Parameter adjusting method and device in material growth process, equipment and medium

Abstract

The invention provides a parameter adjustment method and device in a material growth process, equipment and a medium, and relates to the technical field of material preparation. Comprising the following steps: in a material growth process of a target material, in a first preset duration, according to a growth stage of the target material and material information of the target material, pre-processing a first growth parameter data set acquired each time according to an acquired data acquisition frequency corresponding to the growth stage; inputting the obtained second growth parameter data set into the target neural network model for analysis and processing, and obtaining an output result; and according to the adjustment priority of each piece of growth parameter data and at least one output result within the first preset duration, adjusting a target parameter in the first growth parameter data set, and keeping the target parameter as the adjusted parameter within the second preset duration. According to the method provided by the embodiment of the invention, the flexibility and real-time performance of parameter adjustment can be improved, and the control precision of material growth is improved.

Description

Technical Field

[0001] The present application relates to the technical field of material preparation, and in particular to a method, device, equipment and medium for adjusting parameters in a material growth process. Background Art

[0002] In related technologies, during the material growth process, the control of parameters can rely on a single parameter adjustment framework. Assuming that the influence of a single parameter (such as temperature, airflow, etc.) on the growth process is the dominant factor, the parameter can be adjusted to control the material growth. However, this method lacks real-time and flexibility, and it is necessary to wait until the material preparation is completed to determine whether the adjustment is accurate. Summary of the Invention

[0003] This application provides a method, device, equipment, and medium for adjusting parameters during material growth. The technical solution of this application is as follows:

[0004] In a first aspect, the present application provides a method for adjusting parameters during a material growth process, comprising:

[0005] During the material growth process of the target material, obtaining a growth stage of the target material, and obtaining a data acquisition frequency corresponding to the growth stage according to the growth stage and material information of the target material;

[0006] within a first preset time period, preprocessing the first growth parameter data set collected each time according to the data collection frequency to obtain a second growth parameter data set, wherein the first growth parameter data set includes at least one growth parameter data;

[0007] Inputting the second growth parameter data set into a target neural network model for analysis and processing to obtain an output result, wherein the output result includes information of each category in at least one category and information of each subcategory under the information of each category;

[0008] acquiring, according to at least one of a material type of the target material and a priority setting instruction, an adjustment priority of each growth parameter data in the first growth parameter data set;

[0009] determining a target parameter in the first growth parameter data set according to the adjustment priority and at least one output result within the first preset time period;

[0010] Based on at least one output result within the first preset time length, the target parameter in the first growth parameter data set is adjusted, and the target parameter is maintained as the adjusted parameter within a second preset time length, wherein the second preset time length is adjacent to the first preset time length and away from the material growth start time point of the target material.

[0011] In a possible implementation, wherein the first growth parameter data set includes a first image data subset and a first non-image data subset, preprocessing the first growth parameter data set collected each time to obtain the second growth parameter data set includes:

[0012] performing size adjustment and normalization processing on each image data in the first image data subset to obtain a second image data subset;

[0013] performing normalization processing on each image data in the second image data subset to obtain a third image data subset;

[0014] Processing the third image data subset using an image enhancement technique to obtain a fourth image data subset;

[0015] performing standardization and normalization processing on the first non-image data subset to obtain a second non-image data subset;

[0016] performing missing value filling on the second non-image data subset to obtain a third non-image data subset;

[0017] adding the fourth subset of image data and the third subset of non-image data to a second set of growth parameter data;

[0018] The second growth parameter data set is subjected to feature extraction processing, feature conversion processing, and feature splicing processing to obtain a processed second growth parameter data set.

[0019] In a possible implementation, adjusting the target parameter in the first growth parameter data set according to at least one output result within the first preset time period includes:

[0020] Analyzing and processing at least one output result within the first preset time period to obtain a hierarchical classification result, wherein the hierarchical classification result includes each category information and each subcategory information under each category information;

[0021] Performing statistics on the hierarchical classification results to obtain statistical results, wherein the statistical results include a first quantity of each category of information and a second quantity of each subcategory of information under each category of information;

[0022] determining a target parameter in the first growth parameter data set according to the first quantity, and determining an adjustment strategy for the target parameter according to the second quantity;

[0023] The target parameter is adjusted using the adjustment strategy.

[0024] In one possible implementation, the method further includes:

[0025] Performing focused ion beam cutting on the first historical material to obtain structural performance information and characteristic change information of each layer corresponding to the first historical material;

[0026] Training an initial neural network model based on the structural performance information and characteristic change information of each layer corresponding to the first historical material, and obtaining output information of the initial neural network model;

[0027] Adjusting the loss function corresponding to the initial neural network model according to the output information and the label information of the first historical material to obtain an adjusted loss function;

[0028] The initial neural network model is adjusted according to the adjusted loss function, and the adjusted initial neural network model is continued to be trained using the structural performance information and characteristic change information of each layer corresponding to the second historical material, until the training information corresponding to the initial neural network model meets the model training requirements, and it is determined that the target neural network model is obtained.

[0029] In one possible implementation, the method further includes:

[0030] Obtaining growth status information corresponding to the target material;

[0031] An adjustment strategy corresponding to the target material is determined according to the growth status information and at least one output result within the second preset time period.

[0032] In a possible implementation, preprocessing the first growth parameter data set collected each time to obtain the second growth parameter data set includes:

[0033] When the acquisition devices corresponding to each first growth parameter data in the first growth parameter data set are different, each growth parameter data in the first growth parameter data set is processed according to the acquisition time point of each first growth parameter data to obtain the second growth parameter data set.

[0034] In a second aspect, the present application provides a device for adjusting parameters during a material growth process, comprising:

[0035] a frequency acquisition unit, configured to acquire a growth stage of the target material during its growth process, and acquire a data acquisition frequency corresponding to the growth stage based on the growth stage and material information of the target material;

[0036] a data processing unit, configured to pre-process the first growth parameter data set collected each time according to the data collection frequency within a first preset time period to obtain a second growth parameter data set, wherein the first growth parameter data set includes at least one growth parameter data;

[0037] a data analysis unit, configured to input the second growth parameter data set into a target neural network model for analysis and processing, and obtain an output result, wherein the output result includes information of each category in at least one category and information of each subcategory under the information of each category;

[0038] a parameter adjustment unit, configured to obtain an adjustment priority of each growth parameter data in the first growth parameter data set according to at least one of a material type of the target material and a priority setting instruction;

[0039] The parameter adjustment unit is further configured to determine a target parameter in the first growth parameter data set according to the adjustment priority and at least one output result within the first preset time period;

[0040] The parameter adjustment unit is further used to adjust the target parameters in the first growth parameter data set based on at least one output result within the first preset time length, and maintain the target parameters as the adjusted parameters within a second preset time length, wherein the second preset time length is adjacent to the first preset time length and away from the material growth start time point of the target material.

[0041] In a third aspect, the present application provides an electronic device, comprising:

[0042] processor;

[0043] a memory for storing instructions executable by the processor;

[0044] The processor is configured to execute the instructions to implement the parameter adjustment method during the material growth process described in the first aspect.

[0045] In a fourth aspect, the present application provides a computer-readable storage medium having a computer program stored thereon, characterized in that when the computer program is executed by a processor, the parameter adjustment method in the material growth process described in the first aspect is implemented.

[0046] In a fifth aspect, the present application provides a computer program product, comprising a computer program / instruction, characterized in that when the computer program / instruction is executed by a processor, the parameter adjustment method in the material growth process described in the first aspect is implemented.

[0047] The technical solution disclosed in this application brings at least the following beneficial effects:

[0048] In an embodiment of the present application, during a material growth process of a target material, a growth stage of the target material is obtained, and a data collection frequency corresponding to the growth stage is obtained based on the growth stage and material information of the target material; within a first preset time period, a first growth parameter data set collected each time is preprocessed based on the data collection frequency to obtain a second growth parameter data set, wherein the first growth parameter data set includes at least one growth parameter data; the second growth parameter data set is input into a target neural network model for analysis and processing to obtain an output result, wherein the output result includes information of each category in at least one category and information of each subcategory under each category information; based on at least one of the material type of the target material and a priority setting instruction, an adjustment priority of each growth parameter data in the first growth parameter data set is obtained; based on the adjustment priority and at least one output result within the first preset time period, a target parameter in the first growth parameter data set is determined; based on the at least one output result within the first preset time period, the target parameter in the first growth parameter data set is adjusted, and the target parameter is maintained at the adjusted parameter for a second preset time period, wherein the second preset time period is a time period adjacent to the first preset time period and away from the material growth start time point of the target material. In this way, multiple parameters in the material growth process can be adjusted, and there is no need to wait for the material to be fully prepared to determine whether a certain parameter is adjusted accurately. Multiple parameters can be intelligently adjusted according to the parameter adjustment priority, and the parameter adjustment information within a preset time period can be determined based on the material preparation information within a preset time period. Based on multi-dimensional data input, multiple growth parameters that need to be adjusted can be output. By analyzing multiple output results, the final adjustment method can be determined, and refined real-time adjustment can be performed during the material growth process. It can be applicable to the growth process control of various materials, improve the flexibility and real-time performance of parameter adjustment, and improve the control accuracy, growth efficiency and stability of the material growth process.

[0049] It should be understood that the foregoing general description and the following detailed description are exemplary and explanatory only and are not restrictive of the present application. BRIEF DESCRIPTION OF THE DRAWINGS

[0050] The drawings herein are incorporated into and constitute a part of the specification, illustrate embodiments consistent with the present application, and together with the specification are used to explain the principles of the present application, and do not constitute an improper limitation on the present application.

