Big data-based crystallization detection method, system, device and storage medium

Abstract

A crystallization detection method, system, device and storage medium based on big data, which comprises the following steps: processing, screening and converting basic source data of a crystallization node in a constant diameter and ending process of a single crystal such as a Czochralski single crystal into a plurality of data sets which are easy to identify and mark in the crystallization node and establishing a model, performing multi-dimensional data cleaning and establishing a dimensional data warehouse; obtaining basic source data of a current node and converting the basic source data into process parameters, comparing the process parameters with the model in the dimensional data warehouse, analyzing data of a judgment result, judging whether there is a crystallization abnormality in a current link, and processing according to the judgment result. The technical scheme of the present application can use a crystallization identification function to detect crystallization abnormality in the constant diameter and ending process when the crystallization abnormality occurs in the constant diameter and ending process of the single crystal pulling, reduce the workload and time of personnel inspection and supervision, improve work efficiency and production yield, and prevent abnormal accidents.

Description

Technical Field

[0001] This invention belongs to the field of photovoltaic monocrystalline silicon pulling technology, and in particular relates to crystallization detection methods, systems, equipment and storage media based on big data. Background Technology

[0002] The Czochralski (CZ) single crystal growth process mainly includes steps such as temperature stabilization, crystal introduction, shoulder formation, diameter equalization, and finishing. During the diameter equalization and finishing processes, crystallization can occur, but currently, it cannot be identified in a timely manner. Therefore, on-site supervision by staff is required to inspect each furnace to ensure timely detection and handling. However, this detection method is very labor-intensive, increases the workload of staff, and makes it difficult to promptly detect and address every furnace with abnormalities during inspections.

[0003] Therefore, in order to promptly identify crystallization anomalies during the equal diameter and finishing processes of single crystal pulling, and to proceed with the next step normally or issue an alarm based on the identification results, this invention provides a crystallization detection method, system, device, and storage medium based on big data. Using crystallization identification functions and models, it detects crystallization anomalies during the equal diameter and finishing processes, reducing the workload and time of staff supervision, improving work efficiency and production output, and preventing abnormal accidents. Summary of the Invention

[0004] The problem this invention aims to solve is to provide a crystallization detection method, system, equipment, and storage medium based on big data, especially suitable for solar Czochralski silicon single crystal production. It effectively addresses the problem mentioned in the background art that the current existing technology, which requires staff to inspect and supervise each furnace on-site to find and deal with abnormalities, is very wasteful of manpower, increases the workload of staff, and makes it difficult to promptly detect and deal with each furnace with abnormalities during the inspection process.

[0005] To solve the above-mentioned technical problems, the present invention adopts the following technical solution:

[0006] A crystallization detection method based on big data, the method comprising the following steps:

[0007] S1: Obtain basic source data on the crystallization nodes of each different furnace type, each different series, and each different furnace platform during the equal diameter and finishing process of Czochralski single crystal pulling;

[0008] S2: Process the acquired basic source data, filter and convert it into several parameters that are easy to identify and mark in the crystallization nodes of each different furnace type, each different series, and each different furnace platform, and obtain a dataset of all the parameter values ​​of the crystallization nodes of each different furnace type, each different series, and each different furnace platform.

[0009] S3: Build a model for each parameter in the crystallization node of each different furnace type, each different series, and each different furnace platform through deep learning;

[0010] S4: Analyze, calculate and fit the model in step S3 using deep learning to obtain the normal crystal rod position range, normal crystal rod diameter range and normal furnace spot brightness range during the single crystal pulling process with equal diameter and the final crystallization process.

[0011] S5: Analyze and calculate each model in step S3 using deep learning to obtain basic source data on the crystallization node range, crystallization diameter range, and furnace spot brightness range of the current furnace type, current series, and current furnace platform.

[0012] S6: Process the basic source data of the normal crystal rod position range, normal crystal rod diameter range, and normal furnace spot brightness range obtained in step S5, filter and convert them into easily identifiable and labelable process parameters of the crystal rod position range, crystal rod diameter range, and furnace spot brightness range in the crystallization node of the current furnace type, current series, and current furnace platform.

