CN111755079B - Method and system for determining optimal raw material proportioning scheme of polycrystalline silicon - Google Patents

Method and system for determining optimal raw material proportioning scheme of polycrystalline silicon Download PDF

Info

Publication number
CN111755079B
CN111755079B CN202010639708.7A CN202010639708A CN111755079B CN 111755079 B CN111755079 B CN 111755079B CN 202010639708 A CN202010639708 A CN 202010639708A CN 111755079 B CN111755079 B CN 111755079B
Authority
CN
China
Prior art keywords
raw material
material proportioning
data
proportioning scheme
scheme
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Active
Application number
CN202010639708.7A
Other languages
Chinese (zh)
Other versions
CN111755079A (en
Inventor
李凤莲
张龙
张雪英
黄丽霞
焦江丽
陈桂军
史凯岳
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Taiyuan University of Technology
Original Assignee
Taiyuan University of Technology
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Taiyuan University of Technology filed Critical Taiyuan University of Technology
Priority to CN202010639708.7A priority Critical patent/CN111755079B/en
Publication of CN111755079A publication Critical patent/CN111755079A/en
Application granted granted Critical
Publication of CN111755079B publication Critical patent/CN111755079B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Classifications

    • G—PHYSICS
    • G16—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR SPECIFIC APPLICATION FIELDS
    • G16C—COMPUTATIONAL CHEMISTRY; CHEMOINFORMATICS; COMPUTATIONAL MATERIALS SCIENCE
    • G16C20/00—Chemoinformatics, i.e. ICT specially adapted for the handling of physicochemical or structural data of chemical particles, elements, compounds or mixtures
    • G16C20/70—Machine learning, data mining or chemometrics
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
    • G06Q10/00—Administration; Management
    • G06Q10/04—Forecasting or optimisation specially adapted for administrative or management purposes, e.g. linear programming or "cutting stock problem"
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
    • G06Q50/00—Information and communication technology [ICT] specially adapted for implementation of business processes of specific business sectors, e.g. utilities or tourism
    • G06Q50/04—Manufacturing
    • G—PHYSICS
    • G16—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR SPECIFIC APPLICATION FIELDS
    • G16C—COMPUTATIONAL CHEMISTRY; CHEMOINFORMATICS; COMPUTATIONAL MATERIALS SCIENCE
    • G16C60/00—Computational materials science, i.e. ICT specially adapted for investigating the physical or chemical properties of materials or phenomena associated with their design, synthesis, processing, characterisation or utilisation
    • Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
    • Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
    • Y02P—CLIMATE CHANGE MITIGATION TECHNOLOGIES IN THE PRODUCTION OR PROCESSING OF GOODS
    • Y02P90/00—Enabling technologies with a potential contribution to greenhouse gas [GHG] emissions mitigation
    • Y02P90/30—Computing systems specially adapted for manufacturing

Landscapes

  • Engineering & Computer Science (AREA)
  • Business, Economics & Management (AREA)
  • Theoretical Computer Science (AREA)
  • Computing Systems (AREA)
  • Strategic Management (AREA)
  • Human Resources & Organizations (AREA)
  • Economics (AREA)
  • Marketing (AREA)
  • Health & Medical Sciences (AREA)
  • General Physics & Mathematics (AREA)
  • General Business, Economics & Management (AREA)
  • Physics & Mathematics (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Bioinformatics & Cheminformatics (AREA)
  • Bioinformatics & Computational Biology (AREA)
  • General Health & Medical Sciences (AREA)
  • Tourism & Hospitality (AREA)
  • Game Theory and Decision Science (AREA)
  • Quality & Reliability (AREA)
  • Data Mining & Analysis (AREA)
  • Medical Informatics (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Artificial Intelligence (AREA)
  • Entrepreneurship & Innovation (AREA)
  • Databases & Information Systems (AREA)
  • Operations Research (AREA)
  • Development Economics (AREA)
  • Evolutionary Computation (AREA)
  • Crystallography & Structural Chemistry (AREA)
  • Chemical & Material Sciences (AREA)
  • Software Systems (AREA)
  • Manufacturing & Machinery (AREA)
  • Primary Health Care (AREA)
  • Crystals, And After-Treatments Of Crystals (AREA)

Abstract

The invention discloses a method and a system for determining an optimal raw material proportioning scheme of polycrystalline silicon. The invention can rapidly and efficiently screen out better raw material proportioning combination from hundreds to thousands of raw material proportioning combinations, and improves the production quality and economic benefit of polysilicon.

