Disconnection control method and device, electronic equipment and storage medium

Abstract

Embodiments of the present application provide a wire break control method, device, equipment and medium. The method comprises: in the present Czochralski process, acquiring characteristic data, wherein the characteristic data comprises control data and monitoring data, inputting the characteristic data into a wire break prediction model, wherein the wire break prediction model is trained by characteristic data samples and sample wire break results of the shoulder stage corresponding to the labels; wherein the sample wire break results comprise wire break and no wire break, generating a predicted wire break rate of the shoulder stage from the wire break prediction model according to the characteristic data, and adjusting process parameters in the case that the predicted wire break rate reaches a preset threshold, so as to control no wire break in the shoulder stage, so that the wire break rate of the shoulder stage can be accurately predicted in advance, and the process parameters are adjusted in time accordingly, avoiding the problems of inaccurate and untimely wire break judgment, and reducing the occurrence rate of wire break.

Description

Technical Field

[0001] This invention relates to the field of crystal preparation technology, and in particular to a wire breakage control method, a wire breakage control device, an electronic device, and a storage medium. Background Technology

[0002] The main process for preparing monocrystalline silicon materials is the Czochralski process (CZ), which uses the Czochralski method to refine polycrystalline silicon raw materials into monocrystalline silicon. The process of generating rod-shaped monocrystalline silicon crystals during the Czochralski single crystal growth process includes steps such as loading, heating the molten material, temperature control, crystal pulling, shoulder formation, shoulder rotation, equal diameter shaping, and finishing.

[0003] After the polycrystalline silicon raw material has melted, crystal pulling cannot begin immediately because the temperature is higher than the crystal pulling temperature. It must be cooled down to the appropriate temperature. Crystal pulling involves placing a seed crystal (a single crystal shaped to a specific form) pre-attached to the end of a steel wire rope into contact with the liquid surface. At the crystal pulling temperature, silicon molecules grow along the lattice direction of the seed crystal, forming a single crystal. Shoulder formation involves gradually growing the crystal to the required diameter. During shoulder formation, a section of crystal is pulled out, gradually increasing in length and diameter to approximately the required diameter, to eliminate dislocations. Once the crystal has grown to the required diameter during shoulder formation, it enters the shoulder turning process.

[0004] In the production of monocrystalline silicon rods, current technologies mostly rely on vision systems to determine whether a wire is broken during the shoulder-setting stage, or even rely on operators to visually assess the situation. Both methods are reactive and not timely enough, and the visual method has a high error rate. Summary of the Invention

[0005] In view of the above problems, embodiments of the present invention are proposed to provide a disconnection control method that overcomes or at least partially solves the above problems, so as to solve the problems of inaccurate and untimely disconnection judgment.

[0006] Accordingly, embodiments of the present invention also provide a disconnection control device, an electronic device, and a storage medium to ensure the implementation and application of the above method.

[0007] To address the above problems, this invention discloses a disconnection control method, comprising:

[0008] During this Czochralski single crystal growth process, characteristic data was acquired, including control data and monitoring data.

[0009] The feature data is input into the line break prediction model, wherein the line break prediction model is trained using feature data samples and corresponding labeled line break results of the shoulder release stage; wherein the line break results include broken lines and unbroken lines.

[0010] Based on the feature data, the predicted breakage rate for the shoulder-laying stage is generated by the breakage prediction model;

[0011] If the predicted breakage rate reaches a preset threshold, the process parameters are adjusted to prevent breakage during the shoulder-setting stage.

[0012] Optionally, the feature data sample is the feature data sample before the shoulder formation stage, and the acquisition of feature data during this Czochralski single crystal pulling process includes:

[0013] When the Czochralski single crystal process enters the shoulder formation stage, characteristic data prior to the shoulder formation stage are acquired.

[0014] Optionally, the feature data includes statistical feature data, and acquiring the feature data during this Czochralski single crystal growth process further includes:

[0015] Based on the feature data prior to the shoulder-releasing stage, corresponding statistical feature data is generated;

[0016] Add the statistical feature data to the feature data.

[0017] Optionally, before inputting the feature data into the line break prediction model, the method further includes:

[0018] After training the line breakage prediction model using feature data samples and the line breakage results of the corresponding labeled shoulder release stage samples, the feature weights corresponding to various feature data in the trained line breakage prediction model are obtained.

[0019] The first preset number of important feature data are determined by sorting the feature weights from high to low, and the important feature data are feature data corresponding to controllable process parameters;

[0020] When the predicted wire breakage rate reaches a preset threshold, adjusting the process parameters to prevent wire breakage during the shoulder-setting stage includes:

[0021] The process parameters corresponding to the important feature data are adjusted.

[0022] Optionally, adjusting the process parameters corresponding to the important feature data includes:

[0023] For the process parameters corresponding to various important feature data, determine multiple candidate values ​​of the process parameters within the corresponding value range;

[0024] Based on multiple candidate values ​​of the various process parameters, generate a combination of candidate values ​​for the various process parameters;

[0025] Based on each of the candidate value combinations, the line break prediction model is used to generate the corresponding candidate predicted line break rate.

[0026] The process parameters are adjusted based on the second preset number of candidate value combinations that have the smallest predicted breakage rate.

[0027] Optionally, acquiring feature data during this Czochralski single crystal growth process includes:

[0028] During the shoulder-releasing phase, the control data and monitoring data are collected once every first set time interval;

[0029] The control data and monitoring data collected consecutively for a preset number of times are determined as the feature data.

[0030] Optionally, before inputting the feature data into the line break prediction model, the method further includes:

[0031] Acquire feature data samples and corresponding marked sample disconnection results for the shoulder release stage; wherein, the feature data samples include control data samples and monitoring data samples collected consecutively for a preset number of times, and the sample disconnection results are the disconnection results after the last collection at a second preset time interval from the control data samples and monitoring data samples collected consecutively for a preset number of times;

[0032] Input the feature data samples and the corresponding labeled sample disconnection results into the disconnection prediction model;

[0033] The disconnection prediction model is trained using the disconnection results of the sample labeled with the feature data sample until the loss value of the disconnection prediction model is less than a set loss value, thus obtaining the trained disconnection prediction model.

[0034] This invention also discloses a wire breakage control device, comprising:

[0035] The data acquisition module is used to acquire characteristic data during this Czochralski single crystal growth process, wherein the characteristic data includes control data and monitoring data;

[0036] The data input module is used to input the feature data into the line break prediction model, wherein the line break prediction model is trained by feature data samples and line break results of samples in the corresponding labeled shoulder release stage; wherein the sample line break results include broken lines and unbroken lines.

