CN117286565A

CN117286565A - Parameter setting method, device, electronic equipment and storage medium

Info

Publication number: CN117286565A
Application number: CN202210679805.8A
Authority: CN
Inventors: 赵静楠; 马方强; 李朋朋; 杨东; 张积升; 魏铭琪; 吴苗苗; 周永波; 盛燕
Original assignee: Yinchuan Longi Silicon Materials Co ltd
Current assignee: Yinchuan Longi Silicon Materials Co ltd
Priority date: 2022-06-16
Filing date: 2022-06-16
Publication date: 2023-12-26

Abstract

The embodiment of the invention provides a parameter setting method, device, equipment and medium. The method comprises the following steps: the method comprises the steps of acquiring characteristic data before a temperature adjusting stage in the Czochralski crystal process, inputting the characteristic data into a set recommendation model, setting the recommendation model to obtain through characteristic data samples, and training corresponding marked sample temperature adjusting temperature and sample seeding power, generating the recommended temperature of the temperature adjusting stage and the recommended seeding power of the seeding stage according to the characteristic data by the set recommendation model, setting the temperature adjusting temperature to be the recommended temperature adjusting temperature before the temperature adjusting stage, setting the seeding power to be the recommended seeding power, enabling proper recommended temperature adjusting temperature and recommended seeding power to be automatically generated according to the characteristic data, setting the temperature adjusting temperature and the seeding power according to the characteristics, enabling parameters to be automatically set when entering the temperature adjusting stage, and accordingly improving working efficiency, guaranteeing consistency of temperature adjusting results, and improving yield of products and production line output.

Description

Parameter setting method, device, electronic equipment and storage medium

Technical Field

The present invention relates to the field of crystal manufacturing, and in particular, to a parameter setting method, a parameter setting device, an electronic apparatus, and a storage medium.

Background

The preparation process of the monocrystalline silicon material mainly comprises a Czochralski method (Czochralski process/CZ), and the polycrystalline silicon raw material is refined into monocrystalline silicon by the Czochralski method. The process of generating rod-shaped monocrystalline silicon crystals in the process of pulling up the monocrystalline comprises the steps of charging, melting, charging, temperature adjustment, seeding, shouldering, shoulder turning, constant diameter ending and the like.

Wherein, firstly, the quartz crucible is filled on the crucible, and then the polysilicon material is filled on the quartz crucible. And closing the furnace after the charging is completed, then starting vacuumizing, filling argon after the vacuum requirement is met, and starting a heater to heat under the condition of micro negative pressure. The melting material is to gradually raise the temperature according to the technological requirement of the melting material, reach the temperature of the melting material and melt the solid polysilicon raw material into liquid state. To maintain stable melt and temperature fields and continuous growth of crystals, it is necessary to continuously add polysilicon feedstock to the quartz crucible in the furnace. When the melting of the polysilicon material is completed, seeding cannot be started immediately, and the temperature must be adjusted to the seeding temperature by temperature adjustment because the temperature is higher than the seeding temperature. Seeding is a process in which a seed crystal (i.e., a shaped single crystal) previously loaded onto the end of a wire rope is brought into contact with a liquid surface, and silicon molecules are grown in the lattice direction of the seed crystal at a seeding temperature, thereby forming a single crystal. The shouldering is to gradually grow the crystal diameter to a required diameter, and a section of crystal with the diameter gradually becoming larger to the required diameter or so is pulled out along with the length gradually becoming longer in the shouldering process so as to eliminate crystal dislocation. After the crystal grows to the diameter required by production in the shouldering process, the crystal enters the shouldering process. The shoulder is to control the crystal diameter to the diameter required for production. And after the shoulder turning is finished, the step of controlling the diameter of the crystal is carried out, and in the step, the crystal is grown according to the set diameter by automatically controlling the oxygen content and the temperature.

The temperature regulation is a crucial step in the whole crystal pulling process, the setting of double targets (temperature regulation and seeding power) is particularly critical, and the accuracy of parameter setting determines the success rate of seeding and shouldering of the subsequent crystal pulling, thereby influencing single yield. However, the setting of the current dual targets for temperature adjustment requires a lot of experience of personnel, and continuous inspection is required in the temperature adjustment preparation stage to complete the whole temperature adjustment process. The mode consumes a great deal of effort by personnel, has low working efficiency and poor consistency of temperature regulation results, and finally affects lean production. And under the centralized control platform, a single person corresponds to a plurality of single crystal furnaces, and the double targets of temperature adjustment are manually set, so that the action occupies the effective energy of centralized control personnel.

Disclosure of Invention

In view of the above problems, embodiments of the present invention are provided to provide a parameter setting method for overcoming the above problems or at least partially solving the above problems, so as to solve the problems of low working efficiency and poor consistency of temperature adjustment results due to dependence of personnel experience on parameter setting in temperature adjustment stage.

Correspondingly, the embodiment of the invention also provides a parameter setting device, electronic equipment and a storage medium, which are used for ensuring the realization and application of the method.

In order to solve the above problems, an embodiment of the present invention discloses a parameter setting method, including:

acquiring characteristic data before a temperature regulation stage of the Czochralski crystal process, wherein the characteristic data comprises control data and monitoring data related to the temperature regulation temperature of the temperature regulation stage and seeding power of a seeding stage;

inputting the characteristic data into a set recommendation model, wherein the set recommendation model is obtained through training of characteristic data samples, and corresponding marked sample temperature adjustment temperature and sample seeding power;

generating a recommended temperature regulation temperature of the temperature regulation stage and a recommended seeding power of the seeding stage by the set recommended model according to the characteristic data;

before the tempering stage, the tempering temperature is set to the recommended tempering temperature, and the seeding power is set to a recommended seeding power.

