub-models (see [link to relevant documentation]). Figure 2After 12 rounds of prediction-experiment-correction model, 20 sets of process parameters were finally output. After experimental verification, the fiber strength of 2 sets of process parameters reached the target, and the modulus of 3 sets of process parameters reached the target. However, the strength and modulus did not reach the target at the same time.

[0296] Compared with Example 1, Comparative Example 2 has a different content in the fourth sub-model. The parameters output by the sub-model are prioritized in relation to the high strength and high modulus performance characteristics. Selecting a sub-model with higher priority can speed up process development. After Comparative Example 2 changed the content of the fourth sub-model, its priority obviously decreased from the development process perspective, which led to an increase in the development cycle for achieving high strength and high modulus performance simultaneously.

[0297] Comparative Example 3

[0298] Comparative Example 3 proposes a method for developing a preparation process for high-strength, high-modulus polyacrylonitrile-based carbon fibers, as detailed below:

[0299] (I) Constructing the prediction model (overall model and sub-models)

[0300] Compared to Example 1, only the overall model and sub-model 1 (first sub-model) are retained, while sub-model 2 (second sub-model), sub-model 3 (third sub-model), and sub-model 4 (fourth sub-model) are removed.

[0301] The construction process of the overall model and sub-model 1 is the same as that of Example 1.

[0302] (II) Output Criteria for Overall Model and Sub-models

[0303] Compared to Example 1, only the output criteria of the overall model and sub-model 1 (first sub-model) are retained.

[0304] (III) Setting the parameter range for each process point in the overall process parameters.

[0305] (iv) Fiber test performance and verified optimal overall process parameters

[0306] With carbon fiber strength and modulus as targets (target fiber properties range for Comparative Example 3: tensile strength 5.65-6.0 GPa, tensile modulus 340-380 GPa), a process search is performed based on the prediction outputs of the overall model and one sub-model (see [link to relevant documentation]). Figure 2 After 24 rounds of prediction-experiment-correction modeling, 20 sets of process parameters were finally output. After experimental verification, the fiber performance of one set of process parameters for the entire line was close to the target range.

[0307] Compared with Example 1, Comparative Example 3 retains only one sub-model, which increases the space for process selection, the demand for computing resources, and the probability of process verification trial and error, as well as the development cycle and investment.

[0308] Comparative Example 4

[0309] Comparative Example 4 proposes a method for developing a preparation process for high-strength, high-modulus polyacrylonitrile-based carbon fibers, as detailed below:

[0310] (I) Constructing the prediction model (overall model)

[0311] Compared to Example 1, only the overall model is retained, and the overall model construction process is the same as in Example 1.

[0312] (II) Overall Model Output Criterion

[0313] Compared to Example 1, only the overall model output criteria are retained.

[0314] (III) Setting the parameter range for each process point in the overall process parameters.

[0315] (iv) Fiber test performance and verified optimal overall process parameters

[0316] With carbon fiber strength and modulus as targets (target range of fiber properties in Comparative Example 4: tensile strength 5.65-6.0 GPa, tensile modulus 340-380 GPa), a process search is performed based on the overall model prediction output (see [link]). Figure 2 After 30 rounds of prediction-experiment-correction modeling, 20 sets of process parameters were finally output. Experimental verification showed that the fiber performance of all process parameter groups still had a large gap with the target performance.

[0317] Compared to Example 1, Comparative Example 4, using only a single overall model, suffers from excessively long global process flow, numerous process parameters, and an enormous search volume during the iterative process search, consuming significant computational resources. Furthermore, the low prediction accuracy of the overall model leads to a substantial increase in the probability of trial and error in experimental verification. In summary, despite a significant increase in computational resources and experimental investment compared to Example 1, Comparative Example 4 still fails to achieve satisfactory process development results.

[0318] The above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention in any way. Any simple modifications, equivalent changes, and alterations made to the above embodiments based on the technical essence of the present invention shall still fall within the scope of the technical solution of the present invention.