[0013] S7: Compare the easily identifiable and markable process parameters of the crystal rod position range, crystal rod diameter range, and furnace spot brightness range mentioned in step S6 with the normal crystal rod position range, normal crystal rod diameter range, and normal furnace spot brightness range model mentioned in step S4. Based on the comparison results, determine whether the easily identifiable and markable process parameter values ​​of the crystal rod position range, crystal rod diameter range, and furnace spot brightness range in the crystallization node where the single crystal is located are reasonable.

[0014] S8: Perform data analysis on the judgment results in step S7 using deep learning, return the detection value, determine whether a crystallization abnormality has occurred in the current process based on the detection value, and output an alarm or continue the process based on the judgment result.

[0015] Furthermore, in step S2, each parameter in the crystallization node of each different furnace type, each different series, and each different furnace platform corresponds to the type of all the process parameters in step S6.

[0016] Furthermore, the parameters are established based on the production area, the location of crystal formation, and the characteristics of crystal size.

[0017] Furthermore, all of these parameters are configured and displayed on the terminal display of the single crystal furnace.

[0018] Furthermore, the basic source data for the crystallization nodes of each different furnace type, each different series, and each different furnace platform includes production process data and / or raw material data and / or quality data.

[0019] A crystallization detection system, the system comprising:

[0020] Source data acquisition unit: used to acquire basic source data of crystallization nodes for each different furnace type, each different series, and each different furnace platform during the equal diameter and finishing process of Czochralski single crystal pulling;

[0021] Source data processing unit: Processes the acquired basic source data, filters and converts it into several parameters that are easy to identify and mark in the crystallization nodes of each different furnace type, each different series, and each different furnace platform, and obtains a dataset of all the parameter values ​​of the crystallization nodes of each different furnace type, each different series, and each different furnace platform.

[0022] Model building unit: used to build a model for each parameter in the crystallization node of each different furnace type, each different series, and each different furnace platform through deep learning;

[0023] Data cleaning unit: used to perform multi-dimensional data cleaning on each of the models and establish a dimensional data warehouse for the Czochralski single crystal constant diameter and finishing process;

[0024] Data comparison unit: used to compare the process parameters in the crystallization node of the current furnace type, current series, and current furnace platform with each of the aforementioned models in the crystallization node of each different furnace type, each different series, and each different furnace platform;

[0025] Big Data Platform Unit: This unit performs big data analysis on the judgment results of comparing the process parameters in the crystallization node of the current furnace type, series, and furnace platform with each of the aforementioned models in the crystallization node of each different furnace type, series, and furnace platform. It returns the detection value, determines whether a crystallization abnormality has occurred in the current process based on the detection value, and outputs an alarm or continues to execute the process based on the judgment result.

[0026] Furthermore, each parameter in the crystallization node of each different furnace type, each different series, and each different furnace platform in the source data acquisition unit corresponds to all the process parameter types in the data processing unit;

[0027] The parameters are established based on the production area, the location of crystal formation, and the characteristics of crystal size;

[0028] All of the parameters are configured and displayed on the terminal display of the single crystal furnace.

[0029] Furthermore, the basic source data for the crystallization nodes of each different furnace type, each different series, and each different furnace platform includes production process data and / or raw material data and / or quality data.

[0030] A computer device includes a memory and a processor; the memory stores a computer program; the processor is configured to execute the computer program, and when executing the computer program, cause the processor to perform the steps of the crystallization detection method as described in any of the preceding claims.

[0031] A computer-readable storage medium storing a computer program that, when executed by a processor, causes the processor to perform the steps of the crystallization detection method as described in any of the preceding claims.