Optionally, the set recommendation model includes a first set recommendation model and a loop set recommendation model, and the inputting the feature data into the set recommendation model includes:

if the current Czochralski single crystal process is the first Czochralski single crystal process after the current crucible charging, inputting the characteristic data into the first set recommendation model;

And if the current Czochralski single crystal process is a non-first Czochralski single crystal process after the current crucible charging, inputting the characteristic data into the cycle setting recommendation model.

Optionally, the current Czochralski single crystal process is a first Czochralski single crystal process after the current crucible is charged; the characteristic data comprise sub-characteristic data in the crystal pulling process after the last crucible charging and sub-characteristic data before a temperature adjusting stage in the current Czochralski crystal process; any one characteristic data sample corresponding to the first set recommended model comprises sub-characteristic data in the crystal pulling process after the last crucible charging of the historical crucible charging and sub-characteristic data before the temperature regulating stage in the first straight-pulling monocrystal process after the historical crucible charging.

Optionally, the current Czochralski single crystal process is a non-first Czochralski single crystal process after the current crucible is charged; the characteristic data comprise sub-characteristic data in the last crystal pulling process after the current crucible is charged and sub-characteristic data before a temperature adjusting stage in the current Czochralski crystal pulling process; any one characteristic data sample corresponding to the cycle setting recommendation model comprises sub-characteristic data in the last pulling process of the historical Czochralski single crystal process after the historical crucible is filled, and sub-characteristic data before a temperature adjusting stage in the historical Czochralski single crystal process.

Optionally, before the feature data is input into the set recommendation model, the method further includes:

detecting the actual liquid level brightness before the temperature regulating stage;

and under the condition that the brightness of the actual liquid level is in a rising state and reaches a preset value, starting the set recommendation model.

acquiring a characteristic data sample, and correspondingly marking the sample temperature regulation temperature and the sample seeding power;

inputting the characteristic data sample, the sample temperature regulation temperature of the corresponding mark and the sample seeding power into a set recommendation model;

and training the set recommended model by adopting the characteristic data sample, the marked sample temperature regulation temperature and the marked sample seeding power until the loss value of the set recommended model is smaller than the set loss value, so as to obtain the trained set recommended model.

Optionally, the set recommendation model includes at least two kinds; training the set recommended model until the loss value of the set recommended model is smaller than the set loss value, and obtaining the trained set recommended model after the characteristic data sample, the marked sample temperature and the marked sample seeding power are adopted, wherein the method further comprises:

Determining weights of at least two set recommendation models according to the ratio of loss values of the at least two set recommendation models; wherein the loss value and the weight are in inverse proportion;

and according to the weights of the at least two set recommendation models, carrying out weighted summation on the recommended temperature adjustment temperatures of the at least two set recommendation models to obtain a fused recommended temperature adjustment temperature, and carrying out weighted summation on the recommended seeding power of the at least two set recommendation models to obtain the fused recommended seeding power.

The embodiment of the invention also discloses a parameter setting device, which comprises:

the data acquisition module is used for acquiring characteristic data before a temperature regulation stage of the Czochralski crystal process, wherein the characteristic data comprises control data and monitoring data related to the temperature regulation temperature of the temperature regulation stage and the seeding power of the seeding stage;

the data input module is used for inputting the characteristic data into a set recommendation model, wherein the set recommendation model is obtained through training of characteristic data samples, and corresponding marked sample temperature adjustment temperature and sample seeding power;

the parameter generation module is used for generating a recommended temperature regulation temperature of the temperature regulation stage and a recommended seeding power of the seeding stage by the set recommendation model according to the characteristic data;

And the parameter setting module is used for setting the temperature regulation temperature to be the recommended temperature regulation temperature and setting the seeding power to be the recommended seeding power before the temperature regulation stage.

Optionally, the set recommendation model includes a first set recommendation model and a loop set recommendation model, and the data input module includes:

the first data submodule is used for inputting the characteristic data into the first set recommendation model if the current Czochralski crystal process is a first Czochralski crystal process after the current crucible is charged;

and the second input sub-module is used for inputting the characteristic data into the cycle setting recommendation model if the current Czochralski crystal process is a non-first Czochralski crystal process after the current crucible is charged.

Optionally, the apparatus further comprises:

the brightness detection module is used for detecting the actual liquid level brightness before the temperature adjustment stage before the characteristic data are input into a set recommendation model;

and the model starting module is used for starting the set recommendation model under the condition that the brightness of the actual liquid level is in a rising state and reaches a preset value.

Optionally, the apparatus further comprises:

the sample acquisition module is used for acquiring a characteristic data sample, and a sample temperature regulation temperature and a sample seeding power which correspond to the marks before the characteristic data is input into a set recommendation model;

The sample input module is used for inputting the characteristic data samples, the sample temperature adjustment temperature of the corresponding marks and the sample seeding power into a set recommendation model;

and the model training module is used for training the set recommended model by adopting the characteristic data sample, the marked sample temperature regulation temperature and the marked sample seeding power until the loss value of the set recommended model is smaller than the set loss value, so as to obtain the trained set recommended model.

Optionally, the set recommendation model includes at least two kinds; the apparatus further comprises:

the weight determining module is used for training the set recommendation model until the loss value of the set recommendation model is smaller than the set loss value after the characteristic data sample, the marked sample temperature regulation temperature and the marked sample seeding power are adopted, and determining the weights of at least two set recommendation models according to the ratio of the loss values of the at least two set recommendation models after the trained set recommendation model is obtained; wherein the loss value and the weight are in inverse proportion;

and the fusion module is used for carrying out weighted summation on the recommended temperature adjustment temperatures of the at least two set recommended models according to the weights of the at least two set recommended models to obtain fusion recommended temperature adjustment temperatures, and carrying out weighted summation on the recommended seeding powers of the at least two set recommended models to obtain fusion recommended seeding powers.