[0051] Figure 1 A schematic flow chart of a method for adjusting parameters during a material growth process provided in an embodiment of the present application;

[0052] Figure 2A schematic diagram of a process flow of a parameter adjustment system during a material growth process provided in an embodiment of the present application;

[0053] Figure 3 This is a schematic diagram of an example of a parameter adjustment device during a material growth process provided in an embodiment of the present application;

[0054] Figure 4 This is a schematic diagram of the structure of an electronic device provided in an embodiment of the present application. DETAILED DESCRIPTION

[0055] In order to enable ordinary people in the art to better understand the technical solutions of the present application, the technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the accompanying drawings.

[0056] It should be noted that the terms "first," "second," and the like in the specification and claims of this application and the accompanying drawings are used to distinguish similar objects and are not necessarily used to describe a specific order or precedence. It should be understood that the terms used in this manner are interchangeable where appropriate so that the embodiments of the application described herein can be implemented in an order other than those illustrated or described herein. The implementations described in the following exemplary embodiments do not represent all implementations consistent with the present application. Instead, they are merely examples of apparatus and methods consistent with certain aspects of the present application, as detailed in the appended claims.

[0057] It should be noted that the information (including but not limited to user device information, user personal information, etc.), data (including but not limited to data used for analysis, stored data, displayed data, etc.) and signals involved in this application are all authorized by the user or fully authorized by all parties, and the collection, use and processing of relevant data must comply with the relevant laws, regulations and standards of relevant countries and regions.

[0058] The acquisition, storage, use, and processing of data in this application's technical solution comply with relevant national laws and regulations.

[0059] It should be noted that in the embodiments of the present application, there may be certain software, components, models, etc. that already exist in the industry. They should be considered as exemplary. Their purpose is only to illustrate the feasibility of implementing the technical solution of the present application, but it does not mean that the applicant has or will necessarily use the solution.

[0060] The following describes in detail the technical solutions provided by various embodiments of the present application in conjunction with the accompanying drawings.

[0061] Figure 1 This is a flow chart of a method for adjusting parameters during a material growth process provided in an embodiment of the present application. This method can be applied to scenarios where multiple parameters are controlled during a material preparation process, and includes the following steps:

[0062] S101, during a material growth process of a target material, obtaining a growth stage of the target material, and obtaining a data acquisition frequency corresponding to the growth stage based on the growth stage and material information of the target material;

[0063] According to some embodiments, the technical solutions of the embodiments of the present application may be implemented by, for example, an electronic device. This electronic device is not limited to a specific device. For example, the electronic device may be a computer, a smartphone, a wearable device, or the like. The implementation of the embodiments of the present application may be, for example, a device capable of data communication.

[0064] In some embodiments, a target material can be used to indicate a material being prepared. The target material is not a specific fixed material. For example, if the material type corresponding to the target material changes, the target material can also change accordingly. For example, if the material volume corresponding to the target material changes, the target material can also change accordingly.

[0065] According to some embodiments, a material growth process can be used to indicate the preparation process of a target material. For example, the name of the material growth process is not limited and can also be referred to as a material preparation process. The material growth process can be used to indicate the generation process from raw materials to the target material. The material generation process does not specifically refer to a fixed process. For example, when the target material changes, the material growth process can also change accordingly. For example, when the growth time corresponding to the material growth process changes, the material growth process can also change accordingly.

[0066] In some embodiments, the growth stage can be used to indicate different stages of the target material's growth process. Different growth stages can correspond to different material states, for example, and different growth stages can correspond to different growth parameter data. The growth stage does not specifically refer to a fixed stage. The growth stage can include, for example, an unfinished stage and a near-completed stage. The growth stage can be determined, for example, based on the growth duration or image data corresponding to the target material's growth process captured by a camera.

[0067] According to some embodiments, material information may include, for example, information regarding the preparation requirements for the target material and information regarding the materials required to prepare the target material. This material information is not fixed information. For example, if the type of the target material changes, the material information may also change accordingly. For example, if the material thickness in the material information changes, the material information may also change accordingly.

[0068] In some embodiments, the data acquisition frequency can be used to indicate the frequency at which data is collected during parameter control. This data acquisition frequency does not specifically refer to a fixed frequency. The data acquisition frequency can be determined, for example, based on material information of the target material, or based on a frequency setting instruction, although this is not limited in this embodiment of the present application. For example, different data acquisition frequencies can correspond to different growth stages of different materials.

[0069] In some embodiments, during the material growth process of the target material, the growth stage of the target material is obtained, and based on the growth stage and material information of the target material, the data acquisition frequency corresponding to the growth stage is obtained.

[0070] S102, preprocessing the first growth parameter data set collected each time according to the data collection frequency within a first preset time period to obtain a second growth parameter data set, wherein the first growth parameter data set includes at least one growth parameter data;

[0071] According to some embodiments, the preset duration can be used to indicate the duration for collecting growth parameter data, for example, growth parameter data can be collected according to the data collection frequency within the preset duration. The preset duration does not specifically refer to a fixed duration. For example, when a modification instruction for the preset duration is received, the preset duration can also change accordingly. For example, when the preset duration is determined based on the material type of the target material, the preset duration can also change accordingly if the material type changes.

[0072] In some embodiments, the first preset duration may, for example, indicate the duration of data collection before adjusting growth parameters. The first in the first preset duration is used to distinguish it from other preset durations and does not specifically refer to a fixed duration. For example, when the growth stage changes, the first preset duration may also change accordingly.

[0073] According to some embodiments, growth parameter data may be used to indicate parameters collected during the growth of a target material. The amount of data corresponding to the growth parameter data may be multiple. The growth parameter data is not specifically fixed data. For example, if the target material changes, the growth parameter data may also change accordingly. For example, if the acquisition equipment changes, the growth parameter data may also change accordingly.

[0074] In some embodiments, the first growth parameter data set may be, for example, a collection of at least one growth parameter data set. The first growth parameter data set may include, for example, at least one growth parameter data set collected within a first predetermined time period. For example, the first growth parameter data set may include multiple growth parameter data sets of different dimensions collected at multiple collection time points. The name of the growth parameter data set is not limited and may, for example, be referred to as in-situ characterization data.

[0075] In some embodiments, preprocessing may be, for example, processing performed on the first growth parameter data set before inputting it into the model analysis. This preprocessing does not necessarily refer to a fixed processing method. For example, if the growth parameter data included in the first growth parameter data set changes, the preprocessing may also change accordingly. For example, if the number of processing methods corresponding to the preprocessing changes, the preprocessing may also change accordingly.

[0076] According to some embodiments, the second growth parameter data set can be used to indicate a set obtained after preprocessing the first growth parameter data set. The "second" in the second growth parameter data set is used to distinguish it from the remaining growth parameter data sets. The second growth parameter data set does not specifically refer to a fixed set. For example, when the first growth parameter data set changes, the second growth parameter data set may also change accordingly.

[0077] According to some embodiments, within a first preset time period, the first growth parameter data set collected each time is preprocessed according to the data collection frequency to obtain a second growth parameter data set, wherein the first growth parameter data set includes at least one growth parameter data.

[0078] S103: Inputting the second growth parameter data set into the target neural network model for analysis and processing to obtain an output result, wherein the output result includes information of each category in at least one category and information of each subcategory under each category information;

[0079] In some embodiments, the target neural network model may be, for example, a model that has been trained and can analyze and process a growth parameter data set. The target neural network model does not specifically refer to a fixed model. For example, when the model type corresponding to the target neural network model changes, the target neural network model may also change accordingly. For example, when the model parameters of the target neural network model change, the target neural network model may also change accordingly. Among them, different target materials may correspond to different target neural network models, or may correspond to the same target neural network model, and the embodiments of the present application are not limited to this.

[0080] In some embodiments, the output result may be, for example, the output result of inputting a second growth parameter data set into a target neural network model. In other words, a single growth parameter data set collected may correspond to a single output result. This output result is not a fixed result. For example, when the growth parameter data set changes, the output result may also change accordingly. For example, when the target neural network model changes, the output result may also change accordingly.

[0081] In some embodiments, the at least one category information may be a category obtained by classifying at least one growth parameter data. For example, the at least one category information may include growth temperature and source furnace damper status. The subcategory information may be information obtained by subdividing the category information. For example, the subcategory information corresponding to the growth temperature may include "temperature increase," "temperature maintenance," and "temperature reduction."

[0082] In some embodiments, the second growth parameter data set is input into the target neural network model for analysis and processing to obtain an output result, wherein the output result includes category information in at least one category and subcategory information under each category information.

[0083] In some embodiments, for example, in-situ characterization data (e.g., a single image or a multi-dimensional array) is used as input, and a trained machine learning model analyzes and derives adjustment methods for multiple material growth-related parameters. Adjustment parameters include, but are not limited to, growth temperature, source furnace temperature, gas source flow rate, gas source ratio, pull speed, and pull rotation speed. Each parameter is output through a classification model, with adjustment methods such as increasing or decreasing the temperature, maintaining the temperature, increasing or decreasing the flow rate, etc., to address real-time changes during the material growth process.

[0084] S104, obtaining an adjustment priority of each growth parameter data in the first growth parameter data set according to at least one of a material type of the target material and a priority setting instruction;

[0085] According to some embodiments, the priority setting instruction can be used, for example, to determine the adjustment priority of each growth parameter data. The priority setting instruction is not specifically a fixed instruction. The priority setting instruction includes but is not limited to a voice priority setting instruction, a click priority setting instruction, a drag priority setting instruction, etc.