Claims

1. A method for developing a preparation process for high-strength, high-modulus polyacrylonitrile-based carbon fibers, characterized in that, The development method includes the following steps: The steps for constructing the prediction model are as follows: Based on the process dataset of polyacrylonitrile-based carbon fiber, a prediction model including a total model and multiple sub-models is constructed. The total model is a machine learning algorithm model that takes the entire process parameters of polyacrylonitrile-based carbon fiber preparation as input and outputs the performance data and / or structural data of polyacrylonitrile-based carbon fiber. The sub-models are machine learning algorithm models that take the local process parameters of polyacrylonitrile-based carbon fiber preparation as input and output the fiber performance data and / or structural data after local process treatment. The steps for setting the criteria are as follows: Based on the target performance of the required polyacrylonitrile-based carbon fiber, set the parameter range for each process point in the overall process parameters, set the output criteria for each sub-model, and set the output criteria for the overall model. Process search steps: The parameter range of each process point in the entire process parameters is gridded and iteratively searched for the entire process parameters. The prediction model is used for prediction in each iteration. The steps of using the prediction model for prediction include: first running the sub-models, and then running the overall model when the outputs of all the sub-models meet the corresponding output criteria conditions. When the data of at least one sub-model does not meet the corresponding output criterion, a set of full-line process parameters is generated through a new loop iteration, and the prediction model is used again for prediction. Among them, after running the overall model, if the output of the overall model does not meet the output criterion conditions, a set of process parameters for the entire line will be generated through a new loop iteration, and the prediction model will be used again for prediction. In this process, after running the overall model, when the output of the overall model meets the output criterion conditions, the corresponding full-line process parameters are recorded as the selected full-line process parameters for use in preparing polyacrylonitrile-based carbon fibers with the required properties.

2. The development method for the preparation process of high-strength, high-modulus polyacrylonitrile-based carbon fiber according to claim 1, characterized in that, Following the process search step, the following is also included: Experimental verification steps: Conduct experimental verification of the selected process parameters for the entire production line; If the performance of the prepared polyacrylonitrile-based carbon fiber reaches the target value, then the process development is complete. If the performance of the prepared polyacrylonitrile carbon fiber does not meet the target value, the experimental process data will be added to the process dataset of polyacrylonitrile-based carbon fiber, the overall model and sub-models will be retrained, and the process search and experimental verification steps will be repeated.

3. The method for developing a preparation process for high-strength, high-modulus polyacrylonitrile-based carbon fibers according to claim 1, characterized in that, The process dataset for polyacrylonitrile-based carbon fiber includes: process parameter data for the entire process of preparing polyacrylonitrile-based carbon fiber, fiber structure data, and fiber performance data.

4. The development method of the preparation process of high-strength, high-modulus polyacrylonitrile-based carbon fiber according to claim 3, characterized in that, The full-line process parameter data includes: polymerization reaction process parameters, spinning process parameters, pre-oxidation process parameters, low-temperature carbonization process parameters, and high-temperature carbonization process parameters.

5. The method for developing a preparation process for high-strength, high-modulus polyacrylonitrile-based carbon fibers according to claim 4, characterized in that, The polymerization process parameters include: the polymerization formulation, temperature, and time.

6. The method for developing a preparation process for high-strength, high-modulus polyacrylonitrile-based carbon fibers according to claim 4, characterized in that, The spinning process parameters include: solidification and molding process parameters, water washing process parameters, hot water drawing process parameters, oil concentration, drying and densification process parameters, steam drawing process parameters, and shrinkage heat setting process.

7. The method for developing a preparation process for high-strength, high-modulus polyacrylonitrile-based carbon fibers according to claim 4, characterized in that, The pre-oxidation process parameters include pre-oxidation temperature, pre-oxidation time, and draw ratio.

8. The method for developing a preparation process for high-strength, high-modulus polyacrylonitrile-based carbon fibers according to claim 4, characterized in that, The low-temperature carbonization process parameters include the low-temperature carbonization temperature and the low-temperature carbonization time.

9. The method for developing a preparation process for high-strength, high-modulus polyacrylonitrile-based carbon fibers according to claim 4, characterized in that, The high-temperature carbonization process parameters include the high-temperature carbonization temperature and the high-temperature carbonization time.

10. The method for developing a preparation process for high-strength, high-modulus polyacrylonitrile-based carbon fibers according to claim 3, characterized in that, The fiber performance data includes performance data for polyacrylonitrile precursor fibers and performance data for polyacrylonitrile-based carbon fibers.