The embodiment of the invention also discloses an electronic device which is characterized by comprising a processor, a communication interface, a memory and a communication bus, wherein the processor, the communication interface and the memory are communicated with each other through the communication bus;

a memory for storing a computer program;

and a processor for performing the method steps described above when executing the program stored on the memory.

The embodiment of the invention also discloses a readable storage medium, and when the instructions in the storage medium are executed by a processor of the electronic device, the electronic device can execute one or more of the parameter setting methods in the embodiment of the invention.

The embodiment of the invention has the following advantages:

according to the embodiment of the invention, the characteristic data before the temperature regulation stage of the Czochralski crystal process is obtained, wherein the characteristic data comprises control data and monitoring data related to the temperature regulation temperature of the temperature regulation stage and the seeding power of the seeding stage, the characteristic data is input into a set recommendation model, the set recommendation model is obtained through training of a characteristic data sample, the sample temperature regulation temperature and the sample seeding power which correspond to marks, the set recommendation model is used for generating the recommended temperature regulation temperature of the temperature regulation stage and the recommended seeding power of the seeding stage according to the characteristic data, the temperature regulation temperature is set as the recommended temperature regulation temperature before the temperature regulation stage, the seeding power is set as the recommended seeding power, so that the proper recommended temperature regulation temperature and the recommended seeding power are automatically generated according to the characteristic data, the temperature regulation temperature and the seeding power are set according to the characteristic data, the parameters are automatically set when the sample temperature regulation stage is entered, the working efficiency is low, the consistency of the temperature regulation result is ensured, and the yield of a product is improved.

Drawings

FIG. 1 is a flow chart of steps of an embodiment of a parameter setting method of the present invention;

FIG. 2 is a schematic diagram of an automatic parameter setting process according to the present invention;

FIG. 3 is a flow chart of steps of an embodiment of a parameter setting method of the present invention;

FIG. 4 is a schematic diagram of a set recommendation model construction process according to the present invention;

FIG. 5 is a block diagram of an embodiment of a parameter setting apparatus of the present invention;

FIG. 6 is a block diagram illustrating a computing device for parameter setting, according to an example embodiment.

Detailed Description

In order that the above-recited objects, features and advantages of the present invention will become more readily apparent, a more particular description of the invention will be rendered by reference to the appended drawings and appended detailed description.

Referring to fig. 1, a flowchart illustrating steps of an embodiment of a parameter setting method of the present invention may specifically include the following steps:

and 101, acquiring characteristic data before a temperature regulation stage of the Czochralski crystal process, wherein the characteristic data comprises control data and monitoring data related to the temperature regulation temperature of the temperature regulation stage and the seeding power of the seeding stage.

In the embodiment of the present invention, the Czochralski process is a process of pulling a raw material into a single crystal using a Czochralski method, for example, a process of Czochralski silicon. The process of pulling up the single crystal can be divided into a melting stage, a temperature adjusting stage, a seeding stage, a shoulder stage, and the like.

In the embodiment of the invention, in the process of pulling a single crystal once, the temperature regulation temperature of the temperature regulation stage and the seeding power of the seeding stage need to be set before entering the temperature regulation stage. Before the temperature regulation stage of the Czochralski crystal process, various characteristic data can be obtained. The characteristic data includes control data and monitoring data related to the tempering temperature of the tempering stage and the seeding power of the seeding stage. The control data is a control parameter input to the czochralski crystal growing apparatus, the monitoring data is data monitored during actual operation, and specifically may include any applicable characteristic data, which is not limited in the embodiment of the present invention. The monitoring data may be monitored by industrial sensors. The control data may be collected directly from the associated control device.

In the embodiment of the invention, the types of the characteristic data are consistent with the types of the characteristic data samples adopted in the process of setting the recommended model training.

And 102, inputting the characteristic data into a set recommendation model, wherein the set recommendation model is obtained through training of characteristic data samples, and corresponding marked sample temperature adjustment temperature and sample seeding power.

In the embodiment of the invention, the recommended temperature regulation temperature and the recommended seeding power can adopt a machine learning mode, and a set recommendation model capable of recommending two set parameters of the temperature regulation temperature and the seeding power is obtained according to the correlation between the characteristic data related to the temperature regulation temperature and the seeding power and the temperature regulation temperature and the seeding power.

In order to train the set recommendation model, accurate sample data and corresponding tag data, i.e. the feature data samples acquired before the temperature adjustment stage, as well as the sample temperature adjustment temperature and sample seeding power of the corresponding marks, are required. The characteristic data sample, the sample temperature adjustment temperature and the sample seeding power can be obtained through multiple experiments, and the characteristic data sample, the sample temperature adjustment temperature and the sample seeding power can be selected from historical data. And inputting the characteristic data samples in the training set, the sample temperature adjustment temperature and the sample seeding power corresponding to the marks into a model, and then training the model until the model converges.

In the embodiment of the present invention, the recommendation model may be set by using a class decision tree model, an optical gradient hoister (Light Gradient Boosting Machine, lightGBM) model, a linear regression model, or any suitable model, which is not limited in the embodiment of the present invention. The optical gradient elevator model is a rapid, distributed and high-performance gradient elevator framework based on a decision tree algorithm, and can be used for sorting, classifying, regression and many other machine learning tasks.

In the embodiment of the invention, in the actual application stage, the input of the recommendation model is set as the characteristic data acquired before the temperature adjustment stage, and the recommendation model is input after the characteristic data is acquired each time. The output of the model is a recommended value of the temperature regulation temperature, which is marked as recommended temperature regulation, and a recommended value of the seeding power, which is marked as recommended seeding power.

And step 103, generating a recommended temperature regulation temperature of the temperature regulation stage and recommended seeding power of the seeding stage by the set recommended model according to the characteristic data.