[0086] In some embodiments, the adjustment priority of each growth parameter data in the first growth parameter data set may be obtained based on at least one of a material type of the target material and a priority setting instruction.

[0087] For example, the adjustment priority of each growth parameter data in the first growth parameter data set may be obtained according to the material type of the target material. For example, the adjustment priority of each growth parameter data in the first growth parameter data set may be obtained according to a priority setting instruction.

[0088] S105, determining a target parameter in the first growth parameter data set according to the adjustment priority and at least one output result within the first preset time period;

[0089] According to some embodiments, the target parameter may be, for example, a parameter to be adjusted in the first growth parameter data set. The target parameter is not specifically a fixed parameter. For example, when the growth stage of the target material changes, the target parameter may also change accordingly. For example, when a setting instruction for the target parameter is received, the target parameter may also change accordingly.

[0090] S106. Adjust the target parameters in the first growth parameter data set based on at least one output result within the first preset time length, and maintain the target parameters as the adjusted parameters within the second preset time length, wherein the second preset time length is a time length adjacent to the first preset time length and away from the material growth start time point of the target material.

[0091] In some embodiments, the growth parameter data set collected each time can be input into the target neural network model to obtain an output result according to the data collection frequency, and at least one output result can be obtained within a first preset time period.

[0092] In some embodiments, the second preset time length is adjacent to the first preset time length and is far away from the material growth start time point of the target material, that is, the second preset time length is the next preset time length adjacent to the first preset time length.

[0093] According to some embodiments, the target parameters in the first growth parameter data set are adjusted based on at least one output result within a first preset time length, and the target parameters are maintained as adjusted parameters within a second preset time length, wherein the second preset time length is a time length adjacent to the first preset time length and away from the material growth start time point of the target material.

[0094] In an embodiment of the present application, during the material growth process of the target material, the growth stage of the target material is obtained, and based on the growth stage and the material information of the target material, the data acquisition frequency corresponding to the growth stage is obtained; within a first preset time length, the first growth parameter data set collected each time is preprocessed according to the data acquisition frequency to obtain a second growth parameter data set, wherein the first growth parameter data set includes at least one growth parameter data; the second growth parameter data set is input into the target neural network model for analysis and processing to obtain an output result, wherein the output result includes each category information in at least one category and each subcategory information under each category information; based on at least one output result within the first preset time length, the target parameter in the first growth parameter data set is adjusted, and the target parameter is maintained as the adjusted parameter within a second preset time length, wherein the second preset time length is a time length adjacent to the first preset time length and away from the material growth start time point of the target material. In this way, multiple parameters in the material growth process can be adjusted, and there is no need to wait for the material to be fully prepared to determine whether a certain parameter is adjusted accurately. Multiple parameters can be intelligently adjusted according to the parameter adjustment priority, and the parameter adjustment information within a preset time period can be determined based on the material preparation information within a preset time period. Based on multi-dimensional data input, multiple growth parameters that need to be adjusted can be output. By analyzing multiple output results, the final adjustment method can be determined, and refined real-time adjustment can be performed during the material growth process. It can be applicable to the growth process control of various materials, improve the flexibility and real-time performance of parameter adjustment, and improve the control accuracy, growth efficiency and stability of the material growth process.

[0095] Figure 2 A flow chart of a method for adjusting parameters during material growth provided in an embodiment of the present application includes the following steps:

[0096] S201, during the material growth process of the target material, obtaining a growth stage of the target material, and obtaining a data acquisition frequency corresponding to the growth stage according to the growth stage and material information of the target material;

[0097] The relevant process is as above and will not be repeated here.

[0098] In some embodiments, the technical solutions of the embodiments of the present application can be used in molecular beam epitaxy equipment, vapor deposition equipment, Czochralski ingot and other material preparation equipment.

[0099] In some embodiments, the data acquisition frequency can be set based on the different stages and needs of the growth process. For example, during critical stages of material growth, the data acquisition frequency is higher, but each acquisition can be maintained for a period of time to achieve more detailed data construction and finer-grained real-time control. For example, during the growth of thicker gallium arsenide (GaAs) and aluminum gallium arsenide (AlGaAs) materials, the growth duration is longer, so the data acquisition frequency can be lower. However, for extremely thin InAsInAs and InGaAsInGaAs materials, a higher data acquisition frequency is required.

[0100] S202, preprocessing the first growth parameter data set collected each time according to the data collection frequency within a first preset time period to obtain a second growth parameter data set, wherein the first growth parameter data set includes at least one growth parameter data;

[0101] The relevant process is as above and will not be repeated here.

[0102] In some embodiments, the method for collecting parameter data during the material growth process is not limited. For example, growth parameter data can be collected using sensors such as thermocouples and ion gauges. Growth parameter data can also be referred to as in-situ characterization data, which can be collected using a camera for real-time monitoring. For example, in-situ characterization data can include in-situ fluorescent screen patterns and cavity change data observed through a flange.

[0103] In some instances, when the technical solutions of the embodiments of the present application are applied to a molecular beam epitaxy system, the collection of growth parameter data involves multiple core variables, including but not limited to: using temperature sensors such as thermocouples or infrared pyrometers to monitor and control the material preparation growth temperature in real time within the growth region; using thermocouples to monitor and control the source furnace temperature in real time; and using baffle state controllers (such as proximity switches or displacement sensors) installed at each source furnace location to monitor and control whether the baffle is open or closed in real time. The first growth parameter data set may also include in-situ characterization data, where in-situ characterization data collection is performed by a reflection high-energy electron diffraction (RHEED) system or in conjunction with other in-situ characterization equipment. The RHEED system can acquire electron diffraction patterns on the material surface in real time, capturing microscopic changes in the material surface during growth. Changes in the RHEED signal, particularly during quantum dot (QD) growth, can reflect the material's nucleation status, crystal quality, and surface morphology. This real-time in-situ characterization technology can improve the accuracy of surface structure information acquired during material growth. The system is also equipped with an in-situ camera to monitor the material growth process in real time using image acquisition equipment.

[0104] According to some embodiments, wherein the first growth parameter data set includes a first image data subset and a first non-image data subset, preprocessing each acquired first growth parameter data set to obtain a second growth parameter data set includes:

[0105] performing size adjustment and normalization processing on each image data in the first image data subset to obtain a second image data subset;

[0106] performing normalization processing on each image data in the second image data subset to obtain a third image data subset;

[0107] Processing the third image data subset using an image enhancement technique to obtain a fourth image data subset;

[0108] performing standardization and normalization processing on the first non-image data subset to obtain a second non-image data subset;

[0109] Filling missing values on the second non-image data subset to obtain a third non-image data subset;

[0110] adding a fourth subset of image data and a third subset of non-image data to the second set of growth parameter data;

[0111] The second growth parameter data set is subjected to feature extraction, feature conversion, and feature splicing to obtain a processed second growth parameter data set. Therefore, classifying the growth parameter data and using different processing methods can improve the accuracy of data processing and the accuracy of parameter control.

[0112] In some embodiments, since the in-situ characterization data collected during the material growth process includes a large amount of noise, redundant information and missing values, the collected data is strictly preprocessed before being input into the machine learning model, i.e., the target neural network model, to ensure the quality of the data and improve the predictive ability of the model.

[0113] According to some embodiments, preprocessing of image data includes: resizing and normalizing the image data collected by a reflection high-energy electron diffraction (RHEED) system and an in-situ camera, so that the adjusted image size meets the input requirements of the target neural network model, usually adjusted to a fixed pixel size. Secondly, the image data can also be normalized so that the pixel values are within a uniform range (such as the [0, 1] interval) to avoid interference between brightness differences between different images on neural network model training, so that subsequent neural network models can be processed uniformly. Then, in order to improve the generalization ability of the neural network model and prevent overfitting, image enhancement technology is used to process the image data. Among them, enhancement operations include but are not limited to random rotation, translation, scaling, cropping, color dithering, etc. Finally, additional feature extraction can be performed through image processing methods such as edge detection and Hough transform. Key features extracted from the RHEED image can also be input into the target neural network model for analysis and processing. Among them, key features may include, for example, surface morphology, distribution characteristics of diffraction points, changes in diffraction intensity, etc. Since key features can reflect key information such as the nucleation state, crystal quality, surface flatness, etc. of the material, the accuracy of parameter adjustment can be improved.

[0114] According to some embodiments, for non-image data from other sources, such as material growth temperature, source furnace baffle status, etc., since the collected data is presented in a scattered form, each data at a single moment contains limited information. It can be spliced into curve data by collecting all scattered data from the current moment to the past. Among them, the numerical data preprocessing step is similar to the preprocessing of the above-mentioned image data. The preprocessing steps for the non-image data set include: first, through Z-score standardization and Min-Max normalization, to alleviate the problem of large scale differences in numerical data such as growth temperature and source furnace temperature, ensure that all input features have the same scale, and avoid certain features dominating the model training due to excessive numerical ranges. Then, due to sensor failure or environmental interference, there may be missing values in the data set. They can be filled by interpolation, mean filling, or using the average value of the previous and next time steps to perform additional smoothing on the missing values and ensure the continuity and integrity of the data.

[0115] According to some embodiments, preprocessing the first growth parameter data set collected each time to obtain the second growth parameter data set includes:

[0116] When the acquisition devices corresponding to the first growth parameter data in the first growth parameter data set are different, the growth parameter data in the first growth parameter data set are processed according to the acquisition time point of each first growth parameter data to obtain the second growth parameter data set.