11. The method for developing a preparation process for high-strength, high-modulus polyacrylonitrile-based carbon fibers according to claim 3, characterized in that, The structural data of the fiber includes the thickness of the sheath layer in the fiber core-sheath structure and the roundness of the fiber cross-section.

12. The method for developing a preparation process for high-strength, high-modulus polyacrylonitrile-based carbon fibers according to claim 11, characterized in that, The structural data of the fiber also includes the fiber orientation.

13. The method for developing a preparation process for high-strength, high-modulus polyacrylonitrile-based carbon fibers according to claim 1, characterized in that, In the step of building the prediction model: The machine learning algorithm model is a BP neural network model; and / or Based on the process dataset of polyacrylonitrile-based carbon fiber, a machine learning algorithm model is constructed with the process parameters of the entire polyacrylonitrile-based carbon fiber preparation process as input and the performance data and / or structural data of polyacrylonitrile-based carbon fiber as output, and then trained to obtain the overall model.

14. The method for developing a preparation process for high-strength, high-modulus polyacrylonitrile-based carbon fibers according to claim 1, characterized in that, The plurality of sub-models includes: a first sub-model, a second sub-model, a third sub-model, and a fourth sub-model; wherein, Based on the process dataset of polyacrylonitrile-based carbon fiber, a machine learning algorithm model was constructed with polymerization reaction process parameters and solidification molding process parameters as inputs and the cross-sectional roundness of the nascent fiber as the output. The model was then trained to obtain the first sub-model. Based on the process dataset of polyacrylonitrile-based carbon fiber, a machine learning algorithm model was constructed with polymerization reaction process parameters and spinning process parameters as inputs and performance data of polyacrylonitrile precursor fiber as outputs. The model was then trained to obtain the second sub-model. Based on the process dataset of polyacrylonitrile-based carbon fiber, a machine learning algorithm model was constructed with pre-oxidation process parameters as input and pre-oxidized fiber performance data as output, and trained to obtain the third sub-model. Based on the process dataset of polyacrylonitrile-based carbon fiber, a machine learning algorithm model was constructed with spinning process parameters as input and the sheath thickness data of the core-sheath structure of polyacrylonitrile precursor fiber as output. The model was then trained to obtain the fourth sub-model.

15. The method for developing a preparation process for high-strength, high-modulus polyacrylonitrile-based carbon fibers according to claim 14, characterized in that, The performance data of the polyacrylonitrile precursor includes tensile strength data and tensile modulus data.

16. The method for developing a preparation process for high-strength, high-modulus polyacrylonitrile-based carbon fibers according to claim 14, characterized in that, The performance data of the pre-oxidized fiber includes tensile strength data and tensile modulus data.

17. The method for developing a preparation process for high-strength, high-modulus polyacrylonitrile-based carbon fibers according to claim 14, characterized in that, In the step of setting the criteria conditions: The criterion for the first sub-model is set as follows: the roundness of the cross-section of the nascent fiber is within a set range; wherein, the roundness of the cross-section of the nascent fiber is the ratio of the minor axis to the major axis of the cross-section of the nascent fiber; and / or The criteria for the second sub-model are set as follows: the strength of the polyacrylonitrile precursor fiber is within a set strength range; the modulus of the polyacrylonitrile precursor fiber is within a set modulus range; and / or The criteria for the third sub-model are set as follows: the tensile strength of the pre-oxidized fiber is within a set tensile strength range; the tensile modulus of the pre-oxidized fiber is within a set tensile modulus range; and / or The criterion for the fourth sub-model is set as follows: the sheath thickness of the polyacrylonitrile precursor fiber is a set thickness.

18. The method for developing a preparation process for high-strength, high-modulus polyacrylonitrile-based carbon fibers according to claim 17, characterized in that, The criterion for the first sub-model is set as follows: the roundness of the cross-section of the nascent fiber is 0.75-0.

95.

19. The method for developing a preparation process for high-strength, high-modulus polyacrylonitrile-based carbon fibers according to claim 17, characterized in that, The criteria for the second sub-model are set as follows: the strength of the polyacrylonitrile precursor is 450-900 MPa; the modulus of the polyacrylonitrile precursor is 7-15 GPa.