In the embodiment of the invention, aiming at the current Czochralski crystal process, after the current characteristic data is acquired, the acquired characteristic data is input into a set recommendation model to obtain the output of the set recommendation model, and the output of the set recommendation model is further used as the recommended temperature adjustment temperature of the temperature adjustment stage and the recommended seeding power of the seeding stage.

Step 104, setting the tempering temperature to be the recommended tempering temperature and setting the seeding power to be the recommended seeding power before the tempering stage.

In the embodiment of the invention, the temperature regulation temperature is set as a recommended temperature regulation temperature, and the seeding power is set as a recommended seeding power. And both parameters need to be set on the device before the tempering stage.

In an optional embodiment of the invention, before the feature data is input into the set recommendation model, the method may further include: and before the temperature adjustment stage, detecting the actual liquid level brightness, and starting the set recommendation model under the condition that the actual liquid level brightness is in a rising state and reaches a preset value.

The actual brightness of the liquid surface refers to the actual brightness of the liquid surface after the raw material is melted into a liquid state, and can be detected by a sensor. The device automatically detects the actual level brightness before the tempering stage. When the actual liquid level brightness is in a rising state, namely, the process that the actual liquid level brightness is continuously rising, the actual liquid level brightness can be specifically determined by whether the actual liquid level brightness is higher than the actual liquid level brightness detected last time at regular intervals. And the actual liquid level brightness reaches a preset value, which indicates that the temperature adjustment stage is about to be started, and the recommended model is started.

For example, a schematic diagram of the parameter automatic setting flow shown in fig. 2. Taking the process parameter setting of single crystal production equipment as an example, before entering a temperature regulation stage, the system automatically detects the actual liquid level brightness of the silicon liquid, and if the actual liquid level brightness is in a rising state and the deviation between the actual liquid level brightness and the set liquid level brightness is more than 6 units, a recommended task is triggered, namely, a set recommended model is started. And acquiring data of the current Czochralski single crystal process and data of the last Czochralski single crystal process, extracting features, generating feature data, and inputting the feature data into a set recommendation model. And writing the recommended temperature and the recommended seeding power into a database, automatically inquiring data in the database within 5 minutes by a system, and judging whether the equipment has the latest process parameters. If the special equipment exists, the special equipment is directly issued to the equipment and is directly used in the next step. If the process parameter value does not exist, an alarm task is triggered, and an operator is reminded that the task triggers the recommendation button to set the process parameter value of the next process step.

In an optional embodiment of the present invention, the setting recommendation model includes a first setting recommendation model and a loop setting recommendation model, and inputting the feature data into a specific implementation of the setting recommendation model may include: if the current Czochralski single crystal process is the first Czochralski single crystal process after the current crucible charging, inputting the characteristic data into the first set recommendation model; and if the current Czochralski single crystal process is a non-first Czochralski single crystal process after the current crucible charging, inputting the characteristic data into the cycle setting recommendation model.

Crucible charging means that a quartz crucible is charged on a crucible, and then raw materials are charged in the quartz crucible. For example, in the process of Czochralski silicon, a quartz crucible is charged onto a crucible, and then a polycrystalline silicon material is charged into the quartz crucible. And closing the furnace after the charging is completed, then starting vacuumizing, filling argon after the vacuum requirement is met, and starting a heater to heat under the condition of micro negative pressure.

For the Czochralski single crystal process, after one crucible charge, the first and subsequent Czochralski single crystal processes differ in parameter settings due to differences in the physicochemical environment. The set recommended model available during the first czochralski crystal process is unable to recommend available recommended values during non-first czochralski crystal processes. And vice versa.

The parameter settings of the Czochralski crystal process after a single crucible charge are affected by the last Czochralski crystal process. However, the first Czochralski crystal process after switching to the current crucible charge lacks relevant data and is also greatly affected by the previous Czochralski crystal process than the subsequent Czochralski crystal process. Therefore, a set recommended model is trained for the first Czochralski single crystal process after one crucible charge, and is recorded as a first set recommended model, and a set recommended model is trained for the non-first Czochralski single crystal process after one crucible charge, and is recorded as a cyclic set recommended model.

If the current Czochralski single crystal process is the first Czochralski single crystal process after the current crucible loading, inputting the characteristic data into a first set recommendation model. If the current Czochralski single crystal process is a non-first Czochralski single crystal process after the current crucible loading, inputting the characteristic data into a cycle setting recommendation model.

The training data used for the first set recommendation model and the loop set recommendation model are different. The model types adopted by the first set recommendation model and the loop set recommendation model can be the same or different.

According to the first Czochralski single crystal process and the non-first Czochralski single crystal process after the current crucible is charged, a first-time set recommendation model and a cyclic set recommendation model are respectively adopted, so that the models applicable to both the first-time Czochralski single crystal process and the non-first-time Czochralski single crystal process can be recommended, recommended parameters are more accurate, and the method accords with the unique technological characteristics of the Czochralski single crystal process.

In an alternative embodiment of the present invention, the current Czochralski crystal process is a first Czochralski crystal process after the current crucible charge. The characteristic data comprise sub-characteristic data in the crystal pulling process after the last crucible charging and sub-characteristic data before a temperature adjusting stage in the current Czochralski crystal process; any one characteristic data sample corresponding to the first set recommended model comprises sub-characteristic data in the crystal pulling process after the last crucible charging of the historical crucible charging and sub-characteristic data before the temperature regulating stage in the first straight-pulling monocrystal process after the historical crucible charging.

Whether the application stage or the training stage of the model, feature data of the model, that is, one feature data, is input at a time, and is composed of a plurality of sub-feature data. For example, the characteristic data includes a crucible position, which is one sub-characteristic data, and a growth rate, which is another sub-characteristic data.