[0117] In some embodiments, for the application scenario of each target neural network model, the sample dimension setting can be unified, including single or multiple feature dimensions, such as a multi-dimensional sample consisting of a single data source such as the diffraction feature of the RHEED image, or a multi-dimensional feature consisting of the diffraction feature of the RHEED image, the temperature data collected in real time, the state of the source furnace baffle, etc. Different data types are converted through feature extraction methods so that they can be compared at the same time scale. Finally, after the corresponding preprocessing and feature extraction, all these feature types will be aligned according to the timestamp to ensure that each sample contains data from different sources in the same time window. The fusion of these data not only maintains the independence of various features, but also ensures that they can be used for subsequent analysis and machine learning modeling at the same time scale.

[0118] S203: Inputting the second growth parameter data set into the target neural network model for analysis and processing to obtain an output result, wherein the output result includes information of each category in at least one category and information of each subcategory under each category information;

[0119] The relevant process is as above and will not be repeated here.

[0120] According to some embodiments, the method further comprises:

[0121] Performing focused ion beam cutting on the first historical material to obtain structural performance information and characteristic change information of each layer corresponding to the first historical material;

[0122] Training the initial neural network model based on the structural performance information and characteristic change information of each layer corresponding to the first historical material to obtain output information of the initial neural network model;

[0123] Adjusting the loss function corresponding to the initial neural network model according to the output information and the label information of the first historical material to obtain an adjusted loss function;

[0124] The initial neural network model is adjusted according to the adjusted loss function, and the structural performance information and characteristic change information of each layer corresponding to the second historical material are used to continue training the adjusted initial neural network model until the training information corresponding to the initial neural network model meets the model training requirements, and it is determined that the target neural network model is obtained.

[0125] According to some embodiments, when training the initial neural network model, statistical analysis of historical data is required. For example, focused ion beam cutting can be performed on the prepared material, followed by detailed characterization using electron microscopy or other methods. Detailed structural performance information and characteristic variation information can be obtained for each layer of material growth. These layer structures are closely related to the material growth time, so the preparation results of each layer can be correlated with the material parameter data and characterization data corresponding to that moment. When assigning these sample labels, performance comparisons can be performed on materials with the same structure or characteristics within the same sample. These two components need to differ in material parameter data or in-situ characterization data characteristics, i.e., different performance results. All data acquired during the preparation period of samples with improved performance results can be designated as positive samples. That is, when parameter data or in-situ characterization data corresponding to characteristics similar to those of poorer samples are encountered during the material growth process, the output of the neural network model can be directed to the parameter data corresponding to the material with better performance. In other words, the neural network model can provide parameter adjustment methods for encountering this material preparation state, and assign negative samples when the performance is different. The same method can be used for samples with the same structure, assigning positive or negative samples based on performance changes. This sample contains multi-dimensional samples as input features, and also contains the target output of the neural network model: the top-level category information of the hierarchical classification (such as growth temperature and source furnace baffle status) and the sub-category information below it (such as temperature rise, temperature maintenance, etc.).

[0126] According to some embodiments, after obtaining a complete historical dataset, a suitable machine learning model needs to be constructed. To train the model, all samples can be divided into a training set, a validation set, and a test set. The training set is used for model learning, the validation set is used to adjust model parameters, and the test set is used to evaluate the model's final performance. To train a model capable of multi-classification, an appropriate loss function must be selected. In multi-classification problems, the selected loss function is the cross-entropy loss function. During training, the goal of the neural network model is to continuously optimize the loss function using the backpropagation algorithm to reduce the gap between the predicted and true labels. For example, by using an optimization algorithm (such as the Adam optimizer), the initial neural network model can effectively update its internal weights and parameters, thereby improving classification accuracy. After each training session, if the cross-entropy loss function shows a downward trend, it indicates that the model is continuously learning and optimizing. Accuracy calculation is another key metric for evaluating model performance. In multi-classification problems, accuracy refers to the ratio of the number of samples correctly predicted by the model to the total number of samples. For each category, the initial neural network model calculates a predicted probability and compares it with the true label. If the category information predicted by the initial neural network model is consistent with the actual category information, the sample is considered correctly classified. Finally, the overall accuracy of the initial neural network model is calculated by evaluating the samples of the entire data set. If the overall accuracy meets the accuracy requirements, the target neural network model is determined to be obtained.

[0127] In some embodiments, the input of the target neural network model is a multi-dimensional sample, and the intermediate structure of the target neural network model is to extract image features through a convolutional neural network to form a high-dimensional feature representation. The image features can be directly spliced or spliced with processed time series data features (such as temperature, baffle status, etc.), and then passed into the Transformer encoder (ViT part) to extract high-level features about the growth temperature and the source furnace baffle status. Finally, the output of the target neural network model can, for example, consist of two top-level category information: growth temperature and source furnace baffle status. Each part contains multiple sub-category information. Specifically, the sub-category information of the growth temperature includes "temperature increase", "temperature maintenance" and "temperature reduction", while the sub-category information of the source furnace baffle status includes "on" and "off".

[0128] S204: Analyze and process at least one output result within the first preset time period to obtain a hierarchical classification result, wherein the hierarchical classification result includes information of each category and information of each subcategory under each category;

[0129] The relevant process is as above and will not be repeated here.

[0130] S205: Count the hierarchical classification results to obtain statistical results, wherein the statistical results include a first quantity of each category of information and a second quantity of each subcategory of information under each category of information;

[0131] The relevant process is as above and will not be repeated here.

[0132] S206, determining a target parameter in the first growth parameter data set according to the first quantity, and determining an adjustment strategy for the target parameter according to the second quantity;

[0133] The relevant process is as above and will not be repeated here.

[0134] According to some embodiments, the method further comprises:

[0135] obtaining, according to at least one of a material type of the target material and a priority setting instruction, an adjustment priority of each growth parameter data in the first growth parameter data set;

[0136] The target parameter in the first growth parameter data set is determined based on the adjustment priority and at least one output result within the first preset time period. The adjustment priority may also be referred to as the control priority, which is not limited in the present embodiment.

[0137] S207, adjusting the target parameters using an adjustment strategy;

[0138] The relevant process is as above and will not be repeated here.

[0139] According to some embodiments, the at least one output result may be, for example, 10 output results, which may specifically include:

[0140] 1. (Growth temperature: temperature rising, source furnace damper status: open);

[0141] 2. (Growth temperature: temperature maintenance, source furnace damper status: open);

[0142] 3. (Growth temperature: temperature decreases, source furnace damper status: open);

[0143] 4. (Growth temperature: temperature maintenance, source furnace damper status: open);

[0144] 5. (Growth temperature: temperature rising, source furnace damper status: open);

[0145] 6. (Growth temperature: temperature rising, source furnace damper status: closed);

[0146] 7. (Growth temperature: temperature maintenance, source furnace damper status: open);

[0147] 8. (Growth temperature: temperature rising, source furnace damper status: open);

[0148] 9. (Growth temperature: temperature decreases, source furnace damper status: closed);

[0149] 10. (Growth temperature: temperature maintenance, source furnace damper status: open).

[0150] Statistical analysis of these 10 outputs identifies the most frequent subcategories within each top-level category, providing a basis for subsequent control strategies. First, the frequency of the top-level category outputs was counted: "Temperature Increase" appeared 4 times, "Temperature Maintain" 3 times, "Temperature Reduce" 2 times, "On" 8 times, and "Off" 2 times. Subsequently, the most frequent subcategories were identified. For growth temperature, the target neural network model output was most frequently "Temperature Increase," followed by "Temperature Maintain." Therefore, these two subcategories are considered most important for controlling material growth at this stage. For the source furnace damper status, the target neural network model outputted "On" 8 times, indicating that the source furnace damper should remain open during the current cycle. Based on the statistical analysis results, the top-level categories were ranked: "Temperature Increase" > "Temperature Maintain" > "Temperature Reduce," and "On" > "Off." Finally, based on the statistical analysis and category ranking, a control strategy for material growth can be formulated. For example, in the early stage of material growth, if the classification result of "temperature increase" appears frequently, the target neural network model will determine that the growth temperature needs to be increased, and will therefore adjust the heating power of the growth system or adjust the flow rate; if "temperature maintenance" appears, the target neural network model will choose to keep the temperature stable; and if "temperature reduction" appears more frequently, the target neural network model may indicate to reduce the heating power or adjust the cooling system.

[0151] Regarding the state of the source furnace baffle, since the frequency of "open" is greater than that of "closed", the baffle needs to maintain the "open" state. If the frequencies of "open" and "closed" are equal, the target neural network model needs to maintain the current state first, and then combine the results of the next moment to judge the changing trend of the baffle state and the material growth requirements at the current stage. For example, before the formation of quantum dots, the source furnace baffle may need to remain open. As the quantum dots are close to formation, there will be a trend of gradually increasing "closed" frequency. After the quantum dots are formed, the source furnace baffle may need to be closed to control the material growth rate. The corresponding "closed" frequency is significantly higher than the "on" frequency.

[0152] According to some embodiments, after obtaining the output result and judgment result of the target neural network model, the final adjustment strategy will affect the control system of the electronic device through a feedback mechanism. Since the frequency of changes in the growth temperature output label is more frequent, and the frequency of changes in the source furnace damper state is second, the control priority corresponding to the growth temperature is higher than the source furnace damper. The growth temperature can be adjusted first, and the source furnace damper can be not adjusted in the current time period, and further adjustments can be made based on the judgment results of a subsequent period of time, that is, it is determined based on at least one output result within the next preset time period. For example, if the target neural network model continues to output "temperature rise" during the growth process, the electronic device may automatically increase the heating power; if the target neural network model output frequently displays "damper closed", the electronic device will close the relevant source furnace damper.

