Method and device for determining material growth stage

Abstract

The present application proposes a method and device for determining the growth stage of a material, wherein the method includes: in the material production process, obtaining the analysis diagram of the material at the current moment, and determining the characteristic vector of the analysis diagram, then inputting the characteristic vector into a pre-trained clustering algorithm, obtaining the analysis diagram set to which the characteristic vector output by the clustering algorithm belongs, and determining the annotated growth stage corresponding to the analysis diagram set as the growth stage of the material at the current moment, wherein the clustering algorithm is obtained based on an unsupervised training method, the analysis diagram set is obtained by classifying historical analysis diagrams when training the clustering algorithm based on historical analysis diagrams, and the annotated growth stage corresponding to the analysis diagram set is determined based on the material growth state in each analysis diagram contained in the analysis diagram set. This improves the accuracy of determining the growth stage of the material.

Description

Technical Field

[0001] The present application relates to the technical field of material preparation, and in particular to a method and device for determining a material growth stage. Background Art

[0002] Reflection High-Energy Electron Diffraction (RHEED) is a characterization method used to observe the atomic arrangement on the surface of materials. RHEED is used to collect diffraction patterns during the material growth process. During the material growth process, the substrate rotates continuously, and the RHEED electron beam acts on the surface of the material substrate. After being reflected by the material surface, it acts on the fluorescent screen, and the camera continuously records the imaging of the fluorescent screen to obtain the diffraction pattern during the material growth process. Among them, the diffraction pattern contains periodic material diffraction information. Different periodicities indicate different atomic arrangements on the material surface, that is, different material structures.

[0003] In the process of material growth, the growth stage of the material is usually determined by manually distinguishing the periodicity presented in the diffraction pattern. However, at the initial moment when the material's structure changes, the periodic changes presented in the diffraction pattern may not be obvious and cannot be accurately identified by the human eye. As a result, the growth stage of the material cannot be accurately determined based on the diffraction pattern obtained at the moment when the material's structure changes. Summary of the invention

[0004] This application proposes a method and device for determining the growth stage of a material. The specific scheme is as follows:

[0005] In one aspect, an embodiment of the present application provides a method for determining a material growth stage, comprising:

[0006] During the material production process, obtain the analysis diagram of the material at the current moment and determine the characteristic vector of the analysis diagram;

[0007] Input the feature vector into a pre-trained clustering algorithm, obtain the analysis graph set to which the feature vector output by the clustering algorithm belongs, and determine the annotated growth stage corresponding to the analysis graph set as the growth stage of the material at the current moment;

[0008] Among them, the clustering algorithm is obtained based on an unsupervised training method, the analysis diagram set is obtained by classifying historical analysis diagrams when training the clustering algorithm based on historical analysis diagrams, and the labeled growth stage corresponding to the analysis diagram set is determined based on the material growth state in each analysis diagram contained in the analysis diagram set.

[0009] Another aspect of the present application provides a device for determining a material growth stage, including:

[0010] An acquisition module is used to acquire the analysis diagram of the material at the current moment during the material production process and determine the characteristic vector of the analysis diagram;

[0011] A determination module is used to input the feature vector into a pre-trained clustering algorithm, obtain the analysis graph set to which the feature vector output by the clustering algorithm belongs, and determine the annotated growth stage corresponding to the analysis graph set as the growth stage of the material at the current moment;

[0012] Among them, the clustering algorithm is obtained based on an unsupervised training method, the analysis diagram set is obtained by classifying historical analysis diagrams when training the clustering algorithm based on historical analysis diagrams, and the labeled growth stage corresponding to the analysis diagram set is determined based on the material growth state in each analysis diagram contained in the analysis diagram set.

[0013] Another aspect of the present application provides a computer device, including a processor and a memory;

[0014] The processor runs a program corresponding to the executable program code by reading the executable program code stored in the memory, so as to implement the method as described in the above embodiment.

