A gallium oxide quality prediction method, preparation method and system based on deep learning and the pulling method

Abstract

The application discloses a gallium oxide quality prediction method and system based on deep learning and a pulling method, and a preparation method, and the quality prediction method comprises the following steps: acquiring preparation data of a gallium oxide single crystal prepared by a pulling method, wherein the preparation data comprises seed crystal data, environment data and control data; pre-processing the preparation data to obtain pre-processed preparation data; and inputting the pre-processed preparation data into a trained neural network model to obtain corresponding predicted quality data of the gallium oxide single crystal by the trained neural network model. The quality of the gallium oxide single crystal can be predicted by the trained neural network model, so that the preparation data can be adjusted to obtain the required performance of the gallium oxide single crystal, and the performance of the gallium oxide single crystal is optimized.

Description

Technical Field

[0001] This invention relates to the field of gallium oxide crystal preparation, and in particular to a method, preparation method and system for predicting the quality of gallium oxide based on deep learning and the Czochralski method. Background Technology

[0002] β-Ga₂O₃ (gallium oxide) is a direct wide-bandgap semiconductor material with a bandgap of approximately 4.8–4.9 eV. It possesses numerous advantages, including a large bandgap, high saturated electron drift velocity, high thermal conductivity, high breakdown field strength, and chemical stability, making it a promising candidate for applications in high-temperature, high-frequency, and high-power power electronic devices. Furthermore, it can be used in LED chips, solar-blind ultraviolet detectors, various sensor components, and camera elements.

[0003] The Czochralski method is one of the methods for preparing gallium oxide. However, in the current technology, when preparing gallium oxide using the Czochralski method, it is difficult to control the preparation parameters of gallium oxide, and there are too many factors affecting the performance of gallium oxide products, making it impossible to obtain gallium oxide with better properties.

[0004] Therefore, existing technologies still need to be improved and developed. Summary of the Invention

[0005] The technical problem to be solved by this invention is to provide a method, preparation method and system for predicting the quality of gallium oxide based on deep learning and the Czochralski method, which addresses the shortcomings of the existing technology. The aim is to solve the problem that the existing process of preparing gallium oxide crystals by the Czochralski method relies on the operator's experience to set parameters, which has poor repeatability and results in poor quality and stability of the obtained gallium oxide crystals.

[0006] To solve the above-mentioned technical problems, the technical solution adopted by the present invention is as follows:

[0007] A method for predicting gallium oxide quality based on deep learning and the Czochralski method, comprising the following steps:

[0008] Acquire preparation data for gallium oxide single crystals prepared by the Czochralski method, wherein the preparation data includes seed crystal data, environmental data, and control data;

[0009] The preparation data is preprocessed to obtain preprocessed preparation data;

[0010] The preprocessed data is input into a trained neural network model, and the predicted quality data corresponding to the gallium oxide single crystal is obtained through the trained neural network model.

[0011] The gallium oxide quality prediction method based on deep learning and the Czochralski method includes the step of preprocessing the preparation data to obtain preprocessed preparation data, which includes:

[0012] Based on the seed crystal data, environmental data, and control data, pretreatment preparation data is obtained, which is a matrix formed by the seed crystal data, environmental data, and control data.

[0013] The gallium oxide quality prediction method based on deep learning and Czochralski method, wherein the seed crystal data includes: seed crystal diffraction peak full width at half maximum (FWHM), seed crystal diffraction peak FWHM deviation value, and seed crystal diameter;

[0014] The environmental data includes: thermal resistance of the upper insulation cover, thermal resistance deviation of the upper insulation cover, shape factor of the crystal growth channel, shape factor of the crystal growth observation hole, thermal resistance of the lower insulation cover, thermal resistance deviation of the lower insulation cover, relative height between the crucible and the induction coil, and relative height between the heating ring and the induction coil.

[0015] The control data includes: heating power, cooling power, atmosphere type, cavity pressure, gas flow rate, seed crystal rod pulling speed, crystal rotation speed, crystallization rate, and crystal to crucible diameter ratio.

[0016] The gallium oxide quality prediction method based on deep learning and the Czochralski method includes the following steps for obtaining preprocessing preparation data based on the seed crystal data, environmental data, and control data:

[0017] Based on the seed crystal data, environmental data, and control data, a preparation vector is determined. The first element of the preparation vector is one of the seed crystal diffraction peak full width at half maximum (FWHM), the seed crystal diffraction peak FWHM deviation value, and the seed crystal diameter. The second element of the preparation vector is one of the upper insulation cover thermal resistance value, the upper insulation cover thermal resistance deviation value, the crystal growth channel shape factor, the crystal growth observation hole shape factor, the lower insulation cover thermal resistance value, the lower insulation cover thermal resistance deviation value, and one of the relative height between the crucible and the induction coil, and the relative height between the heating ring and the induction coil. The third element of the preparation vector is one of the heating power, cooling power, atmosphere type, cavity pressure, gas flow rate, seed crystal rod pulling speed, crystal rotation speed, crystallization rate, and the crystal-to-crucible diameter ratio.