[0032] Compared with existing technologies, the crystallization detection method, system, equipment, and storage medium based on big data designed in this invention process, filters, and converts the basic source data of crystallization nodes for each different furnace type, series, and furnace platform during the equal diameter and finishing processes of Czochralski single crystal growth into a dataset of easily identifiable and labelable parameter values ​​corresponding to those in the model for each different furnace type, series, and furnace platform. Simultaneously, deep learning is used to build a model for each parameter in the crystallization nodes of each different furnace type, series, and furnace platform, and multi-dimensional data cleaning is performed on each model to establish a direct crystallization detection system. The dimensional data warehouse for single crystal pulling, including the equal diameter and finishing processes, calculates and obtains the current basic source data of the crystallization nodes of the current furnace type, series, and furnace platform. It then filters and converts the data into easily identifiable and labelable process parameters for the crystallization nodes of the current furnace type, series, and furnace platform. This data is compared with each model in the dimensional data warehouse to determine whether the values ​​of the easily identifiable and labelable process parameters in the node where the single crystal is located are reasonable. The determination results are analyzed using deep learning, and detection values ​​are returned. Based on the detection values, it is determined whether a crystallization abnormality has occurred in the current process, and an alarm is output or the process continues to be executed based on the determination results.

[0033] The technical solution of this invention can detect the crystallization situation during the equal diameter and finishing processes of single crystal pulling. Once an abnormal crystallization situation occurs, it can automatically make a judgment in a timely manner, return the detection value, determine whether an abnormal crystallization has occurred in the current process based on the detection value, and output an alarm or continue to execute the process based on the judgment result. This reduces the workload and time of staff to inspect and supervise, improves work efficiency and production output, and prevents the occurrence of abnormal accidents. Attached Figure Description

[0034] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0035] Figure 1 This is a flowchart of a crystallization detection method based on big data according to an embodiment of the present invention;

[0036] Figure 2 This is a schematic diagram of the structure of a crystallization detection system according to an embodiment of the present invention; Detailed Implementation

[0037] The present invention will be further described below with reference to the embodiments and accompanying drawings.

[0038] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to specific embodiments and the accompanying drawings. It should be understood that these descriptions are merely exemplary and not intended to limit the scope of the invention. Furthermore, descriptions of well-known structures and techniques are omitted in the following description to avoid unnecessarily obscuring the concept of the invention.

[0039] like Figure 1 As shown, this embodiment of the invention provides a crystallization detection method based on big data, the method comprising the following steps:

[0040] S1: Obtain basic source data on the crystallization nodes of each different furnace type, each different series, and each different furnace platform during the equal diameter and finishing process of Czochralski single crystal pulling;

[0041] Specifically, in the process of Czochralski single crystal growth with equal diameter and finalization, each single crystal furnace has individual characteristics at each crystallization node of different furnace types, different series, and different furnace platforms. The basic source data for each crystallization node of different furnace types, different series, and different furnace platforms includes production process data and / or raw material data and / or quality data.

[0042] The production process data includes equipment name, start and end time, batch number, process mode, formula name, diameter measurement value, hot zone temperature value, main heater power measurement, bottom heater power measurement, and actual crystal pulling speed.

[0043] Raw material data includes preparation date, batching sequence number, personnel shift, furnace number, workpiece specifications, crucible type, crucible origin, weight of primary polycrystalline material, percentage of recycled material, and overall weight.

[0044] Quality data includes single crystal number, length, weight, diameter, resistivity, lifetime, oxygen content, carbon content, defects, etc.

[0045] S2: Process the acquired basic source data, filter and convert it into several parameters that are easy to identify and mark in the crystallization nodes of each different furnace type, each different series, and each different furnace platform, and obtain a dataset of all the parameter values ​​of the crystallization nodes of each different furnace type, each different series, and each different furnace platform.

[0046] Specifically, the basic source data is processed, filtered, and converted into several parameters that are easy to identify and mark in the crystallization nodes of each different furnace type, each different series, and each different furnace platform. This results in a dataset of all parameter values ​​for the crystallization nodes of each different furnace type, each different series, and each different furnace platform. In other words, the scattered, messy, and inconsistent source data in the input basic source data is integrated and then converted into a dataset of commonly used parameters in the workpiece process nodes, providing a basis for subsequent parameter comparison and judgment analysis.

[0047] Furthermore, all parameters are established based on the production area, the location of crystal formation, and the characteristics of crystal size.