Description

Method and system for determining optimal raw material proportioning scheme of polycrystalline silicon
Technical Field
The invention relates to the technical field of polysilicon production, in particular to a method and a system for determining an optimal raw material proportioning scheme of polysilicon.
Background
Since the 21 st century, the photovoltaic industry has become one of the fastest growing high-tech industries in the world. Among the various solar cells, polysilicon solar cells are extremely important and share a huge share in the photovoltaic market. In the actual production process of polysilicon, the raw materials have great influence on the quality of the product. If all the raw materials with higher price are used, the product quality is better overall, but the cost is relatively higher; if the raw materials with lower price are used in total, the cost is relatively low, but the product quality is poor as a whole. Therefore, the existing actual production process of the polysilicon can not ensure the product quality and reduce the production cost.
Disclosure of Invention
The invention aims to provide a method and a system for determining an optimal raw material proportioning scheme of polycrystalline silicon, which are used for solving the problems that the product quality cannot be ensured and the production cost is reduced in the conventional actual production process of polycrystalline silicon.
In order to achieve the above object, the present invention provides the following solutions:
a method for determining an optimal raw material proportioning scheme of polycrystalline silicon comprises the following steps:
obtaining a plurality of groups of raw material proportioning schemes in the production process of polysilicon; the raw material proportioning scheme comprises a plurality of different raw material data; the raw material data comprise raw polycrystal lump materials, circulating materials, broken polycrystal materials, purified ingot cores, top skins, rim charge and tailing materials;
calculating the similarity between each group of the raw material proportioning scheme and each of the plurality of groups of the raw material proportioning schemes;
establishing a similarity matrix according to the similarity;
determining polycrystalline silicon quality test data according to the raw material proportioning scheme; the polycrystalline silicon quality test data comprise a head removing length, a tail removing length, a minority carrier lifetime value, a head resistor, a tail resistor, a seed crystal thickness, a minority carrier yield, an infrared reject ratio and a final yield;
adding the polycrystalline silicon quality test data into the similarity matrix to generate mixed data;
Clustering the mixed data by adopting a clustering algorithm to generate k mixed data feature classes;
and evaluating a plurality of raw material proportioning schemes to be tested according to the k mixed data characteristic classes, and determining an optimal raw material proportioning scheme.
Optionally, the calculating the similarity between each group of the raw material proportioning schemes and each group of the raw material proportioning schemes in the multiple groups of the raw material proportioning schemes further includes:
and numbering each raw material data in the raw material proportioning scheme, and replacing the name of the raw material data with a numbered number.
Optionally, the calculating the similarity between each raw material proportioning scheme and each raw material proportioning scheme in the plurality of groups of raw material proportioning schemes specifically includes:
using the formulaCalculating the similarity p (x) between the ith raw material proportioning scheme and the jth raw material proportioning scheme i ,x j ) The method comprises the steps of carrying out a first treatment on the surface of the Wherein x is i Represents the ith raw material proportioning scheme, x j Represents the j-th raw material proportioning scheme, m represents the dimension of the raw material data, 1 is less than or equal to i, j is less than or equal to n, n is the number of the obtained multiple groups of raw material proportioning schemes, and theta d (x i ，x j ) Representing the orthogonality value of the ith and jth feedstock formulation in the d-th dimension, whereinx id ＝x jd The d attribute value representing the ith raw material proportioning scheme and the jth raw material proportioning scheme is the same, x id ≠x jd The d attribute values of the ith raw material proportioning scheme and the jth raw material proportioning scheme are different, and d is more than or equal to 1 and less than or equal to m; x is x id The d attribute value of the ith raw material proportioning scheme; x is x jd The d attribute value of the j-th raw material proportioning scheme.
Optionally, clustering the hybrid data by using a clustering algorithm to generate k hybrid data feature classes, which specifically includes:
selecting a 1 st original feature center in the mixed data by using a minimum maximum distance method;
calculating the distance between each datum in the mixed data and the 1 st original feature center, and selecting the point with the largest distance with the 1 st original feature center as the 2 nd original feature center;