[0037] The line breakage rate generation module is used to generate the predicted line breakage rate of the shoulder release stage based on the feature data and the line breakage prediction model.

[0038] The parameter adjustment module is used to adjust the process parameters when the predicted breakage rate reaches a preset threshold, so as to control the breakage from occurring during the shoulder setting stage.

[0039] Optionally, the feature data sample is a feature data sample before the shoulder formation stage, and the data acquisition module is specifically used to acquire the feature data before the shoulder formation stage when the current Czochralski single crystal process enters the shoulder formation stage.

[0040] Optionally, the feature data includes statistical feature data, and the data acquisition module further includes:

[0041] The data generation submodule is used to generate corresponding statistical feature data based on the feature data prior to the shoulder release stage.

[0042] The data addition submodule is used to add the statistical feature data to the feature data.

[0043] Optionally, the device further includes:

[0044] The weight acquisition module is used to acquire the feature weights corresponding to various feature data in the trained disconnection prediction model before inputting the feature data into the disconnection prediction model, after training the disconnection prediction model with feature data samples and disconnection results of the corresponding labeled shoulder stage samples.

[0045] The important feature determination module is used to determine the first preset number of important feature data that are ranked first from high to low according to the feature weights, and the important feature data are feature data corresponding to controllable process parameters;

[0046] The parameter adjustment module includes:

[0047] The parameter adjustment submodule is used to adjust the process parameters corresponding to the important feature data.

[0048] Optionally, the parameter adjustment submodule includes:

[0049] The candidate value determination unit is used to determine multiple candidate values ​​of the process parameters within the corresponding value range for each of the various important feature data.

[0050] A combination generation unit is used to generate a combination of candidate values ​​for multiple process parameters based on multiple candidate values ​​of multiple process parameters.

[0051] The disconnection rate generation unit is used to generate corresponding candidate predicted disconnection rates based on each of the candidate value combinations and the disconnection prediction model.

[0052] The parameter adjustment unit is used to adjust the process parameters based on the second preset number of candidate value combinations with the smallest predicted breakage rate.

[0053] Optionally, the data acquisition module includes:

[0054] The data acquisition submodule is used to acquire the control data and monitoring data once every first set time interval during the shoulder release stage.

[0055] The data determination submodule is used to determine the control data and monitoring data collected continuously for a preset number of times as the feature data.

[0056] Optionally, the device further includes:

[0057] The sample acquisition module is used to acquire feature data samples and corresponding marked sample disconnection results of the shoulder release stage before inputting the feature data into the disconnection prediction model; wherein, the feature data samples include control data samples and monitoring data samples collected continuously for a preset number of times, and the sample disconnection results are the disconnection results of the control data samples and monitoring data samples collected continuously for a preset number of times, after the last collection at an interval of a second preset time.

[0058] The sample input module is used to input the feature data sample and the corresponding labeled sample disconnection result into the disconnection prediction model;

[0059] The model training module is used to train the disconnection prediction model using the disconnection results of the sample labeled with the feature data sample until the loss value of the disconnection prediction model is less than a set loss value, thus obtaining the trained disconnection prediction model.

[0060] This invention also discloses an electronic device, characterized in that it includes a processor, a communication interface, a memory, and a communication bus, wherein the processor, the communication interface, and the memory communicate with each other through the communication bus;

[0061] Memory, used to store computer programs;

[0062] When a processor executes a program stored in memory, it implements the method steps described above.

[0063] This invention also discloses a readable storage medium, which, when the instructions in the storage medium are executed by the processor of an electronic device, enables the electronic device to execute one or more of the disconnection control methods described in this invention.

[0064] The embodiments of the present invention have the following advantages:

[0065] According to an embodiment of the present invention, feature data, including control data and monitoring data, is acquired during the Czochralski single crystal growth process. This feature data is then input into a wire breakage prediction model, which is trained using feature data samples and corresponding marked wire breakage results from the shoulder-growing stage. The sample wire breakage results include both broken and unbroken wires. Based on the feature data, the wire breakage prediction model generates a predicted wire breakage rate for the shoulder-growing stage. When the predicted wire breakage rate reaches a preset threshold, process parameters are adjusted to prevent wire breakage during the shoulder-growing stage. This allows for accurate prediction of the wire breakage rate in advance, enabling timely adjustment of process parameters and avoiding inaccurate or untimely wire breakage assessments, thus reducing the incidence of wire breakage. Attached Figure Description

[0066] Figure 1 This is a flowchart illustrating the steps of an embodiment of the disconnection control method of the present invention;

[0067] Figure 2 This is a flowchart illustrating the steps of an embodiment of the disconnection control method of the present invention;

[0068] Figure 3 This is a flowchart illustrating the steps of an embodiment of the disconnection control method of the present invention;

[0069] Figure 4 This is a structural block diagram of an embodiment of the wire breakage control device of the present invention;

[0070] Figure 5 This is a structural block diagram of a computing device for disconnection control according to an exemplary embodiment. Detailed Implementation

[0071] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments.

[0072] Reference Figure 1 The flowchart illustrates an embodiment of the disconnection control method of the present invention, which may specifically include the following steps:

[0073] Step 101: During this Czochralski single crystal growth process, characteristic data is acquired, including control data and monitoring data.

[0074] In this embodiment of the invention, the Czochralski single crystal process is the process of refining raw materials into single crystals using the Czochralski method, for example, the process of Czochralski single crystal silicon. The Czochralski single crystal process can be divided into a pre-temperature conditioning stage, a temperature conditioning stage, a crystal pulling stage, and a shoulder formation stage.

[0075] In this embodiment of the invention, various characteristic data can be acquired during the shoulder formation stage of the Czochralski single crystal growth process. These characteristic data include control data and monitoring data. The control data consists of control parameters input to the Czochralski single crystal growth equipment, while the monitoring data is data monitored during actual operation. Specifically, it can include any applicable characteristic data, and this embodiment of the invention does not impose any limitations on this.

[0076] Monitoring data can be obtained through industrial sensors. Control data can be directly collected from relevant control equipment. For example, core features may include: argon flow rate, crystal length, crucible rotation, main heater current, meltsurftemp (liquid surface brightness), meltlevel (liquid outlet distance), crystalpos (crystal position), seedrotation (crystal rotation speed), cruciblepos (boiler position), average pulling speed, diameter (crystal diameter), main heater power, and remaining weight.