[0048] Furthermore, all parameters are configured and displayed on the terminal display of the single crystal furnace.

[0049] S3: Build a model for each parameter in the crystallization node of each different furnace type, each different series, and each different furnace platform through deep learning;

[0050] Specifically, a model is built for each parameter in the crystallization node of each different furnace type, each different series, and each different furnace platform using deep learning methods. This is to monitor and analyze the nodes of all workpieces in all furnace types, series, and furnace platforms during the equal diameter and finishing processes, in order to obtain single crystal workpieces that meet the quality standards.

[0051] S4: Analyze, calculate and fit each model in step S3 using deep learning to obtain the normal crystal rod position range, normal crystal rod diameter range and normal furnace spot brightness range during the single crystal pulling process with equal diameter and the final crystallization process.

[0052] Specifically, deep learning methods are used to analyze, calculate, and fit each model in step S3. The range of crystal rod position, crystal rod diameter, and range of brightness of the light spot in the furnace are combined in the crystallization nodes of each different furnace type, each different series, and each different furnace platform to obtain the normal range of crystal rod position, normal range of crystal rod diameter, and normal range of brightness of the light spot in the furnace.

[0053] S5: Analyze and calculate each model in step S3 using deep learning to obtain basic source data on the crystallization node range, crystallization diameter range, and furnace spot brightness range for the current furnace type, series, and furnace platform.

[0054] S6: Process the basic source data of normal crystal rod position range, normal crystal rod diameter range, and normal furnace spot brightness range obtained in step S5, filter and convert them into easily identifiable and labelable process parameters of crystal rod position range, crystal rod diameter range, and furnace spot brightness range in the crystallization node of the current furnace type, current series, and current furnace platform.

[0055] Furthermore, in step S2, each parameter in the crystallization node of each different furnace type, each different series, and each different furnace platform corresponds to all process parameter types in step S6.

[0056] S7: Compare the easily identifiable and markable process parameters of the crystal rod position range, crystal rod diameter range, and furnace spot brightness range in step S6 with the normal crystal rod position range, normal crystal rod diameter range, and normal furnace spot brightness range model in step S4. Based on the comparison results, determine whether the easily identifiable and markable process parameter values ​​of the crystal rod position range, crystal rod diameter range, and furnace spot brightness range in the crystallization node where the single crystal is located are reasonable.

[0057] S8: Perform data analysis on the judgment results in step S7 using deep learning, return the detection value, determine whether a crystallization abnormality has occurred in the current process based on the detection value, and output an alarm or continue the process based on the judgment result.

[0058] Specifically, each data identification and analysis cycle includes a detection period and a mapping period. The detection period includes the equal diameter and the finishing process. During the mapping period, the CCD program sends an equal diameter finishing process image and related information to the server every 10 seconds for detection.

[0059] Specifically, for each image, the returned detection value can be 0 or 1, where 0 indicates a normal state with no crystallization abnormalities, and 1 indicates a crystallization abnormality. Then, based on the detection value's determination, it is decided whether to issue an alarm or continue the process.

[0060] A crystallization detection system, such as Figure 2 As shown, the system includes:

[0061] Source data acquisition unit: used to acquire basic source data of crystallization nodes for each different furnace type, each different series, and each different furnace platform during the equal diameter and finishing process of Czochralski single crystal pulling;

[0062] Source data processing unit: Process the acquired basic source data, filter and convert it into several parameters that are easy to identify and mark in the crystallization nodes of each different furnace type, each different series, and each different furnace platform, and obtain a dataset of all parameter values ​​of the crystallization nodes of each different furnace type, each different series, and each different furnace platform.