[0153] In some embodiments, when the output changes significantly, the corresponding target parameters can be adjusted first. For example, a significant change in output may include a change from only "temperature increase" output to only "temperature maintenance" output, or from only "on" output to only "off" output. Parameters with significant changes have higher priority and require prior adjustment, such as when quantum dot growth is complete.

[0154] The above embodiment uses two top-level categories of information for parameter adjustment. If three or more top-level categories of information appear, such as when setting changes to the source furnace temperature, the adjustment strategy can be determined by further determining which of the three top-level categories of information changes most frequently, obtaining statistical results, and then sorting these three top-level categories of information and setting adjustment strategies accordingly. Through this continuous output statistical analysis, the target neural network model can not only understand the current real-time status of material growth, but also predict future adjustment needs based on historical output and trends, and perform dynamic control to ensure the efficiency and stability of the material growth process.

[0155] According to some embodiments, after obtaining the corresponding top-level category information and sub-category information of the above-mentioned output frequency changes, the mapping process of the adjustment strategy is to convert this statistical result into an executable control command. Each sub-category information corresponds to a set of preset adjustment rules. For example, "temperature increase" may mean that it is necessary to increase the source furnace heating power or adjust the gas source flow to increase the growth temperature, while "open" means that the damper of the source furnace needs to remain open to supply material. Subsequently, the actual control command will be executed according to these mapping rules.

[0156] S208, maintaining the target parameter as the adjusted parameter within a second preset time period, wherein the second preset time period is a time period adjacent to the first preset time period and away from the material growth start time point of the target material.

[0157] The relevant process is as above and will not be repeated here.

[0158] According to some embodiments, the method further comprises:

[0159] Obtaining growth status information corresponding to the target material;

[0160] An adjustment strategy corresponding to the target material is determined according to the growth status information and at least one output result within a second preset time period.

[0161] In some embodiments, since the material preparation process is not completed in one go and requires a period of time, feedback data from the material growth process can be collected in real time during the implementation of parameter adjustment. These data may include re-collected temperature readings, source furnace gas flow, reflection high-energy electron diffraction (RHEED) data, etc. By continuously processing these real-time data, it is possible to verify whether the parameter adjustment meets the expected effect and repeat further adjustments to the growth parameters. For example, if the temperature or RHEED pattern does not achieve the expected change after the source furnace baffle is closed, the target neural network model will readjust the baffle state or other growth parameters based on these feedback information and continue to execute the adjustment strategy. The furnace baffle will automatically open or close according to the output of the target neural network model, thereby adjusting the material growth rate and the size of the quantum dots.

[0162] According to some embodiments, taking the pulling of gallium antimonide (GaSb) single crystals as an example, growth parameter data may include liquid level temperature, melt temperature, melt volume height, pulling speed, and rotation speed. The liquid level temperature refers to the temperature of the melt surface; changes in this value directly affect the crystal growth rate and quality. The temperature of the melt itself also plays a crucial role in the growth process, directly affecting the material's viscosity and fluidity. The melt volume height reflects the remaining melt volume; an excessively low melt volume height can interrupt crystal growth, thus requiring real-time monitoring of the melt height. The melt height is measured in real time by a laser rangefinder or ultrasonic sensor, and the data is fed into the control system. Pulling speed is a key parameter in the crystal growth process and is typically controlled by a servo motor. Therefore, changes in the pulling speed require real-time monitoring. Precisely controlling the pulling speed can adjust the material growth rate and avoid the generation of crystal defects. During the Czochralski single crystal pulling process, the rotation speed of the melt furnace and crystal ingot plays a crucial role in uniform crystal growth. Real-time data on the rotation speed of the melt furnace and crystal ingot is collected using a speed sensor to ensure that all growth parameters remain within a stable range during the growth process.

[0163] In some embodiments, in-situ needle data is collected by observing the internal environment of the cavity through a transparent flange with a camera. In order to obtain visual information during the material growth process, a high-resolution camera and a filter can be equipped. The camera is installed at the observation window of the growth furnace, which can capture the morphological changes of the melt and the surface characteristics of the crystal in real time, as well as the halo effect that can be observed during the pulling and rotation process. By continuously shooting images, the flow of the melt, the surface state and any possible defects, such as surface unevenness, bubbles or other irregular shapes, can also be monitored. For example, during the pulling process, changes in the melt surface may affect the growth quality of the crystal. The camera can provide real-time feedback on whether the surface is uniform or whether there are defects such as cracks through image data. In some cases, image data can also help the system identify whether the growth of the crystal is normal, and whether there is too fast or too slow growth.

[0164] According to some embodiments, the target neural network model may also be referred to as a machine learning model. In order to ensure the integrity and consistency of the growth parameter data, all growth parameter data are marked with a timestamp to ensure the synchronous processing of different types of growth parameter data to ensure the timeliness and accuracy of the growth parameter data. For example, parameters such as melt temperature, liquid surface temperature, and pulling speed need to be synchronously processed with the image data collected by the camera so that the subsequent machine learning model can obtain accurate input. In the preprocessing stage, the image data can be subjected to denoising, normalization, resizing, and other processing to ensure that the image quality meets the requirements of the model analysis. At the same time, the growth parameter data will also be standardized so that the various growth parameter data can be compared and analyzed on a unified scale.

[0165] In an embodiment of the present application, the image data collected in real time by the camera and the parameter data collected by the sensor can be used as in-situ characterization data. The data can be subjected to a series of preprocessing steps to construct a multi-dimensional sample to ensure that high-quality input is provided for subsequent machine learning model analysis. Since the image data collected by the camera usually has problems such as noise, non-uniform resolution or unclear background, the image data can be preprocessed. First, the noise in the image data is removed by an image denoising algorithm (such as median filtering, Gaussian blur, etc.) to ensure that only useful information is retained. Secondly, in order to improve the learning efficiency of the machine learning model, the image data is uniformly resized (for example, to 256×256 pixels) so that all image data samples have the same dimension. Then, the image data is normalized to adjust each pixel value to between 0 and 1 to avoid affecting subsequent processing due to excessive differences in image brightness. Finally, the image data is converted to a color space (for example, from RGB to a grayscale image) to extract key structural information, so that the machine learning model can focus on the key features of the image rather than unimportant color information. To improve the generalization capability of neural network models and avoid overfitting, data augmentation techniques can also be applied during the preprocessing of in-situ characterization data. For image data, new samples can be generated through methods such as rotation, translation, and scaling. For growth parameter data, the dataset can be augmented through methods such as simulated noise addition and time shifting. In addition to the image data captured by the camera, growth parameter data such as the melt surface temperature, melt temperature, and melt volume height are also collected in real time. To ensure that these growth parameter data can be seamlessly integrated with the image data, each growth parameter data is first normalized. The purpose of normalization is to convert data of different dimensions (such as temperature and height) into a relatively uniform scale to facilitate better analysis by machine learning models. For example, the mean of each temperature value is subtracted and then divided by its standard deviation to ensure that it meets the standards of zero mean and unit variance. In addition, for dynamically changing growth parameter data such as melt volume height, the growth parameter data can be smoothed using a time window to reduce the interference of noise on subsequent analysis. Specifically, for growth parameter data that changes rapidly in real time, such as liquid surface temperature, a sliding average or weighted average method can be used to smooth the growth parameter data to ensure that the trend of the parameter value remains consistent over a longer period of time. For example, rapidly changing growth parameter data can refer to parameters whose change rate exceeds a speed threshold.

[0166] In some embodiments, image data and other growth parameter data can be combined to construct a multi-dimensional sample. The image data collected by each frame of the camera will be combined with other growth parameter data collected at that moment (such as melt temperature, liquid surface temperature, melt capacity height) to form a multi-dimensional sample. For example, the pre-processed image data becomes a grayscale image of 256×256 pixels, and the standardized liquid surface temperature, melt temperature, melt capacity height and other parameters. Assume that each growth parameter data is standardized to form a vector. These image data and other growth parameter data are spliced or merged. For example, if the input of a sample contains image data (256×256) and three other growth parameters (liquid surface temperature, melt temperature, melt capacity height), then the sample will become an input vector containing 256×256 pixel image data and 3 other growth parameter values. Alternatively, you can use feature scaling techniques to convert the image data into a one-dimensional vector (for example, convert a 256×256 pixel image into a one-dimensional array of length 65536), then append three other growth parameters to the front of the array, weight them, and incorporate them into the array within the array, or create three additional one-dimensional arrays of length 65536 and concatenate them, ultimately obtaining a multi-dimensional input vector. Alternatively, you can simply use a 256×256 pixel grayscale image as a single multi-dimensional sample.

[0167] During the Czochralski single crystal pulling process, the material growth process is temporally continuous. During multiple sampling processes, continuous images and other growth parameter data will be arranged in chronological order, forming a series of continuous multidimensional samples. To further enhance the temporal nature of the data, time window technology can be used to treat data within a period of time as a single sample for processing. For example, select images and corresponding other growth parameter data within 5 seconds as a time window, and merge multiple data points within this 5-second period into a single sample for input. Such a time window can be continuously updated using sliding window technology, so that each data input into the machine learning model can reflect the real-time changing state. In addition, the data within each time window can be further dimensionalized. This vector can then be normalized and standardized to form a new multidimensional sample, which can be used for subsequent neural network model training and prediction.

[0168] In some embodiments, the machine learning model of the embodiment of the present application can select a convolutional neural network (CNN) as the basic model structure, combined with a fully connected layer to implement a hierarchical classification task. The CNN structure can effectively extract local features in multi-dimensional samples, and extract hierarchical relationships through multi-layer convolution and pooling operations, while the fully connected layer can further process these features or combine them with other growth parameter data (such as liquid surface temperature, melt temperature, melt capacity height, etc.) to perform final multi-classification prediction. The input of the convolutional neural network model is a preprocessed multi-dimensional sample, including image data and other growth parameter data, and the output is a hierarchical classification result, in which each top-level category information represents a main growth parameter (such as pulling speed or rotation speed), and each sub-category information represents a corresponding adjustment method (such as speed increase, speed reduction or speed maintenance).

[0169] The machine learning model can obtain results through multi-classification calculations using the normalized exponential function Softmax, outputting a result containing multiple top-level category information. Each top-level category information is further subdivided according to the definition of sub-category information. Specific top-level category information includes pulling speed and rotation speed, and specific sub-category information includes speed increase, speed decrease, speed maintenance, rotation speed increase, rotation speed maintenance, and rotation speed decrease. Assume that at a specific time step, the machine learning model analyzes the preprocessed samples and obtains the following 10 consecutive classification output results:

[0170] Output 1: Pulling speed: speed increases, rotation speed: rotation speed increases;

[0171] Output 2: Pulling speed: speed increases, rotation speed: rotation speed maintains;

[0172] Output 3: Pulling speed: maintain speed, rotation speed: increase rotation speed;

[0173] Output 4: Pulling speed: speed maintenance, rotation speed: rotation speed maintenance;

[0174] Output 5: Pulling speed: speed decreases, rotation speed: rotation speed increases;

[0175] Output 6: Pulling speed: speed reduced, rotation speed: rotation speed maintained;

[0176] Output 7: Pulling speed: speed increases, rotation speed: rotation speed decreases;

[0177] Output 8: Pulling speed: speed increases, rotation speed: rotation speed maintains;

[0178] Output 9: Pulling speed: speed increases, rotation speed: rotation speed maintains;

[0179] Output 10: Pulling speed: speed maintained, rotation speed: rotation speed maintained.

[0180] Based on these outputs, the machine learning model performs a statistical analysis of the distribution of each top-level category and subcategory. For example, for the classification of pulling speed, there might be 5 outputs of "speed increase," 3 outputs of "speed maintenance," and 2 outputs of "speed reduction." Similarly, for the classification of rotation speed, there might be 6 outputs of "rotation speed maintenance," 3 outputs of "rotation speed increase," and 1 output of "rotation speed reduction."

[0181] Through such statistical analysis, the machine learning model can derive the frequency of each category of information and then determine the current regulatory measures that need to be taken. Determine the regulatory actions that should be performed in this time step. For example, if the pulling speed frequently shows "speed increase" in 10 consecutive outputs, while the rotation speed frequently shows "rotation speed maintenance", the machine learning model may conclude that it is necessary to increase the pulling speed and maintain the rotation speed at the current stage, maintaining a faster pulling speed to promote crystal growth. If "speed maintenance" accounts for 30%, it indicates that at certain time points, the crystal may need to maintain a certain stability. Therefore, the increase in the pulling speed needs to be adjusted within a certain range and should not change too fast or too slow immediately.

[0182] Frequency statistics of the rotation speed output indicate that "maintain rotation speed" appeared six times, the highest probability. This indicates that maintaining the current rotation speed helps maintain crystal growth stability. This typically occurs in the middle of the ingot pulling process, or after the ingot has formed a certain length, when crystal growth approaches a stable stage and frequent rotation speed adjustments are unnecessary. However, in the early stages of crystal growth, a higher rotation speed may be required to ensure uniform distribution of the melt, often leading to an increase in the frequency of the "increase rotation speed" output. Based on these statistical results and the mapped adjustment strategy, adjustments to growth parameter data can be implemented.

[0183] According to some embodiments, during the Czochralski single crystal growth process, a machine learning model can analyze and classify in-situ characterization data in real time to produce a hierarchical classification result regarding the pulling speed and rotation speed. To further optimize the material growth process, adjustments are made based on the classification results. Specifically, based on the frequency distribution obtained from the previous statistical analysis, combined with a preset mapping adjustment strategy, key parameters in the material growth process are dynamically adjusted. Specifically, this may include:

[0184] Through machine learning model analysis, the following 10 output results were obtained within a specific time period: Pulling speed classification results: 7 "speed increase," 3 "speed maintenance," and 0 "speed reduction"; rotation speed classification results: 8 "rotation speed increase," 2 "rotation speed maintenance," and 0 "rotation speed reduction." Growth parameters were dynamically adjusted based on these frequency statistics and a pre-set adjustment strategy.

[0185] Since "speed increase" accounts for 70% of the pulling speed results and "speed reduction" does not appear, this indicates that under the growth conditions of the material at the current stage, a faster pulling speed is beneficial to the growth of the crystal; when the pulling speed is "speed maintenance" (accounting for 30%), it means that the crystal growth is relatively stable at this time, and the pulling speed can be adjusted to keep it within a suitable range to avoid excessive consumption of the melt due to excessive speed. Therefore, the mapped adjustment strategy will set the pulling speed to increase as much as possible in the current stage, and gradually increase the pulling speed to the predetermined maximum safety range to support efficient single crystal growth. If the pulling speed continues to remain at a high level without decreasing, it can be determined whether the pulling speed needs to be reduced by continuously monitoring the liquid surface temperature and the melt temperature.

[0186] "Increase in rotation speed" accounts for 80%, which shows that at this stage, the rotation speed plays an important role in crystal growth. The increase in rotation speed is usually to ensure the uniform distribution of the melt and the stability of crystal nucleation, reduce the temperature gradient during crystal growth, and ensure the quality of the crystal. Therefore, according to the results of frequency analysis, the rotation speed continues to be increased, especially for the rotation speed of the melt furnace, to enhance the stirring effect of the melt. In the case of "rotation speed maintenance" (accounting for 20%), it means that the current rotation speed is sufficient to maintain the stability of the melt and crystal. At this time, the rotation speed will be maintained unchanged to ensure uniformity during the growth process. Through a dynamic adjustment strategy based on statistical results, the pulling speed and rotation speed can be automatically adjusted during the material growth process to optimize the crystal growth process. This adjustment mechanism based on machine learning classification results and frequency analysis makes the material growth process more stable and efficient, and also provides intelligent support for adjustments at different growth stages.

[0187] After each adjustment cycle (i.e., each preset duration) of material growth, real-time growth data continues to be collected. This data includes, but is not limited to, parameters related to material growth, such as liquid surface temperature, melt temperature, and melt volume height. Other in-situ characterization data is also collected, such as material surface images captured by a camera or melt state data obtained by other sensors. This preprocessed data is then passed back as input to the machine learning model. The machine learning model analyzes the current multi-dimensional data and outputs the latest hierarchical classification results based on the analysis results. These outputs include adjustment strategies for different growth parameter data (such as changes in pull speed and rotation speed). A feedback mechanism enables closed-loop control. When the newly acquired growth parameter data is fed back to the machine learning model, it recalculates and optimizes the adjustment strategy based on the current growth status. For example, suppose the pull speed was increased during the previous adjustment cycle. However, based on the real-time feedback of the liquid surface temperature data, the machine learning model may determine that the temperature is too high and output "temperature reduction" as the adjustment strategy. A new signal is then sent to the control system to automatically reduce the pull speed, thereby preventing excessive temperatures from affecting crystal quality.

[0188] During the growth process, the feedback mechanism relies not only on real-time data but also on trend analysis based on historical data. Through statistical analysis and pattern recognition, it can identify which adjustment strategies perform best under specific conditions and which ones may lead to unstable growth. For example, if certain temperature adjustment strategies have resulted in improved growth efficiency over a period of time, while others have led to unstable growth, the machine learning model will learn and prioritize those successful adjustment methods. This continuous learning and optimization not only ensures optimal adjustment during the current growth process but also improves performance for future material growth. Furthermore, if the machine learning model inputs the "melt capacity height" parameter, and as the melt capacity gradually decreases, the melt supply becomes scarce, the machine learning model needs to dynamically adjust the pull speed based on the melt capacity status when the "melt capacity height is too low" result is displayed. Dynamic decision optimization prioritizes "increasing the pull speed" as the adjustment strategy at this time, ensuring that the melt does not run out even at the beginning of the final growth step, thus avoiding interruptions or instability in crystal growth caused by insufficient melt.

[0189] Among them, the growth parameter data set of each fixed time period (for example, every 1 minute or every 20 minutes) is sent to the machine learning model for analysis, reducing the over-frequent data collection and control that may easily lead to over-adjustment of various parameters, resulting in a new optimization cycle before the parameters adjusted in the previous stage or even the previous stages are deployed and optimized. If the adjustment frequency is too low, it is easy to lose the best opportunity for parameter adjustment.

[0190] In an embodiment of the present application, the electronic device may also be, for example, a chemical vapor deposition (CVD) system. In a CVD system, the deposition rate and growth uniformity of the material can be controlled by adjusting parameters such as reaction gas flow, gas flow ratio, cavity pressure, shielding gas pressure, and growth temperature. In addition, the real-time feedback mechanism of the CVD system is similar to that of a molecular beam epitaxy (MBE) system, and in particular, for reduced pressure CVD or low pressure systems, characterization methods in ultra-high vacuum can still be used. The system can obtain surface images and key growth parameter data of the material in real time through integrated sensors, cameras, and other monitoring equipment, and analyze the data in real time in combination with a machine learning model.

[0191] In one or related embodiments, the hierarchical classification results are counted to obtain statistical results, wherein the statistical results include a first quantity of each category information and a second quantity of each subcategory information under each category information. The target parameters in the first growth parameter data set are determined based on the first quantity, and the adjustment strategy of the target parameters is determined based on the second quantity. Therefore, the target parameters and adjustment strategy can be determined based on the frequency of occurrence of each category information and the frequency of occurrence of each subcategory information within the first preset time length, thereby improving the accuracy of parameter adjustment and improving the accuracy of material growth process control.

[0192] In order to implement the above embodiment, the present application also proposes a parameter adjustment device during the material growth process.

[0193] Figure 3 A schematic structural diagram of a parameter adjustment device during a material growth process provided in an embodiment of the present application.

[0194] like Figure 3 As shown, the parameter adjustment device during the material growth process includes:

[0195] The frequency acquisition unit 301 is used to acquire the growth stage of the target material during the material growth process of the target material, and acquire the data acquisition frequency corresponding to the growth stage according to the growth stage and material information of the target material;

[0196] The data processing unit 302 is configured to pre-process the first growth parameter data set collected each time according to the data collection frequency within a first preset time period to obtain a second growth parameter data set, wherein the first growth parameter data set includes at least one growth parameter data;

[0197] The data analysis unit 303 is configured to input the second growth parameter data set into the target neural network model for analysis and processing, and obtain an output result, wherein the output result includes information of each category in at least one category and information of each subcategory under each category information;

[0198] A parameter adjustment unit 304 is configured to obtain an adjustment priority of each growth parameter data in the first growth parameter data set according to at least one of a material type of a target material and a priority setting instruction;

[0199] The parameter adjustment unit 304 is further configured to determine a target parameter in the first growth parameter data set according to the adjustment priority and at least one output result within the first preset time period;

[0200] The parameter adjustment unit 304 is also used to adjust the target parameters in the first growth parameter data set based on at least one output result within the first preset time length, and maintain the target parameters as the adjusted parameters within the second preset time length, wherein the second preset time length is a time length adjacent to the first preset time length and away from the material growth start time point of the target material.

[0201] In one possible embodiment, the first growth parameter data set includes a first image data subset and a first non-image data subset. The data processing unit 302 is configured to preprocess the first growth parameter data set collected each time, and when obtaining the second growth parameter data set, specifically to:

[0202] performing size adjustment and normalization processing on each image data in the first image data subset to obtain a second image data subset;

[0203] performing normalization processing on each image data in the second image data subset to obtain a third image data subset;

[0204] Processing the third image data subset using an image enhancement technique to obtain a fourth image data subset;

[0205] performing standardization and normalization processing on the first non-image data subset to obtain a second non-image data subset;

[0206] Filling missing values on the second non-image data subset to obtain a third non-image data subset;

[0207] adding a fourth subset of image data and a third subset of non-image data to the second set of growth parameter data;

[0208] The second growth parameter data set is subjected to feature extraction processing, feature conversion processing, and feature splicing processing to obtain a processed second growth parameter data set.

[0209] In one possible embodiment, the parameter adjustment unit 304 is configured to adjust the target parameter in the first growth parameter data set according to at least one output result within the first preset time period, specifically to:

[0210] Analyze and process at least one output result within the first preset time period to obtain a hierarchical classification result, wherein the hierarchical classification result includes information of each category and information of each subcategory under each category;

[0211] Performing statistics on the hierarchical classification results to obtain statistical results, wherein the statistical results include a first quantity of each category of information and a second quantity of each subcategory of information under each category of information;

[0212] determining a target parameter in the first growth parameter data set based on the first quantity, and determining an adjustment strategy for the target parameter based on the second quantity;

[0213] Adopt adjustment strategies to adjust target parameters.

[0214] In a possible embodiment, the data analysis unit 303 is further configured to:

[0215] Performing focused ion beam cutting on the first historical material to obtain structural performance information and characteristic change information of each layer corresponding to the first historical material;

[0216] Training the initial neural network model based on the structural performance information and characteristic change information of each layer corresponding to the first historical material to obtain output information of the initial neural network model;

[0217] Adjusting the loss function corresponding to the initial neural network model according to the output information and the label information of the first historical material to obtain an adjusted loss function;

[0218] The initial neural network model is adjusted according to the adjusted loss function, and the structural performance information and characteristic change information of each layer corresponding to the second historical material are used to continue training the adjusted initial neural network model until the training information corresponding to the initial neural network model meets the model training requirements, and it is determined that the target neural network model is obtained.

[0219] In a possible embodiment, the parameter adjustment unit 304 is further configured to:

[0220] Obtaining growth status information corresponding to the target material;

[0221] An adjustment strategy corresponding to the target material is determined according to the growth status information and at least one output result within a second preset time period.

[0222] In one possible embodiment, the data processing unit 302 is configured to preprocess the first growth parameter data set collected each time, and when obtaining the second growth parameter data set, specifically to:

[0223] When the acquisition devices corresponding to the first growth parameter data in the first growth parameter data set are different, the growth parameter data in the first growth parameter data set are processed according to the acquisition time point of each first growth parameter data to obtain the second growth parameter data set.

[0224] It should be noted that the above explanation of the embodiment of the text analysis method based on a large model is also applicable to the text analysis device based on a large model in this embodiment, and will not be repeated here.

[0225] In one or related embodiments, a frequency acquisition unit is used to acquire the growth stage of the target material during the material growth process of the target material, and to acquire the data acquisition frequency corresponding to the growth stage according to the growth stage and the material information of the target material; a data processing unit is used to pre-process the first growth parameter data set acquired each time according to the data acquisition frequency within a first preset time period to acquire a second growth parameter data set, wherein the first growth parameter data set includes at least one growth parameter data; a data analysis unit is used to input the second growth parameter data set into the target neural network model for analysis and processing to acquire an output result, wherein the output result includes information of each category and each category in at least one category. Each subcategory information under the category information; the parameter adjustment unit is also used to obtain the adjustment priority of each growth parameter data in the first growth parameter data set according to the material type of the target material and at least one of the priority setting instructions; the parameter adjustment unit is also used to determine the target parameter in the first growth parameter data set according to the adjustment priority and at least one output result within the first preset time length; the parameter adjustment unit is used to adjust the target parameter in the first growth parameter data set according to at least one output result within the first preset time length, and keep the target parameter as the adjusted parameter within a second preset time length, wherein the second preset time length is a time length adjacent to the first preset time length and away from the material growth start time point of the target material. In this way, multiple parameters in the material growth process can be adjusted, and there is no need to wait for the material to be fully prepared to determine whether a certain parameter is adjusted accurately. Multiple parameters can be adjusted intelligently, and the parameter adjustment information within a preset time period can be determined based on the material preparation information within a preset time period. Based on multi-dimensional data input, multiple growth parameters that need to be adjusted can be output. By analyzing multiple output results, the final adjustment method can be determined, and refined real-time adjustment can be performed during the material growth process. It can be applicable to the growth process control of various materials, improve the flexibility and real-time performance of parameter adjustment, and improve the control accuracy, growth efficiency and stability of the material growth process.

[0226] According to embodiments of the present application, the present application also discloses an electronic device, a computer-readable storage medium, and a computer program product.

[0227] Figure 4 A schematic block diagram of an example electronic device 400 that can be used to implement embodiments of the present application is shown. The electronic device 400 is intended to represent various forms of digital computers, such as laptop computers, desktop computers, workstations, personal digital assistants, servers, blade servers, mainframe computers, and other suitable computers. The electronic device can also represent various forms of mobile devices, such as personal digital assistants, cellular phones, smartphones, wearable devices, and other similar computing devices. The components shown herein, their connections and relationships, and their functions are merely examples and are not intended to limit the implementation of the present application described and / or claimed herein.

[0228] like Figure 4 As shown, electronic device 400 includes a computing unit 401, which can perform various appropriate actions and processes according to a computer program stored in a read-only memory (ROM) 402 or a computer program loaded from a storage unit 408 into a random access memory (RAM) 403. Various programs and data required for the operation of electronic device 400 can also be stored in RAM 403. Computing unit 401, ROM 402, and RAM 403 are connected to each other via a bus 404. An input / output (I / O) interface 405 is also connected to bus 404.

[0229] Multiple components in the electronic device 400 are connected to the I / O interface 405, including an input unit 406, such as a keyboard, a mouse, etc.; an output unit 407, such as various types of displays, speakers, etc.; a storage unit 408, such as a magnetic disk, an optical disk, etc.; and a communication unit 409, such as a network card, a modem, a wireless communication transceiver, etc. The communication unit 409 allows the electronic device 400 to exchange information / data with other devices via a computer network such as the Internet and / or various telecommunication networks.

[0230] The computing unit 401 can be a variety of general and / or special processing components with processing and computing capabilities. Some examples of the computing unit 401 include, but are not limited to, a central processing unit (CPU), a graphics processing unit (GPU), various dedicated artificial intelligence (AI) computing chips, various computing units that run machine learning model algorithms, digital signal processors (DSPs), and any appropriate processors, controllers, microcontrollers, etc. The computing unit 401 performs the various methods and processes described above, such as the parameter adjustment method during the material growth process. For example, in some embodiments, the parameter adjustment method during the material growth process can be implemented as a computer software program, which is tangibly contained in a machine-readable medium, such as a storage unit 408. In some embodiments, part or all of the computer program can be loaded and / or installed on the electronic device 400 via the ROM 402 and / or the communication unit 409. When the computer program is loaded into the RAM 403 and executed by the computing unit 401, one or more steps of the parameter adjustment method during the material growth process described above can be performed. Alternatively, in other embodiments, the computing unit 401 may be configured to execute the parameter adjustment method during the material growth process in any other appropriate manner (eg, by means of firmware).

[0231] Various embodiments of the systems and techniques described herein can be implemented in digital electronic circuit systems, integrated circuit systems, field programmable gate arrays (FPGAs), application specific integrated circuits (ASICs), application specific standard products (ASSPs), system-on-chip systems (SOCs), programmable logic devices (CPLDs), computer hardware, firmware, software, and / or combinations thereof. These various embodiments can include being implemented in one or more computer programs that are executable and / or interpreted on a programmable system that includes at least one programmable processor, which can be a special purpose or general purpose programmable processor that can receive data and instructions from a storage system, at least one input device, and at least one output device, and transmit data and instructions to the storage system, the at least one input device, and the at least one output device.

[0232] The program code of the computer program product for implementing the method of the present application can be written in any combination of one or more programming languages. Such program code can be provided to a processor or controller of a general-purpose computer, a special-purpose computer, or other programmable data processing device, so that when the program code is executed by the processor or controller, the functions / operations specified in the flow chart and / or block diagram are implemented. The program code can be executed entirely on the machine, partially on the machine, as a stand-alone software package, partially on the machine and partially on a remote machine, or entirely on a remote machine or server.

[0233] In the context of the present application, a computer-readable storage medium can be a tangible medium that can contain or store a program for use by an instruction execution system, device or equipment or used in combination with an instruction execution system, device or equipment. A computer-readable storage medium can be a machine-readable signal medium or a machine-readable storage medium. A computer-readable storage medium can include, but is not limited to, electronic, magnetic, optical, electromagnetic, infrared or semiconductor systems, devices or equipment, or any suitable combination of the foregoing. A more specific example of a computer-readable storage medium can include an electrical connection based on one or more lines, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the foregoing.

[0234] To provide interaction with a user, the systems and techniques described herein can be implemented on a computer having: a display device (e.g., a CRT (cathode ray tube) or LCD (liquid crystal display) monitor) for displaying information to the user; and a keyboard and pointing device (e.g., a mouse or trackball) through which the user can provide input to the computer. Other types of devices can also be used to provide interaction with the user; for example, the feedback provided to the user can be any form of sensory feedback (e.g., visual feedback, auditory feedback, or tactile feedback); and input from the user can be received in any form (including acoustic input, voice input, or tactile input).

[0235] The systems and techniques described herein can be implemented in a computing system that includes back-end components (e.g., as a data server), or a computing system that includes middleware components (e.g., an application server), or a computing system that includes front-end components (e.g., a user computer with a graphical user interface or a web browser through which a user can interact with implementations of the systems and techniques described herein), or a computing system that includes any combination of such back-end components, middleware components, or front-end components. The components of the system can be interconnected by any form or medium of digital data communication (e.g., a communication network). Examples of communication networks include: a local area network (LAN), a wide area network (WAN), the Internet, and a blockchain network.

[0236] A computer system may include a client and a server. The client and server are generally remote from each other and typically interact via a communication network. This client-server relationship is established by computer programs running on the respective computers, establishing a client-server relationship. The server may be a cloud server, also known as a cloud computing server or cloud host, a host product within the cloud computing service ecosystem that addresses the management difficulties and limited scalability of traditional physical hosts and VPS services ("Virtual Private Servers" or simply "VPS"). The server may also be a server in a distributed system or a server integrated with blockchain.

[0237] It should be understood that the various forms of the processes shown above can be used to reorder, add, or delete steps. For example, the steps described in this application can be performed in parallel, sequentially, or in a different order, as long as the desired results of the technical solutions disclosed in this application can be achieved. This is not a limitation herein.

[0238] The above specific embodiments do not constitute a limitation on the scope of protection of this application. Those skilled in the art will appreciate that various modifications, combinations, sub-combinations, and substitutions may be made based on design requirements and other factors. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of this application shall be included within the scope of protection of this application.

Claims

1. A method for adjusting parameters of material growth, characterized in that: include: During the material growth process of the target material, obtaining a growth stage of the target material, and obtaining a data acquisition frequency corresponding to the growth stage according to the growth stage and material information of the target material; within a first preset time period, preprocessing the first growth parameter data set collected each time according to the data collection frequency to obtain a second growth parameter data set, wherein the first growth parameter data set includes at least one growth parameter data; Inputting the second growth parameter data set into a target neural network model for analysis and processing to obtain an output result, wherein the output result includes information of each category in at least one category and information of each subcategory under the information of each category; acquiring, according to at least one of a material type of the target material and a priority setting instruction, an adjustment priority of each growth parameter data in the first growth parameter data set; determining a target parameter in the first growth parameter data set according to the adjustment priority and at least one output result within the first preset time period; Based on at least one output result within the first preset time length, the target parameter in the first growth parameter data set is adjusted, and the target parameter is maintained as the adjusted parameter within a second preset time length, wherein the second preset time length is adjacent to the first preset time length and away from the material growth start time point of the target material.

2. The method according to claim 1, characterized in that in, The first growth parameter data set includes a first image data subset and a first non-image data subset, and the preprocessing of the first growth parameter data set collected each time to obtain the second growth parameter data set includes: performing size adjustment and normalization processing on each image data in the first image data subset to obtain a second image data subset; performing normalization processing on each image data in the second image data subset to obtain a third image data subset; Processing the third image data subset using an image enhancement technique to obtain a fourth image data subset; performing standardization and normalization processing on the first non-image data subset to obtain a second non-image data subset; performing missing value filling on the second non-image data subset to obtain a third non-image data subset; adding the fourth subset of image data and the third subset of non-image data to a second set of growth parameter data; The second growth parameter data set is subjected to feature extraction processing, feature conversion processing, and feature splicing processing to obtain a processed second growth parameter data set.

3. The method according to claim 1, characterized in that The adjusting the target parameter in the first growth parameter data set according to at least one output result within the first preset time period includes: Analyzing and processing at least one output result within the first preset time period to obtain a hierarchical classification result, wherein the hierarchical classification result includes each category information and each subcategory information under each category information; Performing statistics on the hierarchical classification results to obtain statistical results, wherein the statistical results include a first quantity of each category of information and a second quantity of each subcategory of information under each category of information; determining a target parameter in the first growth parameter data set according to the first quantity, and determining an adjustment strategy for the target parameter according to the second quantity; The target parameter is adjusted using the adjustment strategy.

4. The method according to claim 1, wherein The method further comprises: Performing focused ion beam cutting on the first historical material to obtain structural performance information and characteristic change information of each layer corresponding to the first historical material; Training an initial neural network model based on the structural performance information and characteristic change information of each layer corresponding to the first historical material, and obtaining output information of the initial neural network model; Adjusting the loss function corresponding to the initial neural network model according to the output information and the label information of the first historical material to obtain an adjusted loss function; The initial neural network model is adjusted according to the adjusted loss function, and the adjusted initial neural network model is continued to be trained using the structural performance information and characteristic change information of each layer corresponding to the second historical material, until the training information corresponding to the initial neural network model meets the model training requirements, and it is determined that the target neural network model is obtained.

5. The method according to claim 1, characterized in that The method further comprises: Obtaining growth status information corresponding to the target material; An adjustment strategy corresponding to the target material is determined according to the growth status information and at least one output result within the second preset time period.

6. The method according to claim 1, characterized in that The preprocessing of the first growth parameter data set collected each time to obtain the second growth parameter data set includes: When the acquisition devices corresponding to each first growth parameter data in the first growth parameter data set are different, each growth parameter data in the first growth parameter data set is processed according to the acquisition time point of each first growth parameter data to obtain the second growth parameter data set.

7. A parameter adjustment device during material growth, characterized in that: include: a frequency acquisition unit, configured to acquire a growth stage of the target material during its growth process, and acquire a data acquisition frequency corresponding to the growth stage based on the growth stage and material information of the target material; a data processing unit, configured to pre-process the first growth parameter data set collected each time according to the data collection frequency within a first preset time period to obtain a second growth parameter data set, wherein the first growth parameter data set includes at least one growth parameter data; a data analysis unit, configured to input the second growth parameter data set into a target neural network model for analysis and processing, and obtain an output result, wherein the output result includes information of each category in at least one category and information of each subcategory under the information of each category; a parameter adjustment unit, configured to obtain an adjustment priority of each growth parameter data in the first growth parameter data set according to at least one of a material type of the target material and a priority setting instruction; The parameter adjustment unit is further configured to determine a target parameter in the first growth parameter data set according to the adjustment priority and at least one output result within the first preset time period; The parameter adjustment unit is further used to adjust the target parameters in the first growth parameter data set based on at least one output result within the first preset time length, and maintain the target parameters as the adjusted parameters within a second preset time length, wherein the second preset time length is adjacent to the first preset time length and away from the material growth start time point of the target material.

8. An electronic device, characterized in that: include: processor; a memory for storing instructions executable by the processor; The processor is configured to execute the instructions to implement the parameter adjustment method in the material growth process according to any one of claims 1 to 6.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the parameter adjustment method in the material growth process according to any one of claims 1 to 6 is implemented.

Images