[0018] The preprocessing preparation data is determined based on the preparation vector.

[0019] The gallium oxide quality prediction method based on deep learning and Czochralski method includes the following predicted quality data: predicted crack data, predicted impurity data, predicted diffraction peak full width at half maximum (FWHM), predicted radial deviation of predicted FWHM, and predicted axial deviation of predicted FWHM.

[0020] The gallium oxide quality prediction method based on deep learning and the Czochralski method, wherein the trained neural network model is trained using the following steps:

[0021] Acquire training data for preparing gallium oxide single crystals by the Czochralski method, as well as corresponding actual quality data, wherein the training data includes: seed crystal training data, environmental training data, and control training data;

[0022] The training data is preprocessed to obtain preprocessed training data;

[0023] The preprocessed training data is input into a preset neural network model, and the preset neural network model is used to obtain the prediction training generation quality data corresponding to the preprocessed training data.

[0024] The model parameters of the preset neural network model are adjusted and corrected based on the predicted training quality data and the actual quality data to obtain a trained neural network model.

[0025] The gallium oxide quality prediction method based on deep learning and the Czochralski method, wherein the preset neural network model includes: a feature extraction module and a fully connected module.

[0026] The steps of inputting the preprocessed training data into a preset neural network model and obtaining prediction training generation quality data corresponding to the preprocessed training data through the preset neural network model include:

[0027] The preprocessed training data is input into the feature extraction module, and the feature extraction module obtains the feature vector corresponding to the preprocessed training data.

[0028] The feature vector is input into the fully connected module, and the prediction training generation quality data corresponding to the preprocessed training data is obtained through the fully connected module.

[0029] A method for preparing gallium oxide based on deep learning and the Czochralski method, wherein the preparation method includes the following steps:

[0030] Obtain the target quality data of the target gallium oxide single crystal;

[0031] Based on the target quality data and the trained neural network model, the target preparation data corresponding to the target gallium oxide single crystal is determined, wherein the target preparation data includes: target seed crystal data, target environment data, and target control data;

[0032] Based on the Czochralski method, the target gallium oxide single crystal was prepared according to the target preparation data.

[0033] The gallium oxide preparation method based on deep learning and the Czochralski method, wherein determining the target preparation data corresponding to the target gallium oxide single crystal based on the target quality data and the trained neural network model includes the following steps:

[0034] Obtain preset preparation data, and preprocess the preset preparation data to obtain preprocessed preset preparation data;

[0035] The pre-processed preset preparation data is input into a trained neural network model, and the predicted quality data corresponding to the pre-processed preset preparation data is obtained through the trained neural network model.

[0036] Based on the predicted quality data and the target quality data, the preset preparation data is corrected to obtain the target preparation data corresponding to the target gallium oxide single crystal.

[0037] A gallium oxide preparation system based on deep learning and the Czochralski method, comprising a memory and a processor, wherein the memory stores a computer program, characterized in that the processor executes the computer program to implement the steps of the prediction method according to any one of the present invention, or the steps of the preparation method according to any one of the present invention.

[0038] Beneficial Effects: This invention proposes a gallium oxide (GaO) quality prediction method and preparation method based on deep learning and the Czochralski method. First, the preparation data of GaO single crystals prepared by the Czochralski method is preprocessed to obtain preprocessed preparation data. Then, the preprocessed preparation data is input into a trained neural network model, and the predicted quality data corresponding to the GaO single crystal is obtained through the trained neural network model. This invention can predict the quality of GaO single crystals through a trained neural network model, thus allowing for adjustments to the preparation data to obtain the desired performance of the GaO single crystal, thereby optimizing the performance of the GaO single crystal. Attached Figure Description

[0039] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings in the following description are only some embodiments recorded in the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0040] Figure 1 This is a schematic diagram of the structure of a crystal growth furnace for preparing gallium oxide crystals by the Czochralski method provided by the present invention.

[0041] Figure 2 This is a flowchart of a preferred embodiment of a gallium oxide quality prediction method based on deep learning and the Czochralski method provided by the present invention.

[0042] Figure 3 This is a flowchart of a preferred embodiment of a gallium oxide preparation method based on deep learning and the Czochralski method provided by the present invention.

[0043] Figure 4 This is an internal structure diagram of a gallium oxide preparation system based on deep learning and the Czochralski method provided by the present invention. Detailed Implementation

[0044] This invention provides a method and system for preparing gallium oxide crystals based on deep learning and the Czochralski method. To make the objectives, technical solutions, and effects of this invention clearer and more explicit, the following detailed description, with reference to the accompanying drawings and embodiments, further illustrates the invention. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of the invention.