The pulling process after the last crucible charge of the current crucible charge is completed. The sub-characteristic data during the pulling process may therefore include sub-characteristic data before the tempering stage and after the tempering stage. For example, the sub-feature data mainly includes a temperature adjustment stage, a seeding stage, a shoulder stage, and the like. During the first Czochralski crystal process of the current crucible charge, only sub-characteristic data prior to the tempering phase can be obtained. For example, sub-feature data mainly including a frit stage, a temperature adjustment stage, and the like.

The training process of first setting the recommended model requires a large number of feature data samples. Any characteristic data sample corresponding to the first set recommended model is also corresponding, and specifically comprises sub-characteristic data in the crystal pulling process after the last crucible loading of the historical crucible loading and sub-characteristic data before the temperature adjusting stage in the first straight-pulling single crystal process after the historical crucible loading. That is, only the data before and after crucible charging in the history data is used as training data so that the setting recommendation model can learn the parameter setting of the first Czochralski single crystal process after crucible charging.

In an alternative embodiment of the present invention, the current czochralski single crystal process is a non-first czochralski single crystal process after the current crucible charge; the characteristic data comprise sub-characteristic data in the last crystal pulling process after the current crucible is charged and sub-characteristic data before a temperature adjusting stage in the current Czochralski crystal pulling process; any one characteristic data sample corresponding to the cycle setting recommendation model comprises a characteristic data sample in the last pulling process of the historical Czochralski single crystal process after the historical crucible is filled, and sub-characteristic data before a temperature adjustment stage in the historical Czochralski single crystal process.

For non-primary pulling processes after the current crucible charge, the last pulling process is already complete. The sub-feature data during the last pull may therefore include sub-feature data before the tempering stage and after the tempering stage. For example, the sub-feature data mainly includes a temperature adjustment stage, a seeding stage, a shoulder stage, and the like. The current Czochralski crystal process after the current crucible charging is still in the running process, and only the sub-characteristic data before the temperature adjustment stage can be obtained. For example, sub-feature data mainly including a frit stage, a temperature adjustment stage, and the like.

The training process of the loop-setting recommendation model requires a large number of feature data samples. Any characteristic data sample corresponding to the cycle setting recommendation model is also corresponding, and specifically comprises sub-characteristic data in the last pulling process of the history Czochralski single crystal process after the history crucible is filled and sub-characteristic data before a temperature adjustment stage in the history Czochralski single crystal process after the history crucible is filled. That is, only the data before and after each Czochralski single crystal process after one crucible charge in the history data is used as training data, so that the setting recommendation model can learn the parameter setting of the non-first Czochralski single crystal process after the current crucible charge.

Referring to fig. 3, a flowchart illustrating steps of an embodiment of a parameter setting method of the present invention may specifically include the following steps:

step 201, obtaining a characteristic data sample, and correspondingly marked sample temperature regulation temperature and sample seeding power.

In an embodiment of the invention, the prediction process corresponds to a training process. The characteristic data samples are collected control data samples and monitoring data samples. The specific implementation mode is as follows: and performing data cleaning on the historical data, extracting some data to serve as characteristic data samples, performing correlation analysis on the characteristic data samples, the sample temperature regulation temperature and the sample seeding power, and finally obtaining training data of the model.

For example, the full-scale historical data of the last-time pulling is acquired from a SCADA (Supervisory Control And Data Acquisition, data acquisition and monitoring control system) database, and the time sequence data of the melting material, temperature adjustment, seeding and other stages are subjected to basic data cleaning processes such as denoising, filling, smoothing and the like according to equipment number groups so as to facilitate subsequent processes. Taking single crystal production data as an example, the characteristic data are mainly obtained by extracting time sequence data of multiple dimensions such as temperature adjustment, seeding, shouldering and the like, such as crucible position, growth speed, liquid silicon temperature and the like. And performing correlation analysis on the characteristic data sample, the sample temperature regulation temperature and the sample seeding power. Firstly, constructing a two-dimensional characteristic matrix by a plurality of characteristic data such as melting time, high-temperature energy, seeding pulling speed, temperature and the like, and temperature adjustment temperature and seeding power. And then, analyzing the correlation between every two feature dimensions by adopting a Spearman correlation coefficient, and reserving feature data with a correlation coefficient of temperature regulation and seeding power larger than 0.4 to prepare for the next step of building a model.

During model training, the feature data samples may be partitioned, with the training set accounting for 80% of the population and the test set accounting for 20% of the population. Training the model by using the data of the training set, and testing the model by using the data of the testing set. In addition, before the method, the characteristic data sample can be divided into a characteristic set of a first Czochralski single crystal process and a characteristic set of a non-first Czochralski single crystal process according to requirements, and then the two characteristic sets are respectively divided into a training set and a testing set.

For example, a schematic diagram of the set recommendation model construction flow is shown in fig. 4. A data characteristic dataset is input. Firstly, dividing the feature set of the first set recommendation model and the feature set of the cyclic set recommendation model. And splitting the feature set of the first set recommendation model into a training set and a testing set, and splitting the feature set of the cyclic set recommendation model into the training set and the testing set.

And 202, inputting the characteristic data samples, the sample tempering temperature of the corresponding marks and the sample seeding power into a set recommendation model.

In the embodiment of the invention, when the set recommended model is trained, a characteristic data sample, a sample temperature regulation temperature corresponding to the mark and a sample seeding power are input. And inputting a group of data, namely a group of characteristic data samples, the sample temperature regulation temperature of the corresponding mark and the sample seeding power at a time. After inputting a set of data, the model is trained once.

And 203, training the set recommended model by adopting the characteristic data sample, the marked sample temperature regulation temperature and the marked sample seeding power until the loss value of the set recommended model is smaller than the set loss value, and obtaining the trained set recommended model.