[0015] Another aspect of the present application provides a computer-readable storage medium having a computer program stored thereon, wherein the program implements the method of the above embodiment when executed by a processor.

[0016] Additional aspects and advantages of the present application will be given in part in the description below, and in part will become apparent from the description below, or will be learned through the practice of the present application. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] The above and / or additional aspects and advantages of the present application will become apparent and easily understood from the following description of the embodiments in conjunction with the accompanying drawings, in which:

[0018] Figure 1 A schematic diagram of a flow chart of a clustering algorithm training method provided in an embodiment of the present application;

[0019] Figure 2 A schematic diagram of a method for determining a material growth stage provided in an embodiment of the present application;

[0020] Figure 3 A schematic flow chart of another method for determining the growth stage of a material provided in an embodiment of the present application;

[0021] Figure 4 A schematic diagram of the structure of a device for determining the growth stage of a material provided in an embodiment of the present application. DETAILED DESCRIPTION

[0022] The embodiments of the present application are described in detail below, and examples of the embodiments are shown in the accompanying drawings, wherein the same or similar reference numerals throughout represent the same or similar elements or elements having the same or similar functions. The embodiments described below with reference to the accompanying drawings are exemplary and are intended to be used to explain the present application, and should not be construed as limiting the present application.

[0023] The method for determining the material growth stage of the embodiment of the present application is executed by the material growth stage determination device provided by the embodiment of the present application (hereinafter referred to as the determination device), which can be configured in a computer device to improve the accuracy of determining the material growth stage.

[0024] The following describes a training method for obtaining a clustering algorithm used in the embodiments of the present application with reference to the accompanying drawings.

[0025] Figure 1 A flowchart of a clustering algorithm training method provided in an embodiment of the present application.

[0026] like Figure 1 As shown, the clustering algorithm training method includes:

[0027] Step 101, determining the growth stages included in the complete growth process of the material, and setting the number of clusters corresponding to the initial clustering algorithm based on the number of growth stages, wherein the growth stages include a transition stage and a stable growth stage.

[0028] The stable growth stage is a stage in which the material structure is single. The transition stage is a stage in which the material is transformed from one stable growth stage to another stable growth stage. The material structure grown in the transition stage is diverse. The initial clustering algorithm can be any clustering algorithm.

[0029] In the present application, during the material growth process, there may be changes in the growth environment, resulting in different material growth stages, and materials with different structures will grow in different material growth stages. Generally, the growth stages that a certain material may experience during the complete growth process are fixed. Therefore, the growth stages included in the complete material growth process can be pre-set in the system.

[0030] For example, in the epitaxial growth process of gallium arsenide GaAs thin film materials, the stable growth stages that may be experienced include the c(2×4) reconstruction growth stage and the c(4×4) reconstruction growth stage. In addition, during the substrate cooling process in the growth environment, the GaAs thin film material structure gradually changes from c(2×4) reconstruction to c(4×4) reconstruction, that is, a material structure transition stage occurs. At this time, while the c(2×4) reconstruction gradually becomes blurred in the diffraction pattern, the c(4×4) reconstruction appears. Afterwards, while the c(4×4) reconstruction gradually becomes clearer in the diffraction pattern, the c(2×4) reconstruction gradually disappears completely. In the growth environment, during the substrate heating process, the GaAs thin film material structure gradually changes from c(4×4) reconstruction back to c(2×4) reconstruction again. At this time, there is a relatively clear state in the diffraction pattern where both c(2×4) reconstruction and c(4×4) reconstruction exist. The transformation stages corresponding to GaAs include the stage of transition from the c(2×4) reconstruction growth stage to the c(4×4) reconstruction growth stage, and the stage of transition from the c(4×4) reconstruction growth stage to the c(2×4) reconstruction growth stage. Therefore, the growth stages corresponding to the GaAs thin film material can be pre-set in the system to include the above four stages.