[0063] Model building unit: used to build a model for each parameter in the crystallization node of each different furnace type, each different series, and each different furnace platform through deep learning;

[0064] Data cleaning unit: used to perform multi-dimensional data cleaning on each of the models and establish a dimensional data warehouse for the Czochralski single crystal constant diameter and finishing process;

[0065] Data comparison unit: used to compare the process parameters in the crystallization node of the current furnace type, current series, and current furnace platform with each model in the crystallization node of each different furnace type, each different series, and each different furnace platform;

[0066] Big Data Platform Unit: This unit performs big data analysis on the judgment results of comparing the process parameters in the crystallization node of the current furnace type, series, and platform with each model in the crystallization node of each different furnace type, series, and platform. It returns the detection value, determines whether a crystallization abnormality has occurred in the current process based on the detection value, and outputs an alarm or continues to execute the process based on the judgment result.

[0067] Furthermore, each parameter in the crystallization node of each different furnace type, each different series, and each different furnace platform in the source data unit is matched with the corresponding process parameter type in the data processing unit;

[0068] The parameters are established based on the production area, the location of crystal formation, and the characteristics of crystal size;

[0069] All parameters are configured and displayed on the terminal display of the single crystal furnace.

[0070] Furthermore, the basic source data for the crystallization nodes of each different furnace type, each different series, and each different furnace platform includes production process data and / or raw material data and / or quality data.

[0071] A computer device includes a memory and a processor; the memory stores a computer program; the processor is configured to execute the computer program, and when executing the computer program, causes the processor to perform the steps of the crystallization detection method as described above.

[0072] A computer-readable storage medium storing a computer program that, when executed by a processor, causes the processor to perform the steps of any of the above crystal detection methods.

[0073] The advantages and beneficial effects of this invention are:

[0074] 1. This invention designs a crystallization detection method, system, device, and storage medium based on big data. It processes, filters, and converts the basic source data of crystallization nodes for each different furnace type, series, and furnace platform during the equal-diameter and final stages of Czochralski single crystal growth into a dataset of easily identifiable and labelable parameter values ​​corresponding to those in the model. Simultaneously, it uses deep learning to build a model for each parameter in the crystallization nodes of each different furnace type, series, and furnace platform, and performs multi-dimensional data cleaning on each model to establish a model for Czochralski single crystal growth, etc. The dimension data warehouse for the process path and finishing stages calculates and obtains the current basic source data of the crystallization nodes of the current furnace type, series, and furnace platform. It filters and converts the data into easily identifiable and tagged process parameters for the crystallization nodes of the current furnace type, series, and furnace platform, and compares the data with each model in the dimension data warehouse to determine whether the values ​​of the easily identifiable and tagged process parameters in the node where the single crystal is located are reasonable. The determination results are analyzed by deep learning, and a detection value is returned. Based on the detection value, it is determined whether a crystallization abnormality has occurred in the current process, and an alarm is output or the process continues to be executed based on the determination result.

[0075] 2. The technical solution of this invention can use the crystallization identification function to detect crystallization abnormalities during the equal diameter and finishing processes of single crystal pulling. This reduces the workload and time of staff for inspection and supervision, improves work efficiency and production output, and prevents abnormal accidents from occurring.

[0076] It should be understood that the specific embodiments described above are merely illustrative or explanatory of the principles of the invention and do not constitute a limitation thereof. Therefore, any modifications, equivalent substitutions, improvements, etc., made without departing from the spirit and scope of the invention should be included within the protection scope of the invention. Furthermore, the appended claims are intended to cover all variations and modifications falling within the scope and boundaries of the appended claims, or equivalent forms of such scope and boundaries.