In the embodiment of the invention, the marked sample tempering temperature and sample seeding power are adopted to train the set recommendation model. That is, after the model outputs the recommended values of the tempering temperature and the seeding power each time, the recommended values are compared with the actual values (namely, the marked sample tempering temperature and the marked sample seeding power), and the comparison result is input into a loss function to calculate the loss value. The convergence conditions of the model may be: the loss value is less than the set loss value, or the maximum number of iterations is reached. For example, a smaller set loss value is set, the magnitude of the loss value is calculated at the same time during each training, and when the loss value is smaller than the set loss value, the model can be considered to be converged, and then the training can be finished. A relatively large maximum number of iterations, such as 100 iterations, 10000 iterations, 1000000 iterations, etc., is preset, and needs to be selected according to the actual situation, which is not limited in the embodiment of the present invention. After the model is trained for a prescribed number of times, the model can be considered to be trained.

In an alternative embodiment of the present invention, the set recommendation model includes at least two kinds; training the set recommended model until the loss value of the set recommended model is smaller than the set loss value, and obtaining the trained set recommended model, wherein the training method further comprises the following steps: determining weights of at least two set recommendation models according to the ratio of loss values of the at least two set recommendation models; wherein the loss value and the weight are in inverse proportion; and according to the weights of the at least two set recommendation models, carrying out weighted summation on the recommended temperature adjustment temperatures of the at least two set recommendation models to obtain a fused recommended temperature adjustment temperature, and carrying out weighted summation on the recommended seeding power of the at least two set recommendation models to obtain the fused recommended seeding power.

And selecting at least two proper models, and training by adopting the characteristic data samples, the marked sample temperature regulation temperature and the sample seeding power to obtain at least two trained set recommended models. And fusing at least two set recommended models, namely fusing the results output by the at least two set recommended models. Specifically, the weights of at least two set recommendation models are determined according to the ratio of the loss values of the at least two set recommendation models. Since a smaller loss value for a trained model generally indicates a higher accuracy for that model, the loss value and weight should be inversely proportional. For example, if the ratio of the loss value of model A to the loss value of model B is 1:2, the weight of model A is 2/3 and the weight of model B is 1/3.

And then carrying out weighted summation on the recommended temperature adjustment temperatures of the at least two set recommended models according to the weights of the at least two set recommended models, and recording the weighted summation result as a fused recommended temperature adjustment temperature. And carrying out weighted summation on the recommended seeding power of at least two set recommended models, and recording the obtained weighted summation result as the fused recommended seeding power.

For example, as shown in fig. 4, for the feature set of the first set recommendation model, a model is selected from the LightGBM models, which can enhance the generalization ability of the model, use the training set for training, and use the testing set for selecting and storing the model with the minimum loss. And the other model is a class decision tree model, a threshold value of the characteristic data of each dimension is calculated, the decision tree model is sequentially constructed according to the size of the correlation coefficient, verification is carried out by using a test set, the depth of the threshold value of the characteristic data is adjusted, and the model with the minimum loss is saved. And finally, fusing the results of the two models according to the ratio of the loss values of the two models. Aiming at the feature set of the circularly set recommendation model, a model selects a class decision tree model, calculates the threshold value of feature data of each dimension, sequentially constructs the decision tree model according to the size of the correlation coefficient, uses a test set for verification, adjusts the depth of the threshold value of the feature data, and stores the model with the minimum loss. The other model is a regression model, and the regression model with the best test result is selected from improved EN (linear regression of a regularization matrix) models through linear regression model training and checking. And finally, fusing the results of the two models according to the ratio of the loss values of the two models. In addition, in production practice, the model may be updated continuously. After the model is obtained through each training, if the old model exists, the new model and the old model are fused according to the weight proportion, the new model is generated for storage and deployment, and otherwise, the storage and deployment is directly carried out. The weight can be manually adjusted as an adjustable parameter in the process.

In the training process of the class decision tree model, the information gain g (X, Y) =H (X) -H (X|Y) of each feature data is calculated. Wherein, the information entropyConditional entropy->Wherein x= { X ₁ ，x ₂ ，...，x _n Is a feature vector, x _n Is characteristic data, p (x _i ) Representing probability, y= { Y ₁ ，y ₂ ，...，y _n }，p(y _i ) Representing the probability. And selecting the characteristic data with the maximum information gain as the dividing node, continuously calculating the information gain of the residual characteristic data, and sequentially constructing a decision tree to generate a model file.

In an alternative embodiment of the invention, model training requires data accumulation in view of process improvement issues. When the data quantity is insufficient, a weight can be set for each characteristic data manually, and before model training, the weight is multiplied on the information gain value of each characteristic data, so that the problem of inaccurate recommendation caused by insufficient data quantity is avoided. As data accumulates more and more, models become more accurate.

And 204, acquiring characteristic data before a temperature regulation stage of the Czochralski crystal process, wherein the characteristic data comprises control data and monitoring data related to the temperature regulation temperature of the temperature regulation stage and the seeding power of the seeding stage.

In the embodiments of the present invention, the specific implementation manner of this step may be referred to the description in the foregoing embodiments, which is not repeated herein.

And 205, inputting the characteristic data into a set recommendation model, wherein the set recommendation model is obtained through training of characteristic data samples, and corresponding marked sample temperature adjustment temperature and sample seeding power.

And 206, generating a recommended temperature regulation temperature of the temperature regulation stage and recommended seeding power of the seeding stage by the set recommended model according to the characteristic data.

Step 207, setting the tempering temperature to the recommended tempering temperature and setting the seeding power to the recommended seeding power before the tempering stage.