[0045] Those skilled in the art will understand that, unless specifically stated otherwise, the singular forms “a,” “an,” “the,” and “the” used herein may also include the plural forms. It should be further understood that the term “comprising” as used in this specification means the presence of the stated features, integers, steps, operations, elements, and / or components, but does not exclude the presence or addition of one or more other features, integers, steps, operations, elements, components, and / or groups thereof. It should be understood that when we say an element is “connected” or “coupled” to another element, it can be directly connected or coupled to the other element, or there may be intermediate elements. Furthermore, “connected” or “coupled” as used herein can include wireless connections or wireless coupling. The term “and / or” as used herein includes all or any units and all combinations of one or more associated listed items.

[0046] It will be understood by those skilled in the art that, unless otherwise defined, all terms used herein (including technical and scientific terms) have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains. It should also be understood that terms such as those defined in general dictionaries should be understood to have the same meaning as in the context of the prior art, and should not be interpreted in an idealized or overly formal sense unless specifically defined as herein.

[0047] The invention will be further explained below with reference to the accompanying drawings and the description of the embodiments.

[0048] The Czochralski method is a traditional crystal growth method. In this method, the raw material in a crucible is melted by heating, the melt temperature is precisely adjusted, and then a seed crystal is immersed in the melt. Simultaneously, the crystal is pulled and rotated at an appropriate speed to induce crystallization in the crucible. Czochralski crystal growth requires a high level of expertise in temperature field design, heating power stability, and operator experience.

[0049] This invention provides a crystal growth furnace for growing gallium oxide crystals using the Czochralski method, such as... Figure 1As shown, it includes a lower heat insulation cover 10, a crucible 11 disposed inside the lower heat insulation cover 10, and a heating ring 12 disposed outside the crucible 11; an upper heat insulation cover 13 is disposed above the lower heat insulation cover, a crystal growth channel 14 is disposed inside the upper heat insulation cover 13, a crystal growth observation hole 15 is disposed on the side of the upper heat insulation cover, and an induction coil 16 is disposed around the lower heat insulation cover 10.

[0050] Gallium oxide crystals can be prepared using the aforementioned crystal growth furnace. First, the crystal growth furnace for preparing gallium oxide crystals is installed. This includes the selection of the upper insulation cover material, the size of the axial opening of the upper insulation cover, the shape and size of the opening window, the selection of the lower insulation cover material, the height position of the crucible relative to the induction coil, and the relative height of the heating ring and the induction coil. These factors will affect the thermal field distribution inside the device, thereby affecting the product performance of the gallium oxide crystal.

[0051] A β-Ga₂O₃ seed crystal with a specific orientation is placed in a seed crystal fixture and bound and fixed. The orientation of the seed crystal can be

[010] ,

[001] , etc. First, the mechanical pump and diffusion pump are turned on in sequence to evacuate the equipment. When the predetermined vacuum level is reached, the vacuum equipment is turned off, and gas is slowly filled into the crystal growth furnace according to the mixed gas volume ratio. Then, by setting the heating power, the crucible is heated through the heating ring to completely melt the gallium oxide raw material placed in the crucible. Subsequently, the seed crystal rod is slowly lowered to immerse the seed crystal in the melt, while it is pulled up and rotated at an appropriate speed to induce the melt in the crucible to crystallize and obtain gallium oxide crystal. During this process, the lower and upper heat insulation covers are used to maintain a stable thermal field during crystal growth. In the process of preparing gallium oxide crystal, the heating power, cooling power, atmosphere type, cavity pressure, atmosphere cooling water flow rate, seed crystal rod pulling speed, crystal rotation speed, etc., will all affect the product performance of gallium oxide crystal.

[0052] Based on this, this embodiment provides a gallium oxide quality prediction method based on deep learning and the Czochralski method, such as... Figure 2 As shown, it includes the following steps:

[0053] S100. Obtain preparation data for preparing gallium oxide single crystals by the Czochralski method, wherein the preparation data includes seed crystal data, environmental data, and control data.

[0054] Specifically, the preparation data refers to the data used in preparing gallium oxide single crystals using the Czochralski method. This preparation data can be configured as needed. For example, if it is necessary to predict the performance of gallium oxide single crystals obtained under a certain preparation data, it is only necessary to determine the preparation data, preprocess the preparation data to obtain preprocessed preparation data, and then input the preprocessed preparation data into a trained neural network model. The predicted quality data can be obtained through the trained neural network model. In other words, no experiment is required. After determining the preparation data, the quality data of gallium oxide single crystals can be predicted through the trained neural network model.

[0055] In this embodiment, the preparation data includes: seed crystal data, environmental data, and control data. The seed crystal data refers to the data of the seed crystal used in the process of preparing gallium oxide single crystal by the Czochralski method. The environmental data refers to the data of the environment in which the crystal is located during the preparation of gallium oxide single crystal by the Czochralski method. The control data refers to the data of controlling the crystal growth during the preparation of gallium oxide single crystal by the Czochralski method.