[0031] For example, in the process of deoxidation of GaAs substrate, the stable stages that may be experienced include the undeoxidized stage and the deoxidized stage. In addition, during the heating process of the substrate, the undeoxidized state will be transformed into the deoxidized state at a certain temperature point, that is, a transition stage of the material structure will appear. At this time, the diffraction pattern is first a halo image with diffuse scattering without obvious features, and at the same time, a faint diffraction spot pattern will gradually appear in the halo image. Afterwards, when the critical deoxidation temperature point is reached, the diffraction spots will become clearer and brighter. Subsequently, when the deoxidation temperature point is reached, more diffraction spots will appear in the diffraction pattern, and even reconstruction lines will appear directly. The corresponding transition stage of GaAs includes the stage of conversion from undeoxidized to deoxidized. Therefore, the growth stage corresponding to the GaAs nanomaterial can be pre-set in the system to include the above three stages.

[0032] During the growth process of gallium nitride GaN thin film materials, the stable growth stages that may be experienced include the stripe state growth stage and the spot state growth stage. Among them, 2D flat growth (thin film) corresponds to the stripe state growth stage, and 3D island growth (nanowire) corresponds to the spot state growth stage. GaN thin film nanomaterials can only gradually change from the stripe state growth stage to the spot state growth stage. Therefore, the transition stage corresponding to the GaN thin film nanomaterial only includes the stage of transition from the stripe state growth stage to the spot state growth stage. The growth stage corresponding to the GaN thin film nanomaterial can be pre-set in the system to include the above three stages.

[0033] The number of clusters corresponding to the clustering algorithm determines the number of clusters finally formed. In order to obtain a clustering algorithm for determining the growth stage of a material, the application sets the number of clusters corresponding to the initial clustering algorithm to the number of growth stages included in the complete growth process of the material, so as to ensure the accuracy of determining the growth stage of the material using the clustering algorithm obtained through training.

[0034] Optionally, during the closing and opening of the RHEED shutter, the shutter metal baffle may partially or completely block the fluorescent screen where the diffraction pattern is located, resulting in the acquired diffraction pattern being incomplete or not containing valid diffraction information. Alternatively, in the case of RHEED abnormality (for example, when no electron beam is emitted, the diffraction pattern does not change with the substrate rotation cycle, that is, when the substrate is in an abnormal state of being stopped), the acquired diffraction pattern does not contain valid diffraction information.

[0035] Therefore, the type of abnormal analysis graph can be pre-set in the system. After that, the type of abnormal analysis graph can be obtained from the system, and the sum of the number of growth stages and the number of types is set to the number of clusters corresponding to the initial clustering algorithm. So that the abnormal analysis graph is not divided into the analysis graph set corresponding to any growth stage. Thereby improving the accuracy of the clustering algorithm and the analysis graph set generated by clustering. Among them, the abnormal analysis graph is an analysis graph whose diffraction information is incomplete or does not contain effective diffraction information.

[0036] Step 102, determining a feature vector of a historical analysis graph collected during the historical production process of the material.

[0037] In the present application, based on any preset feature vector extraction method, each historical analysis graph collected during the historical production process of the material can be vectorized to determine the feature vector corresponding to each historical analysis graph.

[0038] Optionally, the multi-channel data corresponding to multiple consecutive historical analysis graphs can be combined to generate a high-dimensional vector, and then any principal component extraction method can be used to extract the key vector in the high-dimensional vector. Then, the feature vector of the historical analysis graph is generated based on the key vector. In this way, the data is compressed into several new data that can effectively represent the original data of the historical analysis graph under the premise of minimal information loss. This is conducive to improving the efficiency of training clustering algorithms.

[0039] Step 103: Use the feature vector to train the initial clustering algorithm to obtain a final clustering algorithm, and divide the historical analysis graph into a number of analysis graph sets in the growth stage.