[0077] In this embodiment of the invention, the directly acquired control data samples and monitoring data samples are filtered to obtain feature data samples; wherein, the filtering process includes at least one of the following: text type data removal, redundant data removal, abnormal data removal based on box plot method, and feature filtering based on correlation analysis.

[0078] Many directly obtained control and monitoring data samples are unusable. Therefore, screening and processing of these samples are necessary. Text data removal involves eliminating text-based data. Redundant data removal involves removing duplicate data. Outlier removal based on box plots involves observing the overall data distribution using box plots. Statistics such as the median, 25th percentile, 75th percentile, upper boundary, and lower boundary are used to describe the overall distribution. A box plot is generated by calculating these statistics; the box contains most of the normal data, while data outside the upper and lower boundaries are outliers and are removed. Feature screening based on correlation analysis involves analyzing the correlation between various control and monitoring data samples. Only one strongly correlated data point is retained among the control and monitoring data samples.

[0079] In this embodiment of the invention, after filtering the feature data, statistical feature data and logarithmically transformed feature data can also be generated.

[0080] In an optional embodiment of the present invention, the feature data includes statistical feature data. Acquiring the feature data during this Czochralski single crystal growth process further includes: generating corresponding statistical feature data based on the feature data prior to the shoulder formation stage; and adding the statistical feature data to the feature data.

[0081] Statistical feature data refers to feature data obtained by statistically analyzing directly acquired feature data. For example, the minimum, maximum, mean, and variance of pulling speed are all statistical feature data corresponding to pulling speed. Specifically, any applicable statistical feature data can be set according to actual needs; this embodiment of the invention does not impose any restrictions on this.

[0082] Both directly acquired feature data and statistical feature data are used as input feature data for the line break prediction model. Correspondingly, statistical feature data samples are also needed during the training of the line break prediction model. Both directly acquired feature data and statistical feature data are used as input feature data samples for the line break prediction model.

[0083] In an optional embodiment of the present invention, the feature data includes logarithmically transformed feature data. One implementation of obtaining feature data may further include: performing a logarithmic transformation on the multiple directly obtained control data and monitoring data to obtain the logarithmically transformed feature data, and using the directly obtained control data and monitoring data, and the logarithmically transformed feature data, as feature data input to the line break prediction model.

[0084] In the Czochralski single crystal growth process, some characteristic data are directly obtainable control and monitoring data. Additionally, logarithmic transformation can be used to obtain some characteristic data that cannot be directly obtained. Logarithmic transformation is the operation of taking the logarithm of control and monitoring data, thus converting them into logarithmic data. Since the logarithmic function is a monotonically increasing function within its domain, taking the logarithm does not change the relative relationships of the data. Logarithmic transformation can compress the range of large values ​​and expand the range of small values. It can also mitigate the effects of data distribution skew, allowing the line breakage prediction model to more accurately generate predicted line breakage rates, thereby improving the model's accuracy.

[0085] The directly acquired control data and monitoring data, along with the logarithmically transformed feature data, are all used as input feature data for the line break prediction model. Correspondingly, during the training of the line break prediction model, multiple directly acquired control data samples and monitoring data samples also need to be logarithmically transformed to obtain logarithmically transformed feature data samples. These directly acquired control data samples, monitoring data samples, and logarithmically transformed feature data samples are then used as input feature data samples for the line break prediction model.

[0086] In this embodiment of the invention, before inputting the aforementioned feature data into the line break prediction model, analysis of variance (ANOVA) can be used for further screening. ANOVA determines the influence of controllable factors on the research results by analyzing the contribution of variations from different sources to the total variation. Specifically, the variance of each feature data point can be calculated first, and then feature data with variances greater than a threshold can be selected based on a threshold, thus filtering out feature data that have a greater impact on the results.

[0087] Step 102: Input the feature data into the line break prediction model, wherein the line break prediction model is trained by feature data samples and the line break results of the corresponding labeled shoulder release stage samples; wherein the line break results include broken lines and unbroken lines.

[0088] In this embodiment of the invention, the breakage result is divided into broken line and no broken line. The breakage rate during the shoulder-laying stage can be predicted using machine learning. Based on the correlation between the feature data and the breakage result during the shoulder-laying stage, a breakage prediction model that can predict the breakage rate can be obtained.

[0089] To train the breakage prediction model, accurate sample data and corresponding labeled data are needed. This includes feature data samples acquired during and / or before the shoulder formation stage, and the corresponding labeled breakage results from the shoulder formation stage. Specifically, feature data samples and breakage results can be obtained through multiple experiments, or by selecting feature data samples and breakage results from historical data. The breakage results are obtained by observing the breakage patterns of the crystal.

[0090] In this embodiment of the invention, the disconnection prediction model can adopt a categorical gradient boosting (CATboost) model, a long short-term memory (LSTM) model, or any applicable model. This embodiment of the invention does not limit the specific model used.

[0091] In this embodiment of the invention, during the prediction phase, the input to the breakage prediction model is feature data acquired during or before the shoulder-laying phase. Each time feature data is acquired, it is input into the breakage prediction model. The model outputs a predicted breakage rate for the shoulder-laying phase, denoted as the predicted breakage rate.

[0092] Step 103: Based on the feature data, generate the predicted breakage rate for the shoulder release stage using the breakage prediction model.

[0093] In this embodiment of the invention, the line breakage prediction model makes predictions based on the input feature data to generate a predicted line breakage rate. The output of the line breakage prediction model is obtained and then used as the predicted line breakage rate for the shoulder release stage.

[0094] Step 104: If the predicted breakage rate reaches a preset threshold, adjust the process parameters to control the breakage from occurring during the shoulder-laying stage.

[0095] In this embodiment of the invention, if the predicted breakage rate reaches a preset threshold, it indicates that there is a high probability of breakage during the shoulder-laying stage. Therefore, the process parameters need to be adjusted to prevent breakage during the shoulder-laying stage. If the predicted breakage rate does not reach the preset threshold, it indicates that there is a high probability that breakage will not occur during the shoulder-laying stage. Therefore, there is no need to adjust the process parameters for this reason.

[0096] The process parameters are process-related parameters set for the equipment, including power, average crucible rotation, etc., or any other applicable process parameters. This embodiment of the invention does not limit these parameters.