[0056] S200. The preparation data is preprocessed to obtain preprocessed preparation data.

[0057] Specifically, after obtaining the preparation data, the preparation data is first preprocessed to obtain preprocessed preparation data, which can then be input into a trained neural network model so that the trained neural network model can process the preprocessed preparation data.

[0058] In one embodiment of this application, step S200, preprocessing the preparation data to obtain preprocessed preparation data, includes:

[0059] S210. Based on the seed crystal data, environmental data, and control data, pretreatment preparation data is obtained, wherein the pretreatment preparation data is a matrix formed by the seed crystal data, environmental data, and control data.

[0060] Specifically, after obtaining the preparation data, it is first preprocessed to obtain preprocessed preparation data. Since the various sub-data in the preparation data (such as seed crystal data, environmental data, and control data) can influence each other, but the extent of their mutual influence is currently unclear, it is necessary to preprocess the preparation data to rearrange and recombine the various sub-data to form preprocessed preparation data.

[0061] In another embodiment of this application, the seed crystal data includes: seed crystal diffraction peak full width at half maximum (FWHM), seed crystal diffraction peak FWHM deviation value, and seed crystal diameter; the environmental data includes: upper insulation cover thermal resistance value, upper insulation cover thermal resistance deviation value, crystal growth channel shape factor, crystal growth observation hole shape factor, lower insulation cover thermal resistance value, lower insulation cover thermal resistance deviation value, relative height between crucible and induction coil, and relative height between heating ring and induction coil; the control data includes: heating power, cooling power, atmosphere type, cavity pressure, gas flow rate, seed crystal rod pulling speed, crystal rotation speed, crystallization rate, and crystal to crucible diameter ratio.

[0062] Specifically, the full width at half maximum (FWHM) of the seed crystal diffraction peaks can be measured using an X-ray diffractometer. The FWHM deviation values ​​include the radial deviation and the axial deviation. The radial direction is the direction on the horizontal plane, and the axial direction is the direction perpendicular to the horizontal plane, i.e., the vertical axis. The radial deviation value can be obtained by measuring the FWHM of the seed crystal diffraction peaks radially and calculating the difference between the radial and axial deviations. The axial deviation value can be obtained by measuring the FWHM of the seed crystal axially and calculating the difference between the axial deviations. The seed crystal diameter can be directly measured.

[0063] As mentioned above, in adopting Figure 1 In the crystal growth furnace shown, when preparing gallium oxide single crystals, both the upper and lower insulation covers provide a stable thermal field for the growth of the gallium oxide single crystal. The thermal resistance value of the upper insulation cover refers to the temperature difference between the inside and outside of the insulation cover when a unit of heat passes through the upper insulation cover per unit time. The larger the thermal resistance value of the upper insulation cover, the stronger its ability to resist heat transfer and the better its insulation effect.

[0064] The thermal resistance deviation of the upper insulation cover includes both radial and axial deviations. The radial deviation can be obtained by measuring the thermal resistance on both radial sides of the upper insulation cover and calculating the difference between these values. Similarly, the axial deviation can be obtained by measuring the thermal resistance on both axial sides of the upper insulation cover and calculating the difference between these values.

[0065] Similarly, the thermal resistance value of the lower insulation cover and the deviation value of the thermal resistance value of the lower insulation cover can be obtained by using the same method. The deviation value of the thermal resistance value of the lower insulation cover includes the radial deviation value and the axial deviation value of the thermal resistance value of the lower insulation cover.

[0066] The crystal growth channel shape factor refers to the value of the shape and size of the crystal growth channel. For example, when the crystal growth channel is cylindrical, the crystal growth channel shape factor includes the diameter and height of the crystal growth channel; when the crystal growth channel is cubic, the crystal growth channel shape factor includes the length, height and width of the crystal growth channel. The crystal growth channel shape factor also affects the thermal field distribution of the crystal growth environment, thereby affecting the crystal growth performance.

[0067] The crystal growth observation hole shape factor also refers to the value of the shape and size of the crystal growth observation hole. The shape of the crystal growth observation hole will also affect the thermal field distribution of the crystal growth environment, thereby affecting the crystal growth performance.

[0068] Since the crucible is heated by an induction coil and a heating ring, the relative height between the crucible and the induction coil, and the relative height between the heating ring and the induction coil, will affect the thermal field distribution of the crystal growth environment.

[0069] As the equipment for growing gallium oxide crystals is used, the thermal resistance values ​​of the upper and lower insulation covers, as well as the deviation values ​​of the upper and lower insulation covers, will change. However, these changes will not occur in a short period of time. These environmental data can be retested after a certain number of crystal growth cycles.