20. The method for developing a preparation process for high-strength, high-modulus polyacrylonitrile-based carbon fibers according to claim 17, characterized in that, The criteria for the third sub-model are set as follows: the tensile strength of the pre-oxidized fiber is 350-940 MPa; the tensile modulus of the pre-oxidized fiber is 7-14 GPa.

21. The method for developing a preparation process for high-strength, high-modulus polyacrylonitrile-based carbon fibers according to claim 17, characterized in that, The criterion for the fourth sub-model is set as follows: the sheath thickness of the polyacrylonitrile precursor fiber is 40-600 nm.

22. The method for developing a preparation process for high-strength, high-modulus polyacrylonitrile-based carbon fibers according to claim 1, characterized in that, In the step of setting the criteria conditions: The criteria for the overall model are: the tensile strength of the polyacrylonitrile-based carbon fiber is within a set tensile strength range; and the tensile modulus of the polyacrylonitrile-based carbon fiber is within a set tensile modulus range.

23. The method for developing a preparation process for high-strength, high-modulus polyacrylonitrile-based carbon fibers according to claim 22, characterized in that, In the step of setting the criteria conditions: The criteria for the overall model are: the tensile strength of the polyacrylonitrile-based carbon fiber is 4.6-6.0 GPa; and the tensile modulus of the polyacrylonitrile-based carbon fiber is 340-480 GPa.

24. The method for developing a preparation process for high-strength, high-modulus polyacrylonitrile-based carbon fibers according to claim 1, characterized in that, In the process search step: each set of full-line process parameters generated by iterative iteration is based on roulette wheel betting, a basic genetic algorithm, and / or In the process search step: when the output criterion is not met, a set of process parameters for the entire line is generated by iteratively re-looping according to the genetic algorithm.

25. The method for developing a preparation process for high-strength, high-modulus polyacrylonitrile-based carbon fibers according to claim 24, characterized in that, In the process search step: when the number of selected process parameters for the entire line is less than the set number of selected parameters, the process search step is repeated iteratively until the number of selected process parameters for the entire line reaches the set number of selected parameters.

26. The method for developing a preparation process for high-strength, high-modulus polyacrylonitrile-based carbon fibers according to claim 25, characterized in that, The number of groups is set to 20.

27. The method for developing a preparation process for high-strength, high-modulus polyacrylonitrile-based carbon fibers according to any one of claims 2-26, characterized in that, In the experimental verification steps: Test and verify one or more sets of full-line process parameters selected. If the performance of polyacrylonitrile-based carbon fiber prepared by one or more sets of full-line process parameters selected meets the target value, then the process development is completed. and / or If the performance of the prepared polyacrylonitrile-based carbon fiber fails to meet the target value, the experimental verification data should be added to the process dataset of polyacrylonitrile-based carbon fiber, and the overall model and sub-models should be retrained and corrected. Then, the process search and experimental verification steps are repeated.

28. A high-strength, high-modulus polyacrylonitrile-based carbon fiber, characterized in that, The high-strength, high-modulus polyacrylonitrile-based carbon fiber is prepared by a process developed according to the method for developing the preparation process of high-strength, high-modulus polyacrylonitrile-based carbon fiber as described in any one of claims 1-27.

29. The high-strength, high-modulus polyacrylonitrile-based carbon fiber according to claim 28, characterized in that, The high-strength, high-modulus polyacrylonitrile-based carbon fiber has a tensile strength of 4.6-6.0 GPa, a tensile modulus of 340-480 GPa, a diameter of 4.2-5.7 μm, and a bulk density of 1.71-1.88 g / cm³. 3 .

30. The high-strength, high-modulus polyacrylonitrile-based carbon fiber according to claim 28, characterized in that, The high-strength, high-modulus polyacrylonitrile-based carbon fibers were characterized by small-angle X-ray scattering (SAXS), where the scattering vector q ∈ (0.11, 0.45) Å. -1 When, the scattering intensity I(q)∝q -α α takes values ​​from 2.85 to 3.3; when the scattering vector q ∈ (0.025, 0.08) Å -1 When, the scattering intensity I(q)∝q -α α takes values ​​from 0.75 to 1.1.

Images