[0097] According to an embodiment of the present invention, feature data, including control data and monitoring data, is acquired during the Czochralski single crystal growth process. This feature data is then input into a wire breakage prediction model, which is trained using feature data samples and corresponding marked wire breakage results from the shoulder-growing stage. The sample wire breakage results include both broken and unbroken wires. Based on the feature data, the wire breakage prediction model generates a predicted wire breakage rate for the shoulder-growing stage. When the predicted wire breakage rate reaches a preset threshold, process parameters are adjusted to prevent wire breakage during the shoulder-growing stage. This allows for accurate prediction of the wire breakage rate in advance, enabling timely adjustment of process parameters and avoiding inaccurate or untimely wire breakage assessments, thus reducing the incidence of wire breakage.

[0098] In an optional embodiment of the present invention, before inputting the feature data into the breakage prediction model, the method may further include: after training the breakage prediction model using feature data samples and corresponding labeled breakage results from the shoulder-laying stage, obtaining feature weights corresponding to various feature data in the trained breakage prediction model. A first preset number of important feature data points are determined, ranked from highest to lowest by the feature weights, and these important feature data points are feature data corresponding to controllable process parameters. When the predicted breakage rate reaches a preset threshold, the process parameters are adjusted to prevent breakage during the shoulder-laying stage, including adjusting the process parameters corresponding to the important feature data.

[0099] The training process of a broken line prediction model involves determining the feature weights corresponding to various feature data based on the training data. After the broken line prediction model is trained, the feature weights are determined. The feature weights corresponding to various feature data are then extracted from the trained broken line prediction model. The larger the feature weight, the greater the impact of that feature data on broken lines; conversely, the smaller the feature weight, the smaller the impact of that feature data on broken lines.

[0100] Since adjusting the process parameters corresponding to feature data with higher feature weights will have a more significant impact on line breakage, the feature data are sorted from highest to lowest feature weight. Several top-ranked feature data are then selected and designated as important feature data. Specifically, a first preset number of important feature data can be selected. This first preset number can be set according to actual needs, and this embodiment of the invention does not impose any limitations on it. Furthermore, the important feature data are feature data corresponding to controllable process parameters.

[0101] In an optional embodiment of the present invention, a specific implementation of adjusting the process parameters corresponding to the important feature data may include: determining multiple candidate values ​​for the process parameters corresponding to various important feature data within their respective value ranges; generating multiple combinations of candidate values ​​for the process parameters based on the multiple candidate values; generating corresponding candidate predicted disconnection rates using the disconnection prediction model based on each candidate value combination; and adjusting the process parameters based on a second preset number of candidate value combinations that have the lowest candidate predicted disconnection rates.

[0102] For the process parameters corresponding to various important feature data, the value range of the process parameter is first obtained. Then, multiple candidate values ​​are selected within the value range of each process parameter. The methods for determining multiple candidate values ​​can include various methods, such as selecting a value at preset intervals, selecting multiple commonly used values, or any other applicable methods. This embodiment of the invention does not limit these methods.

[0103] The methods for generating candidate value combinations of multiple process parameters based on multiple candidate values ​​can be varied. For example, each candidate value combination may contain only one process parameter, and the multiple candidate values ​​for each process parameter constitute one candidate value combination. Alternatively, each candidate value combination may have multiple process parameters, and combining the multiple candidate values ​​of these multiple process parameters will yield a candidate value combination in each combination method. Another example is generating all combinations of all candidate values ​​for all process parameters. Specifically, one or more combination methods can be used, and this embodiment of the invention does not limit this. For example, multiple candidate value combinations can be generated based on the multiple candidate values ​​of multiple process parameters using a Cartesian product approach.

[0104] Each candidate value combination is input as feature data into the line break prediction model. Other feature data input into the model retains its original values. Based on the input data, the model generates a predicted line break rate, denoted as the candidate predicted line break rate. The candidate value combinations with the lowest predicted line break rates are selected; specifically, a second preset number of candidate value combinations can be selected. This second preset number can be set according to actual needs, and this embodiment of the invention does not impose any limitations on it. For example, the second preset number is 5.

[0105] The process parameters can be adjusted in various ways based on a second preset number of candidate value combinations. For example, the candidate value combination with the lowest predicted breakage rate can be directly used to adjust the process parameters to the value of that candidate value combination. Another example is providing the selected second preset number of candidate value combinations to the workers so that they can choose the most suitable candidate value combination to adjust the process parameters to that value. Any applicable method can be used, and this embodiment of the invention does not limit this.

[0106] Reference Figure 2 The flowchart illustrates an embodiment of the disconnection control method of the present invention, which may specifically include the following steps:

[0107] Step 201: When the current Czochralski single crystal process enters the shoulder formation stage, acquire the characteristic data before the shoulder formation stage.

[0108] In one specific implementation of this invention, the wire breakage prediction model is activated when the Czochralski single crystal pulling process enters the shoulder formation stage, predicting the wire breakage result for the entire shoulder formation stage. Therefore, the feature data is the feature data collected before the shoulder formation stage. Specifically, control data and monitoring data can be collected multiple times before the shoulder formation stage. For example, based on the understanding of the impact of wire breakage during the Czochralski single crystal pulling process, control data and monitoring data are mainly collected multiple times during the temperature adjustment stage and the crystal pulling stage.

[0109] In this embodiment of the invention, since the feature data used in the training process and the prediction process should be consistent, the feature data sample is also the feature data sample before the release stage, and its type and the time of collection are consistent with the type and time of collection of feature data in the prediction process.

[0110] For example, the Catboost model can be used for line break prediction. During model training, the feature data samples can be divided, with the training set accounting for 80% of the total and the test set accounting for 20%. The model is trained using the training set data and then tested using the test set data. When the accuracy index reaches 0.95 and the AUC (Area Under Curve) index reaches 0.86, it indicates that the line break prediction model performs well on the test set data, and the accuracy of the line break prediction model meets the requirements.

[0111] Step 202: Input the feature data into the line break prediction model, wherein the line break prediction model is trained by feature data samples and the line break results of the corresponding labeled shoulder release stage samples; wherein the line break results include broken lines and unbroken lines.

[0112] Step 203: Based on the feature data, generate the predicted breakage rate for the shoulder release stage using the breakage prediction model.

[0113] Step 204: If the predicted breakage rate reaches a preset threshold, adjust the process parameters to prevent breakage during the shoulder-laying stage.