According to the embodiment of the invention, the characteristic data sample, the sample tempering temperature and the sample seeding power which correspond to the marks are input into a set recommendation model, the set recommendation model is trained by adopting the characteristic data sample, the sample tempering temperature and the sample seeding power which correspond to the marks until the loss value of the set recommendation model is smaller than the set loss value, the trained set recommendation model is obtained, the characteristic data before the tempering stage of the Czochralski single crystal process is obtained, the characteristic data comprises control data and monitoring data related to the tempering temperature of the tempering stage and the seeding power of the seeding stage, the characteristic data is input into a set recommendation model, the set recommendation model is trained by adopting the characteristic data sample, the recommended tempering temperature of the recommended stage and the seeding power of the seeding stage are generated by the set recommendation model according to the characteristic data, the tempering temperature is set as the characteristic data before the stage, the tempering power is automatically regulated to be the recommended power, the set seeding power is automatically regulated to be the recommended power, and the output is enabled to be consistent when the set seeding power is automatically regulated, the set seeding power is enabled to be the recommended power, and the product quality is enabled to be high, and the quality is enabled to be good, and the quality is enabled to be high.

It should be noted that, for simplicity of description, the method embodiments are shown as a series of acts, but it should be understood by those skilled in the art that the embodiments are not limited by the order of acts, as some steps may occur in other orders or concurrently in accordance with the embodiments. Further, those skilled in the art will appreciate that the embodiments described in the specification are presently preferred embodiments, and that the acts are not necessarily required by the embodiments of the invention.

Referring to fig. 5, a block diagram of an embodiment of a parameter setting apparatus according to the present invention is shown, and may specifically include the following modules:

a data acquisition module 301, configured to acquire feature data before a temperature adjustment stage of the present czochralski crystal growing process, where the feature data includes control data and monitoring data related to a temperature adjustment temperature of the temperature adjustment stage and seeding power of the seeding stage;

the data input module 302 is configured to input the feature data into a set recommendation model, where the set recommendation model is obtained through training of a feature data sample, a sample temperature adjustment temperature and a sample seeding power corresponding to the mark;

A parameter generating module 303, configured to generate, according to the feature data, a recommended tempering temperature of the tempering stage and a recommended seeding power of the seeding stage by using the set recommendation model;

the parameter setting module 304 is configured to set the tempering temperature to the recommended tempering temperature and set the seeding power to the recommended seeding power before the tempering stage.

Optionally, the apparatus further comprises:

For the device embodiments, since they are substantially similar to the method embodiments, the description is relatively simple, and reference is made to the description of the method embodiments for relevant points.

Fig. 6 is a block diagram illustrating a configuration of an electronic device 400 for shoulder actuation, according to an example embodiment. For example, electronic device 400 may be a mobile phone, computer, digital broadcast terminal, messaging device, game console, tablet device, medical device, exercise device, personal digital assistant, or the like.

Referring to fig. 6, electronic device 400 may include one or more of the following components: a processing component 402, a memory 404, a power supply component 406, a multimedia component 408, an audio component 410, an input/output (I/O) interface 412, a sensor component 414, and a communication component 416.

The processing component 402 generally controls overall operation of the electronic device 400, such as operations associated with display, telephone calls, data communications, camera operations, and recording operations. The processing component 402 may include one or more processors 420 to execute instructions to perform all or part of the steps of the parameter setting method described above. Further, the processing component 402 can include one or more modules that facilitate interaction between the processing component 402 and other components. For example, the processing component 402 may include a multimedia module to facilitate interaction between the multimedia component 408 and the processing component 402.

Memory 404 is configured to store various types of data to support operations at device 400. Examples of such data include instructions for any application or method operating on electronic device 400, contact data, phonebook data, messages, pictures, videos, and the like. The memory 404 may be implemented by any type or combination of volatile or nonvolatile memory devices 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 memory, flash memory, magnetic or optical disk.

The power component 404 provides power to the various components of the electronic device 400. Power component 404 can include a power management system, one or more power sources, and other components associated with generating, managing, and distributing power for electronic device 400.

The multimedia component 408 includes a screen between the electronic device 400 and the user that provides an output interface. 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 touch screen to receive input signals from a user. The touch panel includes one or more touch sensors to sense touches, swipes, and gestures on the touch panel. The touch sensor may sense not only the boundary of a touch or slide action, but also the duration and pressure associated with the touch or slide operation. In some embodiments, the multimedia component 408 includes a front camera and/or a rear camera. When the electronic device 400 is in an operational 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 camera and rear camera may be a fixed optical lens system or have focal length and optical zoom capabilities.

The audio component 410 is configured to output and/or input audio signals. For example, the audio component 410 includes a Microphone (MIC) configured to receive external audio signals when the electronic device 400 is in an operational mode, such as a call mode, a recording mode, and a voice recognition mode. The received audio signals may be further stored in the memory 404 or transmitted via the communication component 416. In some embodiments, audio component 410 further includes a speaker for outputting audio signals.

The I/O interface 412 provides an interface between the processing component 402 and peripheral interface modules, which may be a keyboard, click wheel, buttons, etc. These buttons may include, but are not limited to: homepage button, volume button, start button, and lock button.

The sensor assembly 414 includes one or more sensors for providing status assessment of various aspects of the electronic device 400. For example, the sensor assembly 414 may detect an on/off state of the device 400, a relative positioning of components, such as a display and keypad of the electronic device 400, a change in position of the electronic device 400 or a component of the electronic device 400, the presence or absence of a user's contact with the electronic device 400, an orientation or acceleration/deceleration of the electronic device 400, and a change in temperature of the electronic device 400. The sensor assembly 414 may include a proximity sensor configured to detect the presence of nearby objects in the absence of any physical contact. The sensor assembly 414 may also include a light sensor, such as a CMOS or CCD image sensor, for use in imaging applications. In some embodiments, the sensor assembly 414 may also include an acceleration sensor, a gyroscopic sensor, a magnetic sensor, a pressure sensor, or a temperature sensor.