[0070] The heating power refers to the heating power of the heating ring on the crucible; the cooling power refers to the power of the liquid used to cool the environment inside the crystal growth furnace; the atmosphere type refers to the type of gas introduced into the crystal growth furnace, including O2, Ar, N2, CO2, etc.; the cavity pressure refers to the pressure inside the crystal growth furnace; the gas flow rate refers to the gas flow rate introduced into the crystal growth furnace; the seed rod pulling speed refers to the speed at which the seed rod is pulled upwards during crystal growth; the crystal rotation speed refers to the speed at which the seed rod drives the crystal to rotate during crystal growth; the crystallization rate refers to the proportion of molten gallium oxide raw material to form crystals; and the crystal-to-crucible diameter ratio refers to the ratio of the diameter of the generated crystal to the diameter of the crucible. These parameters are all control parameters affecting the preparation of crystals using the Czochralski method.

[0071] In one embodiment of this example, step S210, the step of obtaining preprocessing preparation data based on the seed crystal data, environmental data, and control data, includes:

[0072] S211. Based on the seed crystal data, environmental data, and control data, determine the preparation vector; wherein, the first element of the preparation vector is one of the seed crystal diffraction peak full width at half maximum (FWHM), the seed crystal diffraction peak FWHM deviation value, and the seed crystal diameter; the second element of the preparation vector is one of the upper heat insulation cover thermal resistance value, the upper heat insulation cover thermal resistance deviation value, the crystal growth channel shape factor, the crystal growth observation hole shape factor, the lower heat insulation cover thermal resistance value, the lower heat insulation cover thermal resistance deviation value, and one of the relative height between the crucible and the induction coil, and the relative height between the heating ring and the induction coil; the third element of the preparation vector is one of the heating power, cooling power, atmosphere type, cavity pressure, gas flow rate, seed crystal rod pulling speed, crystal rotation speed, crystallization rate, and the crystal-to-crucible diameter ratio;

[0073] S212. Determine the preprocessing preparation data based on the preparation vector.

[0074] Specifically, the preparation vector (A, B, C) is determined based on seed crystal data A, environmental data B, and control data C. Seed crystal data A is selected from: seed crystal diffraction peak full width at half maximum (FWHM) A1, seed crystal diffraction peak FWHM deviation value A2, and seed crystal diameter A3. Environmental data B is selected from: upper insulation cover thermal resistance value B1, upper insulation cover thermal resistance deviation value B2, crystal growth channel shape factor B3, crystal growth observation hole shape factor B4, lower insulation cover thermal resistance value B5, lower insulation cover thermal resistance deviation value B6, and the relative height between the crucible and the induction coil B7, and the relative height between the heating ring and the induction coil B8. Control data C is selected from: heating power C1, cooling power C2, atmosphere type C3, cavity pressure C4, gas flow rate C5, seed crystal rod pulling speed C6, crystal rotation speed C7, crystallization rate C8, and the crystal to crucible diameter ratio C9. In other words, in the preparation vector (A, B, C), A can be one of A1, A2, A3, B can be one of B1, B2, B3, B4, B5, B6, B7, B8, and C can be one of C1, C2, C3, C4, C5, C6, C7, C8, C9. This results in 216 preparation vectors.

[0075] Arrange all the preparation vectors in order of their numbers to form a matrix, and you will get the preprocessed preparation data.

[0076] Specifically, the preparation data for the pretreatment are as follows:

[0077]

[0078] Of course, other arrangements were also used to obtain preprocessed preparation data.

[0079] S300. Input the preprocessed preparation data into the trained neural network model, and obtain the predicted quality data corresponding to the gallium oxide single crystal through the trained neural network model.

[0080] Specifically, the predicted quality data includes predicted crack data, predicted impurity crystal data, predicted half-width at half-maximum (FWHM) of diffraction peaks, predicted radial deviation of FWHM of diffraction peaks, and predicted axial deviation of FWHM of diffraction peaks. Crack data refers to crack grade data, and predicted crack data refers to the predicted crack grade data. For example, cracks can be divided into multiple grades. For instance, if cracks are divided into 3 grades, the crack data would be 1, 2, and 3 respectively.

[0081] Impurity data refers to impurity grade data, while predicted impurity data refers to predicted impurity grade data. For example, impurities can be divided into multiple grades. For instance, if impurities are divided into 3 grades, then the impurity data are 1, 2, and 3 respectively.

[0082] The predicted half-width at half-maximum (WHM) of a diffraction peak refers to the predicted WHM of the diffraction peak. The radial deviation value of the predicted WHM of the diffraction peak refers to the predicted difference in the radial WHM of the diffraction peak. The axial deviation value of the predicted WHM of the diffraction peak refers to the predicted difference in the axial WHM of the diffraction peak.

[0083] In some implementations, the trained neural network model is trained using the following steps:

[0084] S01. Obtain training data for preparing gallium oxide single crystals by the Czochralski method, and the corresponding actual quality data, wherein the training data includes: seed crystal training data, environmental training data, and control training data.

[0085] Specifically, training data refers to the data used for training gallium oxide single crystals prepared by the Czochralski method, while actual quality data refers to the actual quality data of the gallium oxide single crystals prepared by the Czochralski method. A training set is formed using the training data and actual quality data, and a pre-defined neural network model is trained based on this training set to obtain the trained neural network model.

[0086] When collecting data to obtain the training set, gallium oxide single crystals were prepared using the Czochralski method, and the data from the preparation of the gallium oxide single crystals were recorded as training data. After obtaining the gallium oxide single crystals, the quality of the gallium oxide single crystals was analyzed to obtain actual quality data. To facilitate the training of the neural network model, as much data as possible can be collected to form the training set.

[0087] S02. Preprocess the training data to obtain preprocessed training data.

[0088] Specifically, after obtaining the training data, the training data is preprocessed to obtain preprocessed training data. The preprocessing process can be referred to in step S200.

[0089] S03. Input the preprocessed training data into a preset neural network model, and obtain the prediction training generation quality data corresponding to the preprocessed training data through the preset neural network model.

[0090] Specifically, the preprocessed training data is input into a preset neural network model, and the predicted training generated quality data is obtained through the preset neural network model. The predicted training generated quality data includes: predicted training generated crack data, predicted training generated impurity data, predicted training generated diffraction peak full width at half maximum (FWHM), predicted training generated diffraction peak FWHM radial deviation value, and predicted training generated diffraction peak FWHM axial deviation value.

[0091] S04. Adjust and correct the model parameters of the preset neural network model based on the predicted training generated quality data and the actual quality data to obtain the trained neural network model.

[0092] Specifically, based on the predicted training generated quality data and the actual quality data, the model parameters of the preset neural network model are corrected, and the process of inputting the preprocessed training data into the preset neural network model and obtaining the predicted training generated quality data corresponding to the preprocessed training data through the preset neural network model (i.e., step S03) continues until the preset training conditions are met and a trained neural network model is obtained.

[0093] Specifically, based on the predicted training quality data and the actual quality data, the model parameters of the preset neural network model are corrected, and the process continues to execute the step of inputting the preprocessed training data into the preset neural network model, and obtaining the predicted training quality data corresponding to the preprocessed training data through the preset neural network model, until the preset training conditions are met, resulting in a trained neural network model. That is, if the preset neural network model meets the preset training conditions, a trained neural network model is obtained. If the preset neural network model does not meet the preset training conditions, the process returns to step S03 until the preset neural network model meets the preset training conditions, resulting in a trained neural network model.

[0094] In one implementation of this invention, a loss function value for a preset neural network model is determined based on the predicted training generated quality data and the actual quality data. The model parameters of the preset neural network model are then corrected based on the loss function value. Specifically, a gradient-based method is used to correct the parameters of the preset neural network model. After determining the loss function value of the preset neural network model, the corrected parameters of the preset neural network model are determined based on the gradient of the loss function value with respect to the parameters of the preset neural network model, the parameters of the preset neural network model, and a preset learning rate.

[0095] The preset training conditions include: the loss function value meets preset requirements and / or the preset number of training iterations of the neural network model reaches a preset number.

[0096] The preset requirements are determined based on the accuracy and efficiency of the preset neural network model. For example, the loss function value of the preset neural network model reaches its minimum value or no longer changes. The preset number of training iterations is the maximum number of training iterations for the preset neural network model, for example, 4000 times.

[0097] The loss functions of the preset neural network model include: mean squared error, root mean square error, and mean absolute error.

[0098] In one implementation of this application, the preset neural network model includes a feature extraction module and a fully connected module.

[0099] For example, the preset neural network model includes: a first convolutional unit, a second convolutional unit, a third convolutional unit, a fourth convolutional unit, and a fully connected unit. Specifically, the first convolutional unit includes: two convolutional layers and one pooling layer. The second, third, and fourth convolutional units each include three convolutional layers and one pooling layer. The fully connected unit includes three fully connected layers.

[0100] Convolutional and fully connected layers are responsible for mapping and transforming the input data. This process uses parameters such as weights and biases, and also requires activation functions. Pooling layers are fixed-function operations. Specifically, convolutional layers extract features; pooling layers perform pooling operations on the input features, changing their spatial dimensions; and fully connected layers fully connect all the data from the previous layer.

[0101] In some embodiments, step S03, inputting the preprocessed training data into a preset neural network model, and obtaining prediction training generation quality data corresponding to the preprocessed training data through the preset neural network model, includes:

[0102] S031 The preprocessed training data is input into the feature extraction module, and the feature extraction module obtains the feature vector corresponding to the preprocessed training data;

[0103] S032. Input the feature vector into the fully connected module, and obtain the prediction training generation quality data corresponding to the preprocessed training data through the fully connected module.

[0104] Specifically, the preprocessed training data is input into a preset neural network model, the feature extraction module in the preset neural network model outputs the feature vector corresponding to the preprocessed training data, and the feature vector is input into the fully connected module to obtain the prediction training generation quality data corresponding to the preprocessed training data output by the fully connected module.

[0105] Based on the aforementioned gallium oxide quality prediction method based on deep learning and the Czochralski method, this embodiment provides a gallium oxide preparation method based on deep learning and the Czochralski method, such as... Figure 3 As shown, the preparation method includes:

[0106] S10. Obtain the target quality data of the target gallium oxide single crystal.

[0107] Specifically, if a target gallium oxide single crystal is required, the target quality data of the target gallium oxide single crystal can be determined first, that is, the desired quality data of the gallium oxide single crystal can be determined. The target quality data includes: target crack data, target impurity crystal data, target diffraction peak full width at half maximum (FWHM), target diffraction peak radial deviation value, and target diffraction peak axial deviation value.

[0108] S20. Based on the target quality data and the trained neural network model, determine the target preparation data corresponding to the target gallium oxide single crystal, wherein the target preparation data includes: target seed crystal data, target environment data, and target control data.

[0109] Specifically, based on the target quality data and the trained neural network model, the target preparation data corresponding to the target gallium oxide single crystal is determined. It should be noted that since different preparation data can yield the same quality data, the target preparation data for the target gallium oxide single crystal is not unique when determining it based on the target quality data and the trained neural network model. A single target preparation data point is determined based on the controllability of each data point among multiple target preparation data points, thereby facilitating the acquisition of the target gallium oxide single crystal.

[0110] In some embodiments, step S20, determining the target preparation data corresponding to the target gallium oxide single crystal based on the target quality data and the trained neural network model, includes:

[0111] S21. Obtain preset preparation data, and preprocess the preset preparation data to obtain preprocessed preset preparation data;

[0112] S22. Input the pre-processed preset preparation data into the trained neural network model, and obtain the prediction quality data corresponding to the pre-processed preset preparation data through the trained neural network model;

[0113] S23. Based on the predicted quality data and the target quality data, the preset preparation data is corrected to obtain the target preparation data corresponding to the target gallium oxide single crystal.

[0114] Specifically, pre-prepared data can be preset and preprocessed to obtain pre-processed pre-prepared data. The specific preprocessing process can be found in step S200. The pre-processed pre-prepared data is then input into a trained neural network model to obtain predicted quality data that corrects the pre-processed pre-prepared data. The preset prepared data is then corrected based on the predicted quality data and the target quality data. When the difference between the predicted quality data and the target quality data is less than a preset threshold, the corrected preset prepared data can be used as the target prepared data.

[0115] S30. Based on the Czochralski method, a target gallium oxide single crystal is prepared according to the target preparation data.

[0116] Specifically, after obtaining the target preparation data, the target gallium oxide single crystal can be prepared by the Czochralski method based on the target preparation data.

[0117] Based on the above prediction method or the above preparation method, this invention provides a gallium oxide preparation system based on deep learning and the Czochralski method. This system can be a computer device, with an internal structure as follows: Figure 4As shown, the system includes a processor, memory, network interface, display screen, and input devices connected via a system bus. The processor provides computing and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system and computer programs. The internal memory provides the environment for the operation of the operating system and computer programs in the non-volatile storage media. The network interface is used for communication with external terminals via a network connection. When the computer program is executed by the processor, it implements a prediction method or a preparation method for gallium oxide based on deep learning and the Czochralski method. The display screen can be a liquid crystal display (LCD) or an e-ink display. The input devices can be a touch layer covering the display screen, buttons, a trackball, or a touchpad mounted on the system casing, or an external keyboard, touchpad, or mouse.

[0118] Those skilled in the art will understand that Figure 4 The diagram shown is merely a partial structural diagram related to the present application and does not constitute a limitation on the system to which the present application is applied. A specific system may include more or fewer components than those shown in the diagram, or may combine certain components, or may have different component arrangements.

[0119] In one embodiment, a gallium oxide fabrication system based on deep learning and the Czochralski method is provided, including a memory and a processor. The memory stores a computer program, and the processor executes the computer program to implement the steps of the prediction method or the steps of the fabrication method.

[0120] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for predicting gallium oxide quality based on deep learning and the Czochralski method, characterized in that, Including the following steps: Data for preparing gallium oxide single crystals using the Czochralski method is obtained. This data includes seed crystal data, environmental data, and control data. The crystal growth furnace used in the Czochralski method includes a lower heat shield, a crucible housed within the lower heat shield, and a heating ring surrounding the crucible. An upper heat shield is positioned above the lower heat shield, containing a crystal growth channel and a crystal growth observation hole on its side. An induction coil is also positioned around the lower heat shield. The seed crystal data includes the seed crystal diffraction peak half-height. The environmental data includes: thermal resistance of the upper insulation cover, thermal resistance deviation of the upper insulation cover, shape factor of the crystal growth channel, shape factor of the crystal growth observation hole, thermal resistance of the lower insulation cover, thermal resistance deviation of the lower insulation cover, relative height between the crucible and the induction coil, and relative height between the heating ring and the induction coil; the control data includes: heating power, cooling power, atmosphere type, cavity pressure, gas flow rate, seed crystal rod pulling speed, crystal rotation speed, crystallization rate, and crystal to crucible diameter ratio. Based on the seed crystal data, environmental data, and control data, a preparation vector is determined. The first element of the preparation vector is one of the seed crystal diffraction peak full width at half maximum (FWHM), the seed crystal diffraction peak FWHM deviation value, and the seed crystal diameter. The second element of the preparation vector is one of the upper insulation cover thermal resistance value, the upper insulation cover thermal resistance deviation value, the crystal growth channel shape factor, the crystal growth observation hole shape factor, the lower insulation cover thermal resistance value, the lower insulation cover thermal resistance deviation value, and one of the relative heights of the crucible and the induction coil, and the relative heights of the heating ring and the induction coil. The third element of the preparation vector is one of the heating power, cooling power, atmosphere type, cavity pressure, gas flow rate, seed crystal rod pulling speed, crystal rotation speed, crystallization rate, and the crystal-to-crucible diameter ratio. Based on the preparation vector, pretreatment preparation data is determined, wherein the pretreatment preparation data is a matrix formed by rearranging and combining the seed crystal data, environmental data, and control data; The preprocessed data is input into a trained neural network model, and the predicted quality data corresponding to the gallium oxide single crystal is obtained through the trained neural network model. The trained neural network model was obtained using the following training steps: Acquire training data for preparing gallium oxide single crystals by the Czochralski method, as well as corresponding actual quality data, wherein the training data includes: seed crystal training data, environmental training data, and control training data; The training data is preprocessed to obtain preprocessed training data; The preprocessed training data is input into a preset neural network model, and the preset neural network model is used to obtain the prediction training generation quality data corresponding to the preprocessed training data. The model parameters of the preset neural network model are adjusted and corrected based on the predicted training quality data and the actual quality data to obtain a trained neural network model.

2. The gallium oxide quality prediction method based on deep learning and the Czochralski method according to claim 1, characterized in that, The predicted quality data includes: predicted crack data, predicted impurity data, predicted diffraction peak full width at half maximum (FWHM), predicted diffraction peak FWHM radial deviation value, and predicted diffraction peak FWHM axial deviation value.

3. The gallium oxide quality prediction method based on deep learning and the Czochralski method according to claim 1, characterized in that, The preset neural network model includes: a feature extraction module and a fully connected module. The steps of inputting the preprocessed training data into a preset neural network model and obtaining prediction training generation quality data corresponding to the preprocessed training data through the preset neural network model include: The preprocessed training data is input into the feature extraction module, and the feature extraction module obtains the feature vector corresponding to the preprocessed training data. The feature vector is input into the fully connected module, and the prediction training generation quality data corresponding to the preprocessed training data is obtained through the fully connected module.

4. A method for preparing gallium oxide based on deep learning and the Czochralski method, characterized in that, The preparation method includes the following steps: Obtain the target quality data of the target gallium oxide single crystal; Obtain the preparation data as described in claim 1, and preprocess the preparation data to obtain preprocessed preset preparation data; The pre-processed preset preparation data is input into a trained neural network model, and the predicted quality data corresponding to the pre-processed preset preparation data is obtained through the trained neural network model. Based on the predicted quality data and the target quality data, the preset preparation data is corrected to obtain the target preparation data corresponding to the target gallium oxide single crystal, wherein the target preparation data includes: target seed crystal data, target environment data, and target control data; Based on the Czochralski method, a target gallium oxide single crystal was prepared according to the target preparation data. The trained neural network model was obtained using the following training steps: Acquire training data for preparing gallium oxide single crystals by the Czochralski method, as well as corresponding actual quality data, wherein the training data includes: seed crystal training data, environmental training data, and control training data; The training data is preprocessed to obtain preprocessed training data; The preprocessed training data is input into a preset neural network model, and the preset neural network model is used to obtain the prediction training generation quality data corresponding to the preprocessed training data. The model parameters of the preset neural network model are adjusted and corrected based on the predicted training quality data and the actual quality data to obtain a trained neural network model.

5. A gallium oxide preparation system based on deep learning and the Czochralski method, characterized in that, The device includes a memory and a processor, the memory storing a computer program, characterized in that the processor executes the computer program to implement the steps of the prediction method according to any one of claims 1 to 3, or the steps of the preparation method according to claim 4.

Images