[0114] According to an embodiment of the present invention, when the Czochralski single crystal process enters the shoulder-forming stage, feature data prior to the shoulder-forming stage is acquired, and the feature data is input into a wire breakage prediction model. The wire breakage prediction model is trained using feature data samples and corresponding labeled wire breakage results from the shoulder-forming stage. The sample wire breakage results include both broken and unbroken wires. Based on the feature data, the wire breakage prediction model generates a predicted wire breakage rate for the shoulder-forming stage. When the predicted wire breakage rate reaches a preset threshold, process parameters are adjusted to prevent wire breakage during the shoulder-forming stage. This allows for accurate early prediction of the wire breakage rate during the shoulder-forming stage, enabling timely adjustment of process parameters and avoiding inaccurate and untimely wire breakage detection, thus reducing the incidence of wire breakage problems.

[0115] Reference Figure 3 The flowchart illustrates an embodiment of the disconnection control method of the present invention, which may specifically include the following steps:

[0116] Step 301: During the shoulder-releasing stage, the control data and monitoring data are collected once every first set time interval.

[0117] In one specific implementation of this invention, the line breakage prediction model predicts the line breakage result after a second set time interval every first set time interval during the shoulder-laying phase. Therefore, the feature data is collected once for each first set time interval during the shoulder-laying phase, including control and monitoring data. The first set time interval can be set according to actual needs, and this embodiment of the invention does not limit it. For example, the first set time interval is 30 seconds. The second set time interval can be set based on the accuracy of the experimental results, and this embodiment of the invention does not limit it. For example, the second set time interval is 5 minutes.

[0118] Step 302: The control data and monitoring data collected consecutively for a preset number of times are determined as the feature data.

[0119] In this embodiment of the invention, the feature data input to the disconnection prediction model each time is data collected multiple times consecutively. The preset number of times can be set according to actual needs, and this embodiment of the invention does not impose any limitation on this. For example, the preset number of times is set to 5.

[0120] In an optional embodiment of the present invention, before inputting the feature data into the line break prediction model, the method may further include: acquiring feature data samples and corresponding marked line break results for the shoulder-release stage; wherein, the feature data samples include control data samples and monitoring data samples collected consecutively for a preset number of times, and the line break results are line break results collected consecutively for a preset number of times from the control data samples and monitoring data samples, with an interval of a second preset time after the last collection. The feature data samples and corresponding marked line break results are input into the line break prediction model. The line break prediction model is trained using the marked line break results of the feature data samples until the loss value of the line break prediction model is less than a set loss value, thus obtaining the trained line break prediction model.

[0121] The prediction process corresponds to the training process. The feature data samples are control data samples and monitoring data samples collected consecutively for a preset number of times. In order for the line break prediction model to predict the line break rate in advance, the sample line break results are the line break results collected consecutively for a preset number of times from the control data samples and monitoring data samples, with a second preset time interval after the last collection, so that the line break prediction model can predict the line break rate in advance at the second preset time interval.

[0122] When training the line break prediction model, the input consists of feature data samples and corresponding labeled line break results. The neural network of the line break prediction model is trained using the labeled line break results. That is, after each output of the predicted line break rate, the model compares it with the actual value (i.e., the labeled line break result, where 1 represents a broken line and 0 represents no broken line), and the comparison result is input into the loss function to calculate the loss value. The convergence condition of the model can be: the loss value is less than a set loss value, or the maximum number of iterations is reached. For example, a relatively small set loss value can be set, and the loss value is calculated during each training iteration. When the loss value is less than the set loss value, the model is considered to have converged, and training can end. A relatively large maximum number of iterations can be preset, such as 100 iterations, 10,000 iterations, or 1,000,000 iterations, etc., which needs to be selected according to the actual situation. This embodiment of the invention does not impose any restrictions on this. After the model completes the prescribed number of training iterations, the model is considered to have finished training.

[0123] During model training, the feature data samples can be divided, with the training set accounting for 80% and the test set accounting for 20%. For example, an LSTM model can be used to build a time series prediction model to predict line breakage rates. The model is trained using the training set data and then tested using the test set data. When the accuracy index reaches 0.94 and the AUC index reaches 0.8, it indicates that the line breakage prediction model performs well on the test set data, and the accuracy of the line breakage prediction model meets the requirements.

[0124] Step 303: Input the feature data into the line break prediction model, wherein the line break prediction model is trained by feature data samples and the line break results of the corresponding labeled shoulder release stage samples; wherein the line break results include broken lines and unbroken lines.

[0125] Step 304: Based on the feature data, generate the predicted breakage rate for the shoulder release stage using the breakage prediction model.

[0126] Step 305: If the predicted breakage rate reaches a preset threshold, adjust the process parameters to control the breakage from occurring during the shoulder-laying stage.

[0127] According to an embodiment of the present invention, during the shoulder-laying stage, control data and monitoring data are collected once every first predetermined time interval. The control data and monitoring data collected consecutively for a predetermined number of times are determined as the feature data. The feature data is input into a wire breakage prediction model, wherein the wire breakage prediction model is trained using feature data samples and corresponding marked sample wire breakage results of the shoulder-laying stage. The sample wire breakage results include wire breakage and non-wire breakage. Based on the feature data, the wire breakage prediction model generates a predicted wire breakage rate for the shoulder-laying stage. When the predicted wire breakage rate reaches a preset threshold, the process parameters are adjusted to control the wire breakage from occurring during the shoulder-laying stage. This allows for accurate prediction of the wire breakage rate during the shoulder-laying stage, enabling timely adjustment of process parameters and avoiding inaccurate and untimely wire breakage judgments, thereby reducing the incidence of wire breakage problems.

[0128] It should be noted that, for the sake of simplicity, the method embodiments are all described as a series of actions. However, those skilled in the art should understand that the embodiments of the present invention are not limited to the described order of actions, because according to the embodiments of the present invention, some steps can be performed in other orders or simultaneously. Furthermore, those skilled in the art should also understand that the embodiments described in the specification are preferred embodiments, and the actions involved are not necessarily essential to the embodiments of the present invention.

[0129] Reference Figure 4 The diagram shows a structural block diagram of an embodiment of the wire breakage control device of the present invention, which may specifically include the following modules:

[0130] The data acquisition module 401 is used to acquire characteristic data during this Czochralski single crystal growth process, wherein the characteristic data includes control data and monitoring data;

[0131] The data input module 402 is used to input the feature data into the line break prediction model, wherein the line break prediction model is trained by feature data samples and line break results of samples in the corresponding labeled shoulder release stage; wherein the sample line break results include broken lines and unbroken lines.