The communication component 416 is configured to facilitate communication between the electronic device 400 and other devices, either wired or wireless. The electronic device 400 may access a wireless network based on a communication standard, such as WiFi,2G, or 3G, or a combination thereof. In one exemplary embodiment, the communication part 414 receives a broadcast signal or broadcast-related information from an external broadcast management system via a broadcast channel. In an exemplary embodiment, the communication component 414 further includes a Near Field Communication (NFC) module to facilitate short range communications. 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.

In an exemplary embodiment, the electronic device 400 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 elements for executing the above-described parameter setting methods.

In an exemplary embodiment, a non-transitory computer readable storage medium is also provided, such as memory 404, that includes instructions executable by processor 420 of electronic device 400 to perform the above-described parameter setting method. For example, the non-transitory computer readable storage medium may be ROM, random Access Memory (RAM), CD-ROM, magnetic tape, floppy disk, optical data storage device, etc.

A non-transitory computer readable storage medium, which when executed by a processor of a terminal, causes the terminal to perform a parameter setting method, the method comprising:

In this specification, each embodiment is described in a progressive manner, and each embodiment is mainly described by differences from other embodiments, and identical and similar parts between the embodiments are all enough to be referred to each other.

It will be apparent to those skilled in the art that embodiments of the present invention may be provided as a method, apparatus, or computer program product. Accordingly, embodiments of the present invention may take the form of an entirely hardware embodiment, an entirely software embodiment or an embodiment combining software and hardware aspects. Furthermore, embodiments of the invention may take the form of a computer program product on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, etc.) having computer-usable program code embodied therein.

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 flow and/or block of the flowchart illustrations and/or block diagrams, and combinations of flows and/or blocks in the flowchart illustrations and/or block diagrams, can be implemented by computer program instructions. These computer program instructions may 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, create means for implementing the functions specified in the flowchart flow or flows and/or block diagram block or blocks.

These computer program instructions may also be stored in a computer-readable memory that can direct a computer or other programmable data processing apparatus to function in a particular manner, such that the instructions stored in the computer-readable memory produce an article of manufacture including instruction means which implement the function specified in the flowchart flow or flows and/or block diagram block or blocks.

These computer program instructions may also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer implemented process such that the instructions which execute on the computer or other programmable apparatus provide steps for implementing the functions specified in the flowchart flow or flows and/or block diagram block or blocks.

While preferred embodiments of the present invention have been described, additional variations and modifications in those embodiments may occur to those skilled in the art once they learn of the basic inventive concepts. It is therefore intended that the following claims be interpreted as including the preferred embodiment and all such alterations and modifications as fall within the scope of the embodiments of the invention.

Finally, it is further noted that relational terms such as first and second, and the like are used solely to distinguish one entity or action from another entity or action without necessarily requiring or implying any actual such relationship or order between such entities or actions. Moreover, the terms "comprises," "comprising," or any other variation thereof, are intended to cover a non-exclusive inclusion, such that a process, method, article, or terminal that comprises a list of elements does not include only those elements but may include other elements not expressly listed or inherent to such process, method, article, or terminal. Without further limitation, an element defined by the phrase "comprising one … …" does not exclude the presence of other like elements in a process, method, article or terminal device comprising the element.

The above detailed description of a parameter setting method and apparatus, an electronic device and a readable storage medium provided by the present invention, the specific examples are applied to illustrate the principles and embodiments of the present invention, and the above examples are only used to help understand the method and core idea of the present invention; meanwhile, as those skilled in the art will have variations in the specific embodiments and application scope in accordance with the ideas of the present invention, the present description should not be construed as limiting the present invention in view of the above.

Claims

1. A parameter setting method, comprising:

2. The method of claim 1, wherein the set recommendation model includes a first set recommendation model and a loop set recommendation model, and wherein the inputting the feature data into the set recommendation model includes:

3. The method according to claim 2, wherein the current czochralski single crystal process is a first czochralski single crystal process after the current crucible charge; the characteristic data comprise sub-characteristic data in the crystal pulling process after the last crucible charging and sub-characteristic data before a temperature adjusting stage in the current Czochralski crystal process; any one characteristic data sample corresponding to the first set recommended model comprises sub-characteristic data in the crystal pulling process after the last crucible charging of the historical crucible charging and sub-characteristic data before the temperature regulating stage in the first straight-pulling monocrystal process after the historical crucible charging.

4. The method according to claim 2, wherein the current czochralski single crystal process is a non-first czochralski single crystal process after the current crucible charge; the characteristic data comprise sub-characteristic data in the last crystal pulling process after the current crucible is charged and sub-characteristic data before a temperature adjusting stage in the current Czochralski crystal pulling process; any one characteristic data sample corresponding to the cycle setting recommendation model comprises sub-characteristic data in the last pulling process of the historical Czochralski single crystal process after the historical crucible is filled, and sub-characteristic data before a temperature adjusting stage in the historical Czochralski single crystal process.

5. The method of claim 1, wherein prior to said entering the feature data into a set recommendation model, the method further comprises:

6. The method of claim 1, wherein prior to said entering the feature data into a set recommendation model, the method further comprises:

7. The method of claim 6, wherein the set recommendation model includes at least two types; training the set recommended model until the loss value of the set recommended model is smaller than the set loss value, and obtaining the trained set recommended model after the characteristic data sample, the marked sample temperature and the marked sample seeding power are adopted, wherein the method further comprises:

8. A parameter setting apparatus, comprising:

9. The electronic equipment is characterized by comprising a processor, a communication interface, a memory and a communication bus, wherein the processor, the communication interface and the memory are communicated with each other through the communication bus;

A memory for storing a computer program;

a processor for carrying out the method steps of any one of claims 1-7 when executing a program stored on a memory.

10. A readable storage medium, characterized in that instructions in the storage medium, when executed by a processor of an electronic device, enable the electronic device to perform the parameter setting method according to one or more of the method claims 1-7.