[0040] In the present application, the initial clustering algorithm is trained using the feature vectors corresponding to the historical analysis graphs until the final clustering algorithm is obtained. It is understandable that in the process of training the clustering algorithm, the clustering algorithm divides the historical analysis graphs into the number of analysis graph sets of the growth stages. Moreover, each analysis graph set corresponds to the growth stage of a material.

[0041] The following describes a method for determining the material growth stage according to an embodiment of the present application with reference to the accompanying drawings.

[0042] Figure 2 A schematic flow chart of a method for determining a material growth stage provided in an embodiment of the present application.

[0043] like Figure 2 As shown, the method for determining the growth stage of the material includes:

[0044] Step 201, during the material production process, obtain an analysis graph of the material at the current moment, and determine a characteristic vector of the analysis graph.

[0045] The analysis diagram may include a diffraction diagram, a spectrum diagram, etc., which is not limited in this application.

[0046] In the present application, the method for determining the characteristic vector of the analysis graph is the same as the method for determining the characteristic vector of the historical analysis graph collected during the historical production process of the material in the present application.

[0047] Optionally, the multi-channel data corresponding to the analysis graph collected at the current moment and the multi-channel data corresponding to multiple analysis graphs collected before the current moment can be combined to generate a high-dimensional vector. Afterwards, any principal component extraction method can be used to extract the key vector in the high-dimensional vector. Then, the feature vector of the historical analysis graph is generated based on the key vector. Thus, the data is compressed into several new data that can effectively represent the original data of the analysis graph under the premise of minimal information loss. This is conducive to improving the efficiency of determining the growth stage of the material.

[0048] Step 202, input the feature vector into a pre-trained clustering algorithm, obtain the analysis graph set to which the feature vector output by the clustering algorithm belongs, and determine the annotated growth stage corresponding to the analysis graph set as the growth stage of the material at the current moment, wherein the clustering algorithm is obtained based on an unsupervised training method, the analysis graph set is obtained by classifying historical analysis graphs when training the clustering algorithm based on historical analysis graphs, and the annotated growth stage corresponding to the analysis graph set is determined based on the material growth state in each analysis graph contained in the analysis graph set.

[0049] The clustering algorithm may be obtained based on the clustering algorithm training method provided in the above embodiment, or may be obtained by training based on the historical analysis graph using other unsupervised learning training methods, which is not limited in the present application.

[0050] The historical analysis chart is a complete analysis chart collected during the historical growth process of the material, including each growth stage.

[0051] In the present application, when training the clustering algorithm, the clustering algorithm classifies the historical analysis graphs to obtain multiple analysis graph sets. Afterwards, the growth stage of the material corresponding to each analysis graph in each analysis graph set can be determined manually, and the growth stage of the material corresponding to each analysis graph in each analysis graph set can be statistically analyzed to determine the proportion of the analysis graphs corresponding to each growth stage in each analysis graph set. Then, based on the proportion of the analysis graphs corresponding to each growth stage in each analysis graph set, the labeled growth stage corresponding to each analysis graph set is determined. For example, the growth stage corresponding to the highest proportion analysis graph in each analysis graph set is determined as the labeled growth stage corresponding to each analysis graph set. Alternatively, when the difference between the highest proportion and other proportions in a certain analysis graph set is greater than a threshold value, the growth stage corresponding to the highest proportion analysis graph is determined as the labeled growth stage corresponding to the analysis graph set. The labeled growth stage corresponding to other analysis graph sets other than the analysis graph set is determined as a conversion stage. Thereby improving the accuracy of the labeled growth stage, and then helping to improve the accuracy of determining the growth stage of the material.