[0132] The line breakage rate generation module 403 is used to generate the predicted line breakage rate of the shoulder release stage by the line breakage prediction model based on the feature data.

[0133] The parameter adjustment module 404 is used to adjust the process parameters when the predicted breakage rate reaches a preset threshold, so as to control the breakage from occurring during the shoulder setting stage.

[0134] Optionally, the feature data sample is a feature data sample before the shoulder formation stage, and the data acquisition module is specifically used to acquire the feature data before the shoulder formation stage when the current Czochralski single crystal process enters the shoulder formation stage.

[0135] Optionally, the feature data includes statistical feature data, and the data acquisition module further includes:

[0136] The data generation submodule is used to generate corresponding statistical feature data based on the feature data prior to the shoulder release stage.

[0137] The data addition submodule is used to add the statistical feature data to the feature data.

[0138] Optionally, the device further includes:

[0139] The weight acquisition module is used to acquire the feature weights corresponding to various feature data in the trained disconnection prediction model before inputting the feature data into the disconnection prediction model, after training the disconnection prediction model with feature data samples and disconnection results of the corresponding labeled shoulder stage samples.

[0140] The important feature determination module is used to determine the first preset number of important feature data that are ranked first from high to low according to the feature weights, and the important feature data are feature data corresponding to controllable process parameters;

[0141] The parameter adjustment module includes:

[0142] The parameter adjustment submodule is used to adjust the process parameters corresponding to the important feature data.

[0143] Optionally, the parameter adjustment submodule includes:

[0144] The candidate value determination unit is used to determine multiple candidate values ​​of the process parameters within the corresponding value range for each of the various important feature data.

[0145] A combination generation unit is used to generate a combination of candidate values ​​for multiple process parameters based on multiple candidate values ​​of multiple process parameters.

[0146] The disconnection rate generation unit is used to generate corresponding candidate predicted disconnection rates based on each of the candidate value combinations and the disconnection prediction model.

[0147] The parameter adjustment unit is used to adjust the process parameters based on the second preset number of candidate value combinations with the smallest predicted breakage rate.

[0148] Optionally, the data acquisition module includes:

[0149] The data acquisition submodule is used to acquire the control data and monitoring data once every first set time interval during the shoulder release stage.

[0150] The data determination submodule is used to determine the control data and monitoring data collected continuously for a preset number of times as the feature data.

[0151] Optionally, the device further includes:

[0152] The sample acquisition module is used to acquire feature data samples and corresponding marked sample disconnection results of the shoulder release stage before inputting the feature data into the disconnection prediction model; wherein, the feature data samples include control data samples and monitoring data samples collected continuously for a preset number of times, and the sample disconnection results are the disconnection results of the control data samples and monitoring data samples collected continuously for a preset number of times, after the last collection at an interval of a second preset time.

[0153] The sample input module is used to input the feature data sample and the corresponding labeled sample disconnection result into the disconnection prediction model;

[0154] The model training module is used to train the disconnection prediction model using the disconnection results of the sample labeled with the feature data sample until the loss value of the disconnection prediction model is less than a set loss value, thus obtaining the trained disconnection prediction model.

[0155] According to an embodiment of the present invention, feature data, including control data and monitoring data, is acquired during the Czochralski single crystal growth process. This feature data is then input into a wire breakage prediction model, which is trained using feature data samples and corresponding marked wire breakage results from the shoulder-growing stage. The sample wire breakage results include both broken and unbroken wires. Based on the feature data, the wire breakage prediction model generates a predicted wire breakage rate for the shoulder-growing stage. When the predicted wire breakage rate reaches a preset threshold, process parameters are adjusted to prevent wire breakage during the shoulder-growing stage. This allows for accurate prediction of the wire breakage rate in advance, enabling timely adjustment of process parameters and avoiding inaccurate or untimely wire breakage assessments, thus reducing the incidence of wire breakage.

[0156] As the device embodiment is basically similar to the method embodiment, the description is relatively simple, and relevant parts can be found in the description of the method embodiment.

[0157] Figure 5 This is a structural block diagram illustrating an electronic device 700 for shoulder-turn activation according to an exemplary embodiment. For example, the electronic device 700 may be a mobile phone, computer, digital broadcasting terminal, messaging device, game console, tablet device, medical device, fitness equipment, personal digital assistant, etc.

[0158] Reference Figure 5 The electronic device 700 may include one or more of the following components: a processing component 702, a memory 704, a power supply component 706, a multimedia component 708, an audio component 710, an input / output (I / O) interface 712, a sensor component 714, and a communication component 716.

[0159] Processing component 702 typically controls the overall operation of electronic device 700, such as operations associated with display, telephone calls, data communication, camera operation, and recording. Processing component 702 may include one or more processors 720 to execute instructions to complete all or part of the steps of the aforementioned disconnection control method. Furthermore, processing component 702 may include one or more modules to facilitate interaction between processing component 702 and other components. For example, processing component 702 may include a multimedia module to facilitate interaction between multimedia component 708 and processing component 702.

[0160] Memory 704 is configured to store various types of data to support the operation of electronic device 700. Examples of this data include instructions for any application or method operating on electronic device 700, contact data, phonebook data, messages, pictures, videos, etc. Memory 704 can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as static random access memory (SRAM), electrically erasable programmable read-only memory (EEPROM), erasable programmable read-only memory (EPROM), programmable read-only memory (PROM), read-only memory (ROM), magnetic storage, flash memory, magnetic disk, or optical disk.

[0161] Power supply component 706 provides power to various components of electronic device 700. Power supply component 706 may include a power management system, one or more power supplies, and other components associated with generating, managing, and distributing power to electronic device 700.

[0162] Multimedia component 708 includes a screen that provides an output interface between the electronic device 700 and the user. In some embodiments, the screen may include a liquid crystal display (LCD) and a touch panel (TP). If the screen includes a touch panel, the screen may be implemented as a touchscreen to receive input signals from the user. The touch panel includes one or more touch sensors to sense touches, swipes, and gestures on the touch panel. The touch sensors may sense not only the boundaries of the touch or swipe action but also the duration and pressure associated with the touch or swipe operation. In some embodiments, multimedia component 708 includes a front-facing camera and / or a rear-facing camera. When the electronic device 700 is in an operating mode, such as a shooting mode or a video mode, the front-facing camera and / or the rear-facing camera may receive external multimedia data. Each front-facing camera and rear-facing camera may be a fixed optical lens system or have focal length and optical zoom capabilities.