calculating the distance between each data in the mixed data and the 1 st original feature center and the 2 nd original feature center respectively to obtain a first maximum distance farthest from the 1 st original feature center and a second maximum distance farthest from the 2 nd original feature center, and comparing the points corresponding to the minimum value selection values of the first maximum distance and the second maximum distance as the 3 rd original feature center until k original feature centers are selected;
distributing the mixed data to k class clusters which are closest to k original feature centers;
Taking the average value of the attributes of the mixed data in the k class clusters as an updated feature center; one of the original feature centers corresponds to only one of the updated feature centers;
determining a distance between the original feature center and the updated feature center according to the original feature center and the updated feature center;
acquiring the current iteration times;
judging whether the distance is greater than a distance threshold value or not to obtain a first judgment result;
if the first judgment result shows that the distance is larger than a distance threshold, continuing to judge whether the current iteration number is smaller than the iteration number threshold or not, and obtaining a second judgment result;
and if the second judgment result shows that the current iteration number is smaller than an iteration number threshold, returning to the step of distributing the mixed data to k class clusters which are closest to k original feature centers, until the distance is smaller than a distance threshold, or if the current iteration number is larger than the iteration number threshold, outputting the current feature class cluster, and taking the current feature class cluster as the k mixed data feature classes.
Optionally, the evaluating the multiple raw material proportioning schemes to be tested according to the k mixed data feature classes to determine an optimal raw material proportioning scheme specifically includes:
reducing the similarity matrix in the k mixed data feature classes into a raw material proportioning scheme to obtain k raw material feature classes; one of the feedstock characteristics includes a plurality of feedstock proportioning schemes;
calculating the average value of the polycrystalline silicon quality test data in k raw material characteristic classes, and taking the average value of the polycrystalline silicon quality test data as k identification values of k raw material characteristic classes;
dividing the grades of k raw material characteristic classes according to k identification values; the raw material feature class rating includes: normal class, worse class, and abnormal class;
and evaluating a plurality of to-be-tested polycrystalline silicon raw material proportioning schemes according to the grades, and determining an optimal raw material proportioning scheme.
Optionally, the evaluating the multiple raw material proportioning schemes to be tested according to the evaluation level, and determining the optimal raw material proportioning scheme specifically includes:
acquiring a raw material proportioning scheme to be tested;
calculating the similarity of the to-be-detected raw material proportioning scheme and each raw material proportioning scheme in each raw material characteristic class, and determining the average similarity of the to-be-detected raw material proportioning scheme and each raw material characteristic class;
And comparing the k average similarity, determining the raw material characteristic class corresponding to the maximum average similarity, acquiring the rating of the raw material characteristic class corresponding to the maximum average similarity, and taking the rating of the raw material characteristic class corresponding to the maximum average similarity as the rating of the to-be-detected polycrystalline silicon raw material proportioning scheme.
A polysilicon optimal raw material proportioning scheme determining system comprises:
the raw material proportioning scheme acquisition module is used for acquiring a plurality of groups of raw material proportioning schemes in the production process of the polysilicon; the raw material proportioning scheme comprises a plurality of raw material data; the raw material data comprise raw polycrystal blocks, circulating materials, broken polycrystal blocks, purified ingot cores, top skins, rim charge and tails;
the similarity determining module is used for calculating the similarity between each raw material proportioning scheme and each raw material proportioning scheme in the plurality of groups of raw material proportioning schemes;
the similarity matrix establishing module is used for establishing a similarity matrix according to the similarity;
the polycrystalline silicon quality test data determining module is used for determining polycrystalline silicon quality test data according to the raw material proportioning scheme; the polycrystalline silicon quality test data comprