Optimization method and device for sintering process of optical fiber preform

Abstract

The application discloses a sintering process optimization method and device for an optical fiber preform, and belongs to the technical field of machine learning and process optimization. The method comprises the following steps: S1: obtaining the deposition diameter before sintering, the sintering formula and the sintering diameter in a sintering production line workshop, and performing cleaning treatment according to the sintering optical rod time; S2: obtaining fused sintering features according to the extracted category features and numerical features; S3: after performing difference value processing on the sintering features obtained in S2 under time series and adjacent samples, a difference value sintering feature is obtained by using a Person analysis method, and the difference value sintering feature is used as a model input parameter for predicting a difference value sintering diameter; S4: a difference value sintering diameter prediction model is established according to the difference value sintering feature, and the difference value sintering diameter is predicted; and S5: based on the difference value sintering diameter prediction model in S4 and a genetic algorithm network, a parameter intelligent recommendation algorithm is constructed, and the recommended parameters are input into the sintering production line. The application realizes accurate prediction and dynamic optimization of the sintering process of the optical fiber preform through the above scheme.

Description

Technical Field

[0001] This application relates to the fields of machine learning and process optimization technology, and in particular to a method and apparatus for optimizing the sintering process of optical fiber preforms. Background Technology

[0002] As the core raw material for manufacturing silica optical fibers, the production process of optical fiber preforms is crucial to the development of the optical fiber communication industry. Since the mass production of optical fibers, optical fiber preform manufacturing technology has continuously evolved, from the early one-step method to the now commonly used two-step method, which involves first manufacturing the core rod and then adding an outer cladding to form the preform. In this process, various processes such as improved chemical vapor deposition (CVD), axial vapor deposition (CVD), extra-rod CVD, and microwave plasma-activated CVD have emerged and continued to be improved.

[0003] In the sintering process of optical fiber preforms, diameter control is a key factor affecting the quality and performance of the preforms. Currently, the industry is making numerous efforts to optimize the diameter sintering process. Some companies are improving sintering equipment, such as adopting more precise temperature control systems, hoping to achieve fine-grained adjustment of the sintering process to stabilize the sintering diameter of the preforms; others are attempting to innovate in the type and flow control of process gases, trying to reduce diameter fluctuations caused by unstable gas environments.

[0004] However, this field still faces many challenges: On the one hand, existing processes struggle to predict the diameter changes of preforms during sintering accurately and in real time. The sintering process involves complex physicochemical reactions, including thermal expansion and contraction of materials and changes in the deposition rate of gaseous substances. These intertwined factors result in large errors in diameter prediction methods based on traditional empirical models, failing to meet the demands of high-precision production. On the other hand, even minor changes in the external environment during production, such as fluctuations in workshop temperature and humidity, significantly impact the sintering diameter. Existing technologies struggle to effectively compensate for and correct these environmental disturbances, making it difficult to guarantee the consistency of preform sintering diameter, thus affecting the quality stability and production efficiency of optical fiber products. Furthermore, traditional process optimization relies on a "trial and error-test-adjustment" cycle, with each parameter adjustment requiring 4-8 hours including sintering and testing processes. Each adjustment only covers a single variable, leading to low efficiency in parameter adjustment. Summary of the Invention

[0005] This application provides a method and apparatus for optimizing the sintering process of optical fiber preforms, which is used to solve at least one of the above-mentioned technical problems.

[0006] The technical solution adopted in this application is as follows:

[0007] On the one hand, this application provides a method for optimizing the sintering process of optical fiber preforms, the method comprising:

[0008] S1: Obtain the pre-sintering deposition diameter, sintering formula, and sintering diameter in the sintering production line workshop, and perform cleaning treatment according to the sintering time of the bright rod.

[0009] S2: Based on the extracted categorical and numerical features, the fused sintering features are obtained;

[0010] S3: After processing the sintering characteristics obtained in S2 with the difference between adjacent samples under time series, the Person analysis method is used to obtain the difference sintering characteristics, which are used as the model input parameters for predicting the difference sintering diameter;

[0011] S4: Based on the characteristics of differential sintering, establish a differential sintering diameter prediction model to predict the differential sintering diameter;

[0012] S5: Based on the difference sintering diameter prediction model and genetic algorithm network of S4, a parameter intelligent recommendation algorithm is constructed to input the recommended parameters into the sintering production line.

[0013] In one possible implementation of this application, S1 includes:

[0014] S11: Obtain the sintering diameter of the optical fiber preform, sort it by time, and filter out abnormal diameter data and empty data to obtain the sintering diameter after cleaning.

[0015] S12: Obtain the pre-sintering deposition diameter of the optical fiber preform, filter out abnormal and empty data, and calculate the sample intersection according to the sample number of the optical fiber preform and the sintering diameter in S11 to obtain the pre-sintering deposition diameter and sintering diameter.

[0016] S13: Based on the sintering formula under the adjustment of sintering cycle, according to the intersection method of sample numbering in S12, a common sample set of pre-sintering deposition diameter, sintering diameter and sintering formula is obtained.

[0017] In one possible implementation of this application, S1 further includes:

[0018] S14: When the axial coordinates of the observation data of the sintering diameter and the pre-sintering deposition diameter are inconsistent, the pre-sintering deposition diameter and the sintering diameter are aligned by a one-dimensional interpolation fitting method, and merged into a unified feature table with the extracted sintering formula features; wherein, the sintering formula features include at least one or more of the following: design outer diameter, core-to-core ratio of mandrel, mandrel diameter, deposition rate of a single torch, range of loose body outer diameter, deposition equipment type, sintering equipment type, loose body density, and sintering six temperature zones.

[0019] In one possible implementation of this application, S2 includes:

[0020] S21: Extract numerical features with significant correlations through Person correlation, including any one or more of the following: design outer diameter, loose body density, single torch deposition rate, loose body outer diameter range, sintering temperature in six temperature zones, and deposition diameter.

[0021] S22: Analyze the variance relationship between sintering equipment type, deposition equipment type and sintering diameter using analysis of variance to determine the degree of influence of sintering equipment type and deposition equipment type on sintering diameter;

[0022] S23: Obtain numerical and categorical features related to the sintering diameter, and analyze the sintering features after fusion.

[0023] In one possible implementation of this application, the sintering characteristics in S23 include any one or more of the following: design outer diameter, bulk density, single torch deposition rate, bulk outer diameter range, sintering six-zone temperature, deposition diameter, sintering equipment type, and deposition equipment type.

[0024] In one possible implementation of this application, the differential sintering features in S3 include any one or more of the differential design outer diameter, the differential sintering six-zone temperature, and the differential deposition diameter.

[0025] In one possible implementation of this application, the process of establishing the differential sintering diameter prediction model in S4 includes:

[0026] S41: Difference sintering features obtained from S3 are divided into training set, validation set and test set according to a ratio of 8:1:1. The training set and validation set are used as model training and accuracy verification data.

[0027] S42: XG Boost is used as the fitting regression model for the differential sintering diameter, and the root mean square error is used as the evaluation index to train and obtain the differential sintering diameter prediction model.

[0028] In one possible implementation of this application, the process of establishing the differential sintering diameter prediction model in S4 further includes:

[0029] S43: Define the method for calculating overlap: Where mse is the mean square error between the predicted difference sintering diameter and the actual difference sintering diameter, and the calculation method is as follows: Where, x i Let x be the predicted difference sintering diameter at the i-th point along the axial direction. gt The actual difference in sintering diameter between corresponding points along the axis;

[0030] S44: Save the model that meets the overlap index as the differential sintering diameter prediction model.

[0031] In one possible implementation of this application, S5 includes:

[0032] S51: A parameter intelligent recommendation algorithm is built using a genetic mutation algorithm to establish the recommendation logic for temperature parameters in the six sintering zones. The fitness index is set as the reciprocal of the mean square error between the predicted difference sintering diameter and the target difference sintering diameter, i.e.: Where mse is the mean square error between the difference sintering diameter predicted by the model in S44 and the actual difference sintering diameter, the objective constraint is set to minimize mse: min{mse(Dd tg )}, where d tg The target value for the differential sintering diameter;

[0033] S52: Initial population generation, generating 100 initial temperature combinations based on historical best process parameters;

[0034] S53: Iterative optimization, using the GAN algorithm to perform 50 generations of evolution, outputting the optimal temperature combination {T1,T2,...,T6};

[0035] S54: Simulation verification, calculation and prediction of the difference sintering diameter uniformity index: σ ≥ 85% is required; otherwise, return to S53 for further iterative optimization.

[0036] S55: Parameter distribution, which transmits optimized parameters to the sintering furnace control system in real time through the service interface.

[0037] Secondly, this application also provides a device for optimizing the sintering process of optical fiber preforms, the device comprising:

[0038] The acquisition module acquires the pre-sintering deposition diameter, sintering formula, and sintering diameter in the sintering production line workshop, and performs cleaning treatment according to the sintering time of the polished rod.

[0039] The analysis module obtains the fused sintering characteristics based on the extracted categorical and numerical features.

[0040] The processing module performs time series difference processing on the sintering features obtained from the analysis module, and then uses the Person analysis method to obtain the difference sintering features, which are used as the model input parameters for predicting the difference sintering diameter.

[0041] The module is constructed to establish a differential sintering diameter prediction model based on the differential sintering characteristics, so as to predict the differential sintering diameter;

[0042] The construction module, based on the differential sintering diameter prediction model and genetic algorithm network, constructs a parameter intelligent recommendation algorithm to input the recommended parameters into the sintering production line.

[0043] The sintering process optimization method and apparatus for optical fiber preforms provided in this application have the following beneficial effects:

[0044] This paper focuses on improving the efficiency of sintering parameter recommendation and the stability of optical fiber preform diameter. By analyzing the characteristics and difference features of the optical fiber preform, and scientifically selecting difference features based on correlation and variance analysis, a highly accurate optical fiber preform diameter difference prediction model is constructed. The XG Boost machine learning algorithm is used to deeply mine the potential relationships between data features, achieving accurate prediction of the optical fiber preform diameter. A genetic algorithm is employed to intelligently recommend temperature parameters for the sintering process, efficiently exploring the optimal parameter combination by simulating a natural evolutionary process. This application provides a systematic solution for optimizing optical fiber preform sintering process parameters, effectively improving the optimization efficiency of the sintering process and significantly enhancing the stability of the sintered diameter of the optical fiber preform. It has significant application value and technological advantages in the field of optical fiber manufacturing. Simultaneously, through the deep integration of data-driven and intelligent algorithms, it fills the technological gap in accurate prediction and dynamic optimization of existing optical fiber preform sintering processes, providing core technological support for the large-scale production of high-specification optical fiber preforms.

[0045] In summary, this application can accurately simulate the sintering results of optical fiber preforms, quickly and efficiently recommend sintering formula temperature parameters, improve the efficiency and accuracy of optical fiber production, and thus reduce the cost of manpower and material resources. Attached Figure Description

[0046] To more clearly illustrate the technical solutions in this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort. In the drawings:

[0047] Figure 1 A flowchart of an optimized sintering process for optical fiber preforms provided in this application;

[0048] Figure 2 The relevant flowchart of Embodiment 1 provided in this application;

[0049] Figure 3 Another related flowchart of Embodiment 1 provided in this application;

[0050] Figure 4 A schematic diagram of a sintering process optimization device for optical fiber preforms provided in this application. Detailed Implementation

[0051] To enable those skilled in the art to better understand the technical solutions in this application, the technical solutions in this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments of this specification, all other embodiments obtained by those skilled in the art without creative effort should fall within the scope of protection of this application.

[0052] This application discloses a method and apparatus for optimizing the sintering process of optical fiber preforms, focusing on improving the efficiency of sintering parameter recommendation and the stability of the preform diameter. By analyzing the preform characteristics and difference characteristics, and scientifically screening the difference characteristics based on correlation and variance analysis, a highly accurate difference preform diameter prediction model is constructed. The XG Boost machine learning algorithm is used to deeply mine the potential relationships between data features, achieving accurate prediction of the preform diameter. A genetic algorithm is employed to intelligently recommend temperature parameters for the sintering process, efficiently exploring the optimal parameter combination by simulating a natural evolutionary process. This application provides a systematic solution for optimizing optical fiber preform sintering process parameters, effectively improving process optimization efficiency and significantly enhancing the stability of the sintered preform diameter, possessing significant application value and technological advantages in the field of optical fiber manufacturing.

[0053] The following is a detailed description with reference to the accompanying drawings.

[0054] Figure 1 A flowchart of an optimized sintering process for optical fiber preforms provided in this application is shown below. Figure 1 As shown, the sintering process optimization method in this application includes at least the following steps:

[0055] S1: Obtain the pre-sintering deposition diameter, sintering formula, and sintering diameter in the sintering production line workshop, and perform cleaning treatment according to the sintering time of the bright rod.

[0056] S2: Based on the extracted categorical and numerical features, the fused sintering features are obtained;

[0057] S3: After processing the sintering characteristics obtained in S2 with the difference between adjacent samples under time series, the Person analysis method is used to obtain the difference sintering characteristics, which are used as the model input parameters for predicting the difference sintering diameter;

[0058] S4: Based on the characteristics of differential sintering, establish a differential sintering diameter prediction model to predict the differential sintering diameter;

[0059] S5: Based on the difference sintering diameter prediction model and genetic algorithm network of S4, a parameter intelligent recommendation algorithm is constructed to input the recommended parameters into the sintering production line.

[0060] Example 1

[0061] Combination Figure 2-3 A more detailed explanation of S1-S5 will be provided below. Specifically:

[0062] S1: Obtain data on the pre-sintering deposition diameter, sintering formula, and sintering diameter in the sintering production line workshop, and perform cleaning treatment according to the sintering time of the bright rod.

[0063] The data acquisition and cleaning process in S1 is as follows:

[0064] S11: Obtain the sintering diameter data of the optical fiber preform, sort it by time, filter out abnormal diameter data and empty data, and obtain the sintering diameter data of the preform after cleaning.

[0065] S12: Obtain the deposition diameter data of the optical fiber preform before sintering, filter out abnormal data and empty data, and calculate the sample intersection according to the optical fiber preform sample number in S11 to obtain the deposition diameter and sintering diameter before and after sintering.

[0066] S13: Based on the sintering formula data under sintering cycle adjustment, obtain the common sample set of sintering diameter, deposition diameter and sintering formula according to the sample number intersection method in S12.

[0067] To address the inconsistency in axial coordinates between the observed sintering diameter and deposition diameter, a one-dimensional interpolation fitting method was used to align the deposition diameter and sintering diameter, and then merge them into a unified feature table along with the sintering formulation characteristics. These sintering formulation characteristics include the designed outer diameter, core-to-core ratio, core diameter, deposition rate per torch, range of loose bulk outer diameter, deposition equipment type, sintering equipment type, loose bulk density, and the temperatures of the six sintering zones.

[0068] S2: Analyze the fused sintering characteristics based on categorical and numerical features. First, extract significantly correlated numerical features using Person correlation, including design outer diameter, bulk density, single torch deposition rate, bulk outer diameter range, sintering six-zone temperature, and deposition diameter. Then, analyze the variance relationship between sintering equipment type, deposition equipment type, and sintering diameter using analysis of variance methods, such as the ANOVA algorithm. The results show that sintering equipment type and deposition equipment type have a significant impact on the sintering diameter. Therefore, obtain the sintering numerical features significantly correlated with the sintering diameter and the categorical features with significant influence, namely: design outer diameter, bulk density, single torch deposition rate, bulk outer diameter range, sintering six-zone temperature, deposition diameter, sintering equipment type, and deposition equipment type.

[0069] Specifically, the sintering characteristics of optical fiber preforms include Person correlation, deposition diameter under ANVOA variance analysis, and sintering formulation, namely: design outer diameter, bulk density, single torch deposition rate, bulk outer diameter range, sintering six-zone temperature, deposition diameter, sintering equipment type, and deposition equipment type. The target feature for analysis is the sintering diameter.

[0070] S3: Based on the sintering numerical characteristics obtained in S2, including: design outer diameter, loose body density, single torch deposition rate, loose body outer diameter range, sintering six-zone temperature, deposition diameter, and sintering diameter (predicted target), the differences between adjacent samples in the time series are processed. Then, the Person analysis method is used to obtain significantly correlated difference sintering characteristics, namely difference design outer diameter, difference sintering six-zone temperature, and difference deposition diameter, which are used as model input parameters for predicting the difference sintering diameter.

[0071] It should be noted that, due to the adjacent difference processing, the sintering equipment type and the deposition equipment type are consistent before and after, so this type feature is eliminated.

[0072] Among them, the aforementioned optical fiber preform differential sintering characteristics include differential deposition diameter and differential sintering formula under Person correlation analysis, namely: differential design outer diameter, differential sintering six-temperature zone temperature, and differential deposition diameter, with the differential sintering diameter being the target feature of the analysis.

[0073] S4: Based on the characteristics of differential sintering, a differential sintering diameter prediction model is established to predict the differential sintering diameter using the differential design outer diameter, the six-zone temperature of differential sintering, and the differential deposition diameter. Simultaneously, to verify the prediction accuracy, a curve overlap index is constructed to evaluate the prediction accuracy of the differential sintering diameter. Finally, the model that meets the accuracy requirements is saved.

[0074] The process of constructing and validating the sintering difference diameter prediction model in S4 is as follows:

[0075] S41: Difference design outer diameter, difference sintering six-zone temperature, difference deposition diameter and difference sintering diameter data obtained from S3 are divided into training set, validation set and test set according to the ratio of 8:1:1. The training set and validation set are used as model training and accuracy verification data.

[0076] S42: The XG Boost algorithm is used as the fitting regression model for the differential sintering diameter, and the root mean square error (RMSE) is used as the evaluation index to train the differential sintering diameter prediction model.

[0077] S43: To facilitate the evaluation of the accuracy of model predictions, a method for calculating overlap is defined: Where mse is the mean square error between the predicted difference sintering diameter and the actual difference sintering diameter, and its calculation method is as follows: Where xi Let x be the predicted difference sintering diameter at the i-th point along the axial direction. gt The actual difference in sintered diameter between corresponding points along the axis is considered. When the overlap is ≥90%, the model is considered a prediction model with a good fit.

[0078] S44: Save the prediction model that satisfies the curve overlap index as the fitting model for predicting the difference sintering diameter, that is, obtain the difference sintering diameter prediction model.

[0079] S5: Based on the optical fiber preform differential prediction model and genetic algorithm network, such as using GAN algorithm, construct a parameter intelligent recommendation algorithm. By designing the differential outer diameter, differential deposition diameter, and differential sintering diameter, and based on the sintering six-temperature zone temperature of the sample at the previous moment, recommend the sintering six-temperature zone temperature of the current sample. Then, based on the sintering diameter of the sample at the previous moment, calculate the uniformity index of the predicted sintering diameter at the recommended six-temperature zone temperature at the current moment. After meeting the accuracy requirements, input the parameters into the sintering production line.

[0080] The steps for intelligent recommendation in S5 are as follows:

[0081] S51: Recommendation algorithm construction, using the genetic mutation algorithm, also known as the GAN algorithm, to build the recommendation logic for the temperature parameters of the six sintering zones. The fitness index is set as the reciprocal of the mean square error between the predicted difference sintering diameter and the target difference sintering diameter. Where mse is the mean square error between the difference sintering diameter predicted by the model in S44 and the actual difference sintering diameter. The objective constraint is set to minimize mse, i.e., min{mse(Dd)} tg )}, where d tg This represents the target value for the differential sintering diameter.

[0082] S52: Initial population generation, generating 100 initial temperature combinations based on historically optimal process parameters.

[0083] S53: Iterative optimization, using the GAN algorithm to perform 50 generations of evolution, outputting the optimal temperature combination {T1,T2,...,T6}.

[0084] S54: Simulation verification, calculation of predicted diameter uniformity index: σ ≥ 85% is required; otherwise, return to S53 to continue iterative optimization.

[0085] S55: Parameter distribution, which transmits optimized parameters to the sintering furnace control system in real time through the service interface.

[0086] The above-mentioned solution can accurately simulate the sintering results of optical fiber preforms, quickly and efficiently recommend sintering formula temperature parameters, improve the efficiency and accuracy of optical fiber production, and thus reduce the cost of manpower and material resources.

[0087] Based on the same inventive concept, this application also provides a device for optimizing the sintering process of optical fiber preforms, the structure of which is as follows: Figure 4 As shown.

[0088] Figure 4 A schematic diagram of a sintering process optimization device for optical fiber preforms provided in this application. Figure 4 As shown, the optical fiber preform sintering process optimization device 400 in this application specifically includes:

[0089] The module 401 acquires the pre-sintering deposition diameter, sintering formula, and sintering diameter in the sintering production line workshop, and performs cleaning treatment according to the sintering time of the polished rod.

[0090] Analysis module 402 obtains the fused sintering characteristics based on the extracted category features and numerical features;

[0091] The processing module 403 processes the sintering features obtained from the analysis module by performing time series adjacent sample difference processing, and then uses the Person analysis method to obtain the difference sintering features, which are used as model input parameters for predicting the difference sintering diameter.

[0092] Module 404 is constructed to establish a differential sintering diameter prediction model based on the differential sintering characteristics in order to predict the differential sintering diameter.

[0093] The construction module 404 constructs a parameter intelligent recommendation algorithm based on the differential sintering diameter prediction model and genetic algorithm network of the construction module, so as to input the recommended parameters into the sintering production line.

[0094] The various embodiments in this application are described in a progressive manner. Similar or identical parts between embodiments can be referred to mutually. Each embodiment focuses on describing the differences from other embodiments. In particular, the device embodiments are basically similar to the method embodiments, so the description is relatively simple; relevant parts can be referred to the descriptions of the method embodiments.

[0095] The equipment and method provided in this application are one-to-one correspondences. Therefore, the equipment also has similar beneficial technical effects as its corresponding method. Since the beneficial technical effects of the method have been described in detail above, the beneficial technical effects of the equipment will not be repeated here.

[0096] Those skilled in the art will understand that embodiments of this application can be provided as methods, apparatus, systems, or computer program products. Therefore, this application can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, this application can take the form of a computer program product implemented on one or more computer-usable storage media containing computer-usable program code.

[0097] It should also be noted that the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus. Unless otherwise specified, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes that element.

[0098] The above are merely embodiments of this application and are not intended to limit the scope of this application. Various modifications and variations can be made to this application by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the protection scope of this application.

Claims

1. A method for optimizing the sintering process of optical fiber preforms, characterized in that, The method includes: S1: Obtain the pre-sintering deposition diameter, sintering formula, and sintering diameter in the sintering production line workshop, and perform cleaning treatment according to the sintering time of the bright rod. S2: Based on the extracted categorical and numerical features, the fused sintering features are obtained; S3: After processing the sintering characteristics obtained in S2 by the difference between adjacent samples under time series, the Pearson analysis method is used to obtain the difference sintering characteristics, which are used as the model input parameters for predicting the difference sintering diameter; the difference sintering characteristics include the difference design outer diameter, the difference sintering six temperature zones temperature, and the difference deposition diameter; S4: Based on the characteristics of differential sintering, establish a differential sintering diameter prediction model to predict the differential sintering diameter; S5: Based on the differential sintering diameter prediction model and genetic algorithm of S4, construct a parameter intelligent recommendation algorithm to input the recommended parameters into the sintering production line. This includes: using differential design outer diameter, differential deposition diameter, and differential sintering diameter, recommending the temperature of the current sample's sintering six-temperature zone based on the temperature of the previous sample's sintering six-temperature zone, and then calculating the uniformity index of the predicted sintering diameter at the current time's recommended six-temperature zone temperature based on the sintering diameter of the previous sample. After meeting the accuracy requirements, input the parameters into the sintering production line. S5 includes: S51: A parameter intelligent recommendation algorithm is built using a genetic mutation algorithm to establish the recommendation logic for temperature parameters in the six sintering zones. The fitness index is set as the reciprocal of the mean square error between the predicted difference sintering diameter and the target difference sintering diameter, i.e.: Where mse is the mean square error between the predicted difference sintering diameter and the actual difference sintering diameter, the objective constraint is to minimize mse: ,in The target value for the differential sintering diameter; S52: Initial population generation, generating 100 initial temperature combinations based on historical best process parameters; S53: Iterative optimization, performing 50 generations of evolution through the GA algorithm, outputting the optimal temperature combination {T1, T2, ..., T6}; S54: Simulation verification, calculation and prediction of the difference sintering diameter uniformity index: ,Require Otherwise, return to S53 to continue iterative optimization; S55: Parameter distribution, which transmits optimized parameters to the sintering furnace control system in real time through the service interface.

2. The method for optimizing the sintering process of an optical fiber preform according to claim 1, characterized in that, S1 includes: S11: Obtain the sintering diameter of the optical fiber preform, sort it by time, and filter out abnormal diameter data and empty data to obtain the sintering diameter after cleaning. S12: Obtain the pre-sintering deposition diameter of the optical fiber preform, filter out abnormal and empty data, and calculate the sample intersection according to the sample number of the optical fiber preform and the sintering diameter in S11 to obtain the pre-sintering deposition diameter and sintering diameter. S13: Based on the sintering formula under the adjustment of sintering cycle, according to the intersection method of sample numbering in S12, a common sample set of pre-sintering deposition diameter, sintering diameter and sintering formula is obtained.

3. The method for optimizing the sintering process of an optical fiber preform according to claim 2, characterized in that, S1 further includes: S14: When the axial coordinates of the observation data of the sintering diameter and the pre-sintering deposition diameter are inconsistent, the pre-sintering deposition diameter and the sintering diameter are aligned by a one-dimensional interpolation fitting method, and merged into a unified feature table with the extracted sintering formula features; wherein, the sintering formula features include the designed outer diameter, core-to-core ratio of the mandrel, mandrel diameter, deposition rate of a single torch, range of loose body outer diameter, deposition equipment type, sintering equipment type, loose body density, and sintering six-zone temperature.

4. The method for optimizing the sintering process of an optical fiber preform according to claim 1, characterized in that, S2 includes: S21: Extract numerical features with significant correlations through Pearson correlation, including any one or more of the following: design outer diameter, loose body density, single torch deposition rate, loose body outer diameter range, sintering temperature in six temperature zones, and deposition diameter. S22: Analyze the variance relationship between sintering equipment type, deposition equipment type and sintering diameter using analysis of variance to determine the degree of influence of sintering equipment type and deposition equipment type on sintering diameter; S23: Obtain numerical and categorical features related to the sintering diameter, and analyze the sintering features after fusion.