[0163] Audio component 710 is configured to output and / or input audio signals. For example, audio component 710 includes a microphone (MIC) configured to receive external audio signals when electronic device 700 is in an operating mode, such as call mode, recording mode, and voice recognition mode. The received audio signals may be further stored in memory 704 or transmitted via communication component 716. In some embodiments, audio component 710 also includes a speaker for outputting audio signals.

[0164] I / O interface 712 provides an interface between processing component 702 and peripheral interface modules, such as keyboards, click wheels, buttons, etc. These buttons may include, but are not limited to, home buttons, volume buttons, power buttons, and lock buttons.

[0165] Sensor assembly 714 includes one or more sensors for providing state assessments of various aspects of electronic device 700. For example, sensor assembly 714 can detect the on / off state of electronic device 700, the relative positioning of components such as the display and keypad of electronic device 700, changes in position of electronic device 700 or a component of electronic device 700, the presence or absence of user contact with electronic device 700, orientation or acceleration / deceleration of electronic device 700, and temperature changes of electronic device 700. Sensor assembly 714 may include a proximity sensor configured to detect the presence of nearby objects without any physical contact. Sensor assembly 714 may also include a light sensor, such as a CMOS or CCD image sensor, for use in imaging applications. In some embodiments, sensor assembly 714 may also include an accelerometer, gyroscope, magnetometer, pressure sensor, or temperature sensor.

[0166] Communication component 716 is configured to facilitate wired or wireless communication between electronic device 700 and other devices. Electronic device 700 can access wireless networks based on communication standards, such as WiFi, 2G, or 3G, or combinations thereof. In one exemplary embodiment, communication component 716 receives broadcast signals or broadcast-related information from an external broadcast management system via a broadcast channel. In one exemplary embodiment, communication component 716 also includes a near-field communication (NFC) module to facilitate short-range communication. For example, the NFC module may be implemented based on radio frequency identification (RFID) technology, Infrared Data Association (IrDA) technology, ultra-wideband (UWB) technology, Bluetooth (BT) technology, and other technologies.

[0167] In an exemplary embodiment, the electronic device 700 may be implemented by one or more application-specific integrated circuits (ASICs), digital signal processors (DSPs), digital signal processing devices (DSPDs), programmable logic devices (PLDs), field-programmable gate arrays (FPGAs), controllers, microcontrollers, microprocessors, or other electronic components to perform the above-described disconnection control method.

[0168] In an exemplary embodiment, a non-transitory computer-readable storage medium including instructions is also provided, such as a memory 704 including instructions, which can be executed by a processor 720 of an electronic device 700 to complete the aforementioned disconnection control method. For example, the non-transitory computer-readable storage medium may be a ROM, random access memory (RAM), CD-ROM, magnetic tape, floppy disk, and optical data storage device, etc.

[0169] A non-transitory computer-readable storage medium, when instructions in the storage medium are executed by a terminal's processor, enables the terminal to execute a disconnection control method, the method comprising:

[0170] During this Czochralski single crystal growth process, characteristic data was acquired, including control data and monitoring data.

[0171] The feature data is input into the line break prediction model, wherein the line break prediction model is trained using feature data samples and corresponding labeled line break results of the shoulder release stage; wherein the line break results include broken lines and unbroken lines.

[0172] Based on the feature data, the predicted breakage rate for the shoulder-laying stage is generated by the breakage prediction model;

[0173] If the predicted breakage rate reaches a preset threshold, the process parameters are adjusted to prevent breakage during the shoulder-setting stage.

[0174] Optionally, the feature data sample is the feature data sample before the shoulder formation stage, and the acquisition of feature data during this Czochralski single crystal pulling process includes:

[0175] When the Czochralski single crystal process enters the shoulder formation stage, characteristic data prior to the shoulder formation stage are acquired.

[0176] Optionally, the feature data includes statistical feature data, and acquiring the feature data during this Czochralski single crystal growth process further includes:

[0177] Based on the feature data prior to the shoulder-releasing stage, corresponding statistical feature data is generated;

[0178] Add the statistical feature data to the feature data.

[0179] Optionally, before inputting the feature data into the line break prediction model, the method further includes:

[0180] After training the line breakage prediction model using feature data samples and the line breakage results of the corresponding labeled shoulder release stage samples, the feature weights corresponding to various feature data in the trained line breakage prediction model are obtained.

[0181] The first preset number of important feature data are determined by sorting the feature weights from high to low, and the important feature data are feature data corresponding to controllable process parameters;

[0182] When the predicted wire breakage rate reaches a preset threshold, adjusting the process parameters to prevent wire breakage during the shoulder-setting stage includes:

[0183] The process parameters corresponding to the important feature data are adjusted.

[0184] Optionally, adjusting the process parameters corresponding to the important feature data includes:

[0185] For the process parameters corresponding to various important feature data, determine multiple candidate values ​​of the process parameters within the corresponding value range;

[0186] Based on multiple candidate values ​​of the various process parameters, generate a combination of candidate values ​​for the various process parameters;

[0187] Based on each of the candidate value combinations, the line break prediction model is used to generate the corresponding candidate predicted line break rate.

[0188] The process parameters are adjusted based on the second preset number of candidate value combinations that have the smallest predicted breakage rate.

[0189] Optionally, acquiring feature data during this Czochralski single crystal growth process includes:

[0190] During the shoulder-releasing phase, the control data and monitoring data are collected once every first set time interval;

[0191] The control data and monitoring data collected consecutively for a preset number of times are determined as the feature data.

[0192] Optionally, before inputting the feature data into the line break prediction model, the method further includes:

[0193] Acquire feature data samples and corresponding marked sample disconnection results for the shoulder release stage; wherein, the feature data samples include control data samples and monitoring data samples collected consecutively for a preset number of times, and the sample disconnection results are the disconnection results after the last collection at a second preset time interval from the control data samples and monitoring data samples collected consecutively for a preset number of times;

[0194] Input the feature data samples and the corresponding labeled sample disconnection results into the disconnection prediction model;

[0195] The disconnection prediction model is trained using the disconnection results of the sample labeled with the feature data sample until the loss value of the disconnection prediction model is less than a set loss value, thus obtaining the trained disconnection prediction model.

[0196] The various embodiments in this specification are described in a progressive manner, with each embodiment focusing on the differences from other embodiments. The same or similar parts between the various embodiments can be referred to each other.

[0197] Those skilled in the art will understand that embodiments of the present invention can be provided as methods, apparatus, or computer program products. Therefore, embodiments of the present invention can take the form of entirely hardware embodiments, entirely software embodiments, or embodiments combining software and hardware aspects. Furthermore, embodiments of the present invention can take the form of computer program products implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0198] Embodiments of the present invention are described with reference to flowchart illustrations and / or block diagrams of methods, terminal devices (systems), and computer program products according to embodiments of the invention. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing terminal device to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing terminal device, generate instructions for implementing the flowchart illustrations and / or block diagrams. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0199] These computer program instructions may also be stored in a computer-readable storage medium capable of directing a computer or other programmable data processing terminal device to operate in a predictive manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0200] These computer program instructions can also be loaded onto a computer or other programmable data processing terminal equipment, causing a series of operational steps to be performed on the computer or other programmable terminal equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable terminal equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0201] Although preferred embodiments of the present invention have been described, those skilled in the art, upon learning the basic inventive concept, can make other changes and modifications to these embodiments. Therefore, the appended claims are intended to be interpreted as including the preferred embodiments as well as all changes and modifications falling within the scope of the embodiments of the present invention.

[0202] Finally, it should be noted that in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or terminal device that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or terminal device. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or terminal device that includes said element.

[0203] The present invention has provided a detailed description of a disconnection control method and apparatus, an electronic device, and a readable storage medium. Specific examples have been used to illustrate the principles and implementation methods of the present invention. The descriptions of the above embodiments are only for the purpose of helping to understand the method and core ideas of the present invention. At the same time, for those skilled in the art, there will be changes in the specific implementation methods and application scope based on the ideas of the present invention. Therefore, the content of this specification should not be construed as a limitation of the present invention.

Claims

1. A method for controlling wire disconnection, characterized in that, include: During this Czochralski single crystal growth process, characteristic data was acquired, including control data and monitoring data. The feature data is input into the line break prediction model, wherein the line break prediction model is trained using feature data samples and corresponding labeled line break results of the shoulder release stage; wherein the line break results include broken lines and unbroken lines. Based on the feature data, the predicted breakage rate for the shoulder-laying stage is generated by the breakage prediction model; If the predicted breakage rate reaches a preset threshold, the process parameters are adjusted to prevent breakage during the shoulder-setting stage.

2. The method according to claim 1, characterized in that, The feature data sample refers to the feature data sample prior to the shoulder formation stage. The acquisition of feature data during this Czochralski single crystal pulling process includes: When the Czochralski single crystal process enters the shoulder formation stage, characteristic data prior to the shoulder formation stage are acquired.

3. The method according to claim 2, characterized in that, The feature data includes statistical feature data. Acquiring the feature data during this Czochralski single crystal growth process further includes: Based on the feature data prior to the shoulder-releasing stage, corresponding statistical feature data is generated; Add the statistical feature data to the feature data.

4. The method according to claim 1, characterized in that, Before inputting the feature data into the line break prediction model, the method further includes: After training the line break prediction model using feature data samples and the line break results of the corresponding labeled shoulder release stage samples, the feature weights corresponding to various feature data in the trained line break prediction model are obtained. The first preset number of important feature data are determined by sorting the feature weights from high to low, and the important feature data are feature data corresponding to controllable process parameters; When the predicted wire breakage rate reaches a preset threshold, adjusting the process parameters to prevent wire breakage during the shoulder-setting stage includes: The process parameters corresponding to the important feature data are adjusted.

5. The method according to claim 4, characterized in that, The adjustment of the process parameters corresponding to the important feature data includes: For the process parameters corresponding to various important feature data, determine multiple candidate values ​​of the process parameters within the corresponding value range; Based on multiple candidate values ​​of the various process parameters, generate a combination of candidate values ​​for the various process parameters; Based on each of the candidate value combinations, the line break prediction model is used to generate the corresponding candidate predicted line break rate. The process parameters are adjusted based on the second preset number of candidate value combinations that have the smallest predicted breakage rate.

6. The method according to claim 1, characterized in that, The characteristic data acquired during this Czochralski single crystal growth process includes: During the shoulder-releasing stage, the control data and monitoring data are collected once every first set time interval; The control data and monitoring data collected consecutively for a preset number of times are determined as the feature data.

7. The method according to claim 6, characterized in that, Before inputting the feature data into the line break prediction model, the method further includes: Acquire feature data samples and corresponding marked sample disconnection results for the shoulder release stage; wherein, the feature data samples include control data samples and monitoring data samples collected consecutively for a preset number of times, and the sample disconnection results are the disconnection results after the last collection at a second preset time interval from the control data samples and monitoring data samples collected consecutively for a preset number of times; Input the feature data samples and the corresponding labeled sample disconnection results into the disconnection prediction model; The disconnection prediction model is trained using the disconnection results of the sample labeled with the feature data sample until the loss value of the disconnection prediction model is less than a set loss value, thus obtaining the trained disconnection prediction model.

8. A wire breakage control device, characterized in that, include: The data acquisition module is used to acquire characteristic data during this Czochralski single crystal growth process, wherein the characteristic data includes control data and monitoring data; The data input module is used to input the feature data into the line break prediction model, wherein the line break prediction model is trained by feature data samples and line break results of samples in the corresponding labeled shoulder release stage; wherein the sample line break results include broken lines and unbroken lines. The line breakage rate generation module is used to generate the predicted line breakage rate of the shoulder release stage based on the feature data and the line breakage prediction model. The parameter adjustment module is used to adjust the process parameters when the predicted breakage rate reaches a preset threshold, so as to control the breakage from occurring during the shoulder setting stage.

9. An electronic device, characterized in that, It includes a processor, a communication interface, a memory, and a communication bus, wherein the processor, the communication interface, and the memory communicate with each other through the communication bus; Memory, used to store computer programs; A processor, when executing a program stored in memory, implements the method described in any one of claims 1-7.

10. A readable storage medium, characterized in that, When the instructions in the storage medium are executed by the processor of the electronic device, the electronic device is able to perform one or more of the disconnection control methods as described in claims 1-7.

Images