Claims

1. A crystallization detection method based on big data, characterized in that, The method includes the following steps: S1: Obtain basic source data of crystallization nodes for each different furnace type, each different series, and each different furnace platform during the equal diameter and finishing process of Czochralski single crystal pulling. The basic source data includes production process data and / or raw material data and / or quality data. S2: Process the acquired basic source data, filter and convert it into parameters that are easy to identify and mark in the crystallization nodes of each different furnace type, each different series, and each different furnace platform, and obtain a dataset of all the parameter values ​​of the crystallization nodes of each different furnace type, each different series, and each different furnace platform. S3: Build a model for each parameter in the crystallization node of each different furnace type, each different series, and each different furnace platform through deep learning; S4: Analyze, calculate and fit the model in step S3 using deep learning to obtain the normal crystal rod position range, normal crystal rod diameter range and normal furnace spot brightness range during the single crystal pulling process with equal diameter and the final crystallization process. S5: Analyze and calculate each model in step S3 using deep learning to obtain basic source data on the crystallization node range, crystallization diameter range, and furnace spot brightness range of the current furnace type, current series, and current furnace platform. S6: Process the basic source data of the crystal rod position range, crystal rod diameter range, and furnace spot brightness range obtained in step S5, filter and convert them into easily identifiable and labelable process parameters of the crystal rod position range, crystal rod diameter range, and furnace spot brightness range in the crystallization node of the current furnace type, current series, and current furnace platform. S7: Compare the easily identifiable and markable process parameters of the crystal rod position range, crystal rod diameter range, and furnace spot brightness range mentioned in step S6 with the normal crystal rod position range, normal crystal rod diameter range, and normal furnace spot brightness range model mentioned in step S4. Based on the comparison results, determine whether the easily identifiable and markable process parameter values ​​of the crystal rod position range, crystal rod diameter range, and furnace spot brightness range in the crystallization node where the single crystal is located are reasonable. S8: Perform data analysis on the judgment results in step S7 using deep learning, return the detection value, determine whether a crystallization abnormality has occurred in the current process based on the detection value, and output an alarm or continue the process based on the judgment result.

2. The crystallization detection method based on big data according to claim 1, characterized in that: In step S2, each parameter in the crystallization node of each different furnace type, each different series, and each different furnace platform corresponds to the type of all the process parameters in step S6.

3. The crystallization detection method based on big data according to claim 2, characterized in that: The parameters are established based on the production area, the location of crystal formation, and the characteristics of crystal size.

4. The crystallization detection method based on big data according to claim 3, characterized in that: All of the parameters are configured and displayed on the terminal display of the single crystal furnace.

5. A crystallization detection system, characterized in that, The system includes: Source data acquisition unit: used to acquire basic source data of crystallization nodes of each different furnace type, each different series, and each different furnace platform during the Czochralski single crystal equal diameter and finishing process. The basic source data includes production process data and / or raw material data and / or quality data. Source data processing unit: Processes the acquired basic source data, filters and converts it into parameters that are easy to identify and mark in the crystallization nodes of each different furnace type, each different series, and each different furnace platform, and obtains a dataset of all parameter values ​​of the crystallization nodes of each different furnace type, each different series, and each different furnace platform. Model building unit: used to build a model for each parameter in the crystallization node of each different furnace type, each different series, and each different furnace platform through deep learning; Data cleaning unit: used to perform multi-dimensional data cleaning on each of the models and establish a dimensional data warehouse for the Czochralski single crystal constant diameter and finishing process; Data comparison unit: used to compare the process parameters in the crystallization node of the current furnace type, current series, and current furnace platform with each of the aforementioned models in the crystallization node of each different furnace type, each different series, and each different furnace platform; Big Data Platform Unit: This unit performs big data analysis on the judgment results of comparing the process parameters in the crystallization node of the current furnace type, series, and furnace platform with each of the aforementioned models in the crystallization node of each different furnace type, series, and furnace platform. It returns the detection value, determines whether a crystallization abnormality has occurred in the current process based on the detection value, and outputs an alarm or continues to execute the process based on the judgment result.

6. The crystallization detection system according to claim 5, characterized in that: Each parameter in the crystallization node of each different furnace type, each different series, and each different furnace platform in the source data acquisition unit corresponds to all the process parameter types in the data processing unit; The parameters are established based on the production area, the location of crystal formation, and the characteristics of crystal size; All of the parameters are configured and displayed on the terminal display of the single crystal furnace.

7. A computer device, characterized in that: It includes a memory and a processor; the memory stores a computer program; the processor is used to execute the computer program, and when executing the computer program, causes the processor to perform the steps of the crystallization detection method as described in any one of claims 1-4.

8. A computer-readable storage medium, characterized in that: The device contains a computer program that, when executed by a processor, causes the processor to perform the steps of the crystallization detection method as described in any one of claims 1-4.

Images