[0052] Optionally, when training the clustering algorithm, the clustering algorithm classifies the historical analysis graphs to obtain multiple analysis graph sets and the identification (such as number) of each analysis graph set. Afterwards, the labeled growth stage corresponding to each identification is manually determined to convert the identification of each analysis graph set into a labeled growth stage. Optionally, after determining the growth stage of the material at the current moment, when the growth stage is the specified growth stage, a prompt message for adjusting the growth parameters is generated to prompt relevant personnel to adjust the growth parameters in time to avoid material quality degradation. For example, in the epitaxial growth process of gallium arsenide GaAs thin film materials, the specified growth stage can be a conversion stage. When the growth stage is the conversion stage, a prompt message for adjusting the growth parameters is generated, and the prompt message is sent to the terminal device corresponding to the relevant personnel to prompt the relevant personnel to adjust the growth parameters in time.

[0053] In this application, in the material production process, the analysis graph of the material at the current moment is obtained, and the characteristic vector of the analysis graph is determined. After that, the characteristic vector is input into the pre-trained clustering algorithm, the analysis graph set to which the characteristic vector output by the clustering algorithm belongs is obtained, and the annotated growth stage corresponding to the analysis graph set is determined as the growth stage of the material at the current moment, wherein the clustering algorithm is obtained based on an unsupervised training method, the analysis graph set is obtained by classifying the historical analysis graph when training the clustering algorithm based on the historical analysis graph, and the annotated growth stage corresponding to the analysis graph set is determined based on the material growth state in each analysis graph contained in the analysis graph set. Thus, after obtaining the clustering algorithm and the analysis graph set through unsupervised training of the historical analysis graph, the annotated growth stage corresponding to the analysis graph set is determined, which ensures the accuracy of the annotated growth stage. Then, the analysis graph at the current moment is classified and processed by the clustering algorithm, which can capture the subtle differences and changing trends in the analysis graph that are difficult for the human eye to perceive, and provide more accurate judgments. Thereby improving the accuracy of determining the growth stage of the material.

[0054] Figure 3 A schematic flow chart of a method for determining a material growth stage provided in an embodiment of the present application.

[0055] like Figure 3 As shown, the method for determining the growth stage of the material includes:

[0056] Step 301, during the material production process, obtain an analysis diagram of the material at the current moment.

[0057] In the present application, the specific implementation process of step 301 can be found in the detailed description of any embodiment of the present application, and will not be repeated here.

[0058] Step 302, determining whether the analysis graph is an abnormal analysis graph.

[0059] In the present application, when there is an obvious change in the boundary between light and dark in the analysis graph, the analysis graph can be determined to be an abnormal analysis graph. When all pixels in the analysis graph are the same, it means that the analysis graph does not contain useful information (diffraction fringes), and the analysis graph can be determined to be an abnormal analysis graph. When the morphology of the diffraction fringes in the analysis graph meets the preset conditions, the analysis graph can be determined to be an abnormal analysis graph. Among them, the morphology includes diffraction shape, brightness, distribution, etc. The preset condition is that the diffraction spot or shape does not show periodic changes, or the average brightness of the diffraction spot does not change. . Optionally, when the diffraction imaging of the material in the analysis graph is spot-like, and the distribution state of the diffraction spots in the analysis graph meets the preset conditions, the analysis graph can be determined to be an abnormal analysis graph. Among them, the preset condition is the distribution state of the diffraction spots collected when the imaging fluorescent screen is partially blocked.

[0060] Optionally, when the pixels in the analysis graph are not completely the same and the discontinuous state of the diffraction fringes in the analysis graph does not meet the preset conditions, it means that the analysis graph contains valid information and the information is complete. Therefore, it can be determined that the analysis graph is not an abnormal analysis graph.

[0061] Optionally, a discriminant model for determining whether an analysis graph is an abnormal analysis graph can be obtained in advance based on sample set training. Afterwards, the discriminant model is used to determine whether each analysis graph obtained during the material production process is an abnormal analysis graph. The sample set includes multiple historical analysis graphs and a label corresponding to each historical analysis graph indicating whether it is an abnormal analysis graph.

[0062] Step 303: if the analysis chart is an abnormal analysis chart, generate an abnormal prompt.

[0063] In this application, the growth stage of the material cannot be determined based on an analysis diagram that does not contain valid information. Determining the growth stage of the material based on an analysis diagram with incomplete information may lead to inaccurate determination of the growth stage of the material. Therefore, in the case where the analysis diagram is an abnormal analysis diagram, an abnormal prompt can be generated to prompt the operation and maintenance personnel to check each device in the material production environment, promptly repair the fault that causes the abnormal analysis diagram, and ensure the validity of the collected analysis diagram. Thereby improving the accuracy of determining the growth stage of the material.

[0064] Optionally, if the analysis graph at the current moment is an abnormal analysis graph, the processing of the analysis graph may be terminated. Afterwards, when the analysis graph at the next moment is collected, the growth stage of the material at the next moment is determined based on the analysis graph at the next moment, thereby improving the efficiency of determining the growth stage of the material.

[0065] Step 304: if the analysis graph is not an abnormal analysis graph, determine a characteristic vector of the analysis graph.

[0066] Step 305, input the feature vector into a pre-trained clustering algorithm, obtain the analysis graph set to which the feature vector output by the clustering algorithm belongs, and determine the annotated growth stage corresponding to the analysis graph set as the growth stage of the material at the current moment, wherein the clustering algorithm is obtained based on an unsupervised training method, the analysis graph set is obtained by classifying historical analysis graphs when training the clustering algorithm based on historical analysis graphs, and the annotated growth stage corresponding to the analysis graph set is determined based on the material growth state in each analysis graph contained in the analysis graph set.

[0067] In the present application, the specific implementation process of step 304-step 305 can be found in the detailed description of any embodiment of the present application, and will not be repeated here.

[0068] In this application, during the material production process, the analysis graph of the material at the current moment is obtained, and it is determined whether the analysis graph is an abnormal analysis graph. After that, if the analysis graph is an abnormal analysis graph, an abnormal prompt is generated. If the analysis graph is not an abnormal analysis graph, the characteristic vector of the analysis graph is determined, and the characteristic vector is input into a pre-trained clustering algorithm, and the analysis graph set to which the characteristic vector output by the clustering algorithm belongs is obtained, and the annotated growth stage corresponding to the analysis graph set is determined as the growth stage of the material at the current moment. This improves the accuracy and efficiency of determining the growth stage of the material.

[0069] In order to implement the above embodiment, the embodiment of the present application also proposes a device for determining the growth stage of a material. Figure 4 A schematic diagram of the structure of a device for determining the growth stage of a material provided in an embodiment of the present application.

[0070] like Figure 4 As shown, the device 400 for determining the material growth stage includes:

[0071] The acquisition module 410 is used to acquire the analysis graph of the material at the current moment during the material production process and determine the characteristic vector of the analysis graph;

[0072] Determination module 420 is used to input the feature vector into a pre-trained clustering algorithm, obtain the analysis graph set to which the feature vector output by the clustering algorithm belongs, and determine the annotated growth stage corresponding to the analysis graph set as the growth stage of the material at the current moment; wherein, the clustering algorithm is obtained based on an unsupervised training method, the analysis graph set is obtained by classifying historical analysis graphs when training the clustering algorithm based on historical analysis graphs, and the annotated growth stage corresponding to the analysis graph set is determined based on the material growth state in each analysis graph contained in the analysis graph set.

[0073] In a possible implementation manner of another aspect of the present application, a training module is further included, which is used to:

[0074] Determine the growth stages included in the complete growth process of the material, and set the number of clusters corresponding to the initial clustering algorithm based on the number of growth stages, the growth stages include a transition stage and a stable growth stage;

[0075] Determine the characteristic vector of the historical analysis graph collected during the historical production process of the material;

[0076] The initial clustering algorithm is trained using the feature vectors to obtain the final clustering algorithm, and the historical analysis graph is divided into a number of analysis graph sets in the growth stage.

[0077] In a possible implementation of another aspect of the present application, the training module is used to:

[0078] Get the type of abnormal analysis graph;

[0079] The sum of the number of growth stages and the number of types is set to the number of clusters corresponding to the initial clustering algorithm.

[0080] In a possible implementation manner of another aspect of the present application, a preprocessing module is further included, which is used to:

[0081] Determine whether the analysis chart is an abnormal analysis chart;

[0082] When the analysis graph is an abnormal analysis graph, the processing of the analysis graph is terminated.

[0083] In a possible implementation of another aspect of the present application, the preprocessing module is used to:

[0084] When all pixels in the analysis graph are the same, or the discontinuous state of the diffraction fringes in the analysis graph meets the preset conditions, the analysis graph is determined to be an abnormal analysis graph;

[0085] When the pixels in the analysis graph are not completely the same and the discontinuous state of the diffraction fringes in the analysis graph does not meet the preset conditions, it is determined that the analysis graph is not an abnormal analysis graph.

[0086] In a possible implementation of another aspect of the present application, the acquisition module 410 is used to:

[0087] The analysis graph is divided into a plurality of sub-analysis graphs, and the sub-analysis graphs are linearly combined to obtain a plurality of orthogonal vectors;

[0088] determining a key vector among the plurality of orthogonal vectors based on the variance between each of the orthogonal vectors;

[0089] Generate feature vectors of the analysis graph based on the key vectors.

[0090] It should be noted that the above explanation of the embodiment of the method for determining the material growth stage is also applicable to the device for determining the material growth stage of this embodiment, so it will not be repeated here.

[0091] In this application, in the material production process, the analysis graph of the material at the current moment is obtained, and the characteristic vector of the analysis graph is determined. After that, the characteristic vector is input into the pre-trained clustering algorithm, the analysis graph set to which the characteristic vector output by the clustering algorithm belongs is obtained, and the annotated growth stage corresponding to the analysis graph set is determined as the growth stage of the material at the current moment, wherein the clustering algorithm is obtained based on an unsupervised training method, the analysis graph set is obtained by classifying the historical analysis graph when training the clustering algorithm based on the historical analysis graph, and the annotated growth stage corresponding to the analysis graph set is determined based on the material growth state in each analysis graph contained in the analysis graph set. Thus, after obtaining the clustering algorithm and the analysis graph set through unsupervised training of the historical analysis graph, the annotated growth stage corresponding to the analysis graph set is determined, which ensures the accuracy of the annotated growth stage. Then, the analysis graph at the current moment is classified and processed by the clustering algorithm, which can capture the subtle differences and changing trends in the analysis graph that are difficult for the human eye to perceive, and provide more accurate judgments. Thereby improving the accuracy of determining the growth stage of the material.

[0092] In order to implement the above embodiment, the embodiment of the present application also proposes a computer device, including a processor and a memory;

[0093] The processor runs a program corresponding to the executable program code by reading the executable program code stored in the memory, so as to implement the method for determining the material growth stage as described in the above embodiment.

[0094] In order to implement the above embodiments, the embodiments of the present application further provide a computer-readable storage medium on which a computer program is stored. When the program is executed by a processor, the method for determining the material growth stage as described in the above embodiments is implemented.

[0095] Although the embodiments of the present application have been shown and described above, it can be understood that the above embodiments are exemplary and cannot be understood as limitations on the present application. Ordinary technicians in this field can change, modify, replace and modify the above embodiments within the scope of the present application.

Claims

1. A method for determining a material growth stage, characterized in that: include: During the material production process, obtaining an analysis graph of the material at a current moment, and determining a characteristic vector of the analysis graph; Input the feature vector into a pre-trained clustering algorithm, obtain the analysis graph set to which the feature vector output by the clustering algorithm belongs, and determine the annotated growth stage corresponding to the analysis graph set as the growth stage of the material at the current moment; Wherein, the clustering algorithm is obtained based on an unsupervised training method, the analysis graph set is obtained by classifying the historical analysis graphs when training the clustering algorithm based on the historical analysis graphs, and the annotated growth stage corresponding to the analysis graph set is determined based on the material growth state in each analysis graph contained in the analysis graph set; The determining of the characteristic vector of the analysis graph comprises: Combining the analysis graph and multi-channel data corresponding to multiple analysis graphs collected before the analysis graph to generate a high-dimensional vector; Extracting a key vector from the high-dimensional vector; A feature vector of the analysis graph is generated based on the key vector.

2. The method according to claim 1, characterized in that The training process of the clustering algorithm is as follows: Determining the growth stages included in the complete growth process of the material, and setting the number of clusters corresponding to the initial clustering algorithm based on the number of the growth stages, wherein the growth stages include a transition stage and a stable growth stage; Determine a feature vector of a historical analysis graph collected during the historical production process of the material; The initial clustering algorithm is trained using the feature vector to obtain a final clustering algorithm, and the historical analysis graph is divided into a number of analysis graph sets in the growth stage.

3. The method according to claim 1, characterized in that The step of setting the number of clusters corresponding to the initial clustering algorithm based on the number of growth stages includes: Get the type of abnormal analysis graph; The sum of the number of the growth stages and the number of the types is set as the number of clusters corresponding to the initial clustering algorithm.

4. The method according to claim 1, characterized in that Before inputting the feature vector into a pre-trained clustering algorithm and obtaining the analysis graph set to which the feature vector output by the clustering algorithm belongs, the method further includes: Determining whether the analysis graph is an abnormal analysis graph; When the analysis chart is an abnormal analysis chart, an abnormal prompt is generated.

5. The method according to claim 4, characterized in that The determining whether the analysis graph is an abnormal analysis graph comprises: When the shape of the diffraction fringes in the analysis graph meets a preset condition, determining that the analysis graph is an abnormal analysis graph; When the shape of the diffraction fringes in the analysis graph does not satisfy a preset condition, it is determined that the analysis graph is not an abnormal analysis graph.

6. A device for determining a material growth stage, characterized in that: include: An acquisition module, used to acquire the analysis graph of the material at the current moment during the material production process, and determine the characteristic vector of the analysis graph; A determination module, used for inputting the feature vector into a pre-trained clustering algorithm, obtaining the analysis graph set to which the feature vector output by the clustering algorithm belongs, and determining the annotated growth stage corresponding to the analysis graph set as the growth stage of the material at the current moment; Wherein, the clustering algorithm is obtained based on an unsupervised training method, the analysis graph set is obtained by classifying the historical analysis graphs when training the clustering algorithm based on the historical analysis graphs, and the annotated growth stage corresponding to the analysis graph set is determined based on the material growth state in each analysis graph contained in the analysis graph set; The acquisition module is further used to combine the analysis graph and multi-channel data corresponding to multiple analysis graphs collected before the analysis graph to generate a high-dimensional vector; Extracting a key vector from the high-dimensional vector; A feature vector of the analysis graph is generated based on the key vector.

7. The device according to claim 6, characterized in that Also includes training modules for: Determining the growth stages included in the complete growth process of the material, and setting the number of clusters corresponding to the initial clustering algorithm based on the number of the growth stages, wherein the growth stages include a transition stage and a stable growth stage; Determine a feature vector of a historical analysis graph collected during the historical production process of the material; The initial clustering algorithm is trained using the feature vector to obtain a final clustering algorithm, and the historical analysis graph is divided into a number of analysis graph sets in the growth stage.

8. A computer device, characterized in that: including a processor and a memory; The processor runs a program corresponding to the executable program code by reading the executable program code stored in the memory, so as to implement the method according to any one of claims 1 to 5.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the program is executed by a processor, the method according to any one of claims 1 to 5 is implemented.

Images