5. The method for optimizing the sintering process of an optical fiber preform according to claim 4, characterized in that, The sintering characteristics in S23 include the design outer diameter, bulk density, single torch deposition rate, bulk outer diameter range, sintering six-zone temperature, deposition diameter, sintering equipment type, and deposition equipment type.

6. The method for optimizing the sintering process of an optical fiber preform according to claim 1, characterized in that, The process of establishing the differential sintering diameter prediction model in S4 includes: S41: Difference sintering features obtained from S3 are divided into training set, validation set and test set according to a ratio of 8:1:

1. The training set and validation set are used as model training and accuracy verification data. S42: XG Boost is used as the fitting regression model for the differential sintering diameter, and the root mean square error is used as the evaluation index to train and obtain the differential sintering diameter prediction model.

7. The method for optimizing the sintering process of an optical fiber preform according to claim 6, characterized in that, The process of establishing the differential sintering diameter prediction model in S4 also includes: S43: Define the method for calculating overlap: Where mse is the mean square error between the predicted difference sintering diameter and the actual difference sintering diameter, and is calculated as follows: ,in, The sintering diameter is the predicted difference at the i-th point along the axial direction. The actual difference in sintering diameter between corresponding points along the axis; S44: Save the model that meets the overlap index as the differential sintering diameter prediction model.

8. A device for optimizing the sintering process of optical fiber preforms, characterized in that, The device includes: The acquisition module acquires the pre-sintering deposition diameter, sintering formula, and sintering diameter in the sintering production line workshop, and performs cleaning treatment according to the sintering time of the polished rod. The analysis module obtains the fused sintering characteristics based on the extracted categorical and numerical features. The processing module performs time-series difference processing on the sintering features obtained from the analysis module, and then uses the Pearson analysis method to obtain the difference sintering features, which are used as model input parameters for predicting the difference sintering diameter. The difference sintering features include the difference design outer diameter, the difference sintering six-zone temperature, and the difference deposition diameter. The module is constructed to establish a differential sintering diameter prediction model based on the differential sintering characteristics, so as to predict the differential sintering diameter; The construction module, based on the differential sintering diameter prediction model and genetic algorithm, constructs a parameter intelligent recommendation algorithm to input the recommended parameters into the sintering production line. This includes: using differential design outer diameter, differential deposition diameter, and differential sintering diameter, recommending the temperature of the current sample's sintering six-temperature zone based on the temperature of the previous sample's sintering six-temperature zone, and then calculating the uniformity index of the predicted sintering diameter at the recommended six-temperature zone temperature at the current moment based on the sintering diameter of the sample at the previous moment. After meeting the accuracy requirements, the parameters are input into the sintering production line. The building module includes: A parameter intelligent recommendation algorithm was built, employing a genetic mutation algorithm to establish the recommendation logic for temperature parameters in the six sintering zones. The fitness index was set as the reciprocal of the mean square error between the predicted difference sintering diameter and the target difference sintering diameter, i.e.: Where mse is the mean square error between the predicted difference sintering diameter and the actual difference sintering diameter, the objective constraint is to minimize mse: ,in The target value for the differential sintering diameter; Initial population generation: 100 initial temperature combinations are generated based on historically optimal process parameters; Iterative optimization is performed using the GA algorithm for 50 generations of evolution, outputting the optimal temperature combination {T1, T2, ..., T6}. Simulation verification and calculation of the predicted difference sintering diameter uniformity index: ,Require Otherwise, continue iterative optimization; The parameters are sent out, and the optimized parameters are transmitted to the sintering furnace control system in real time through the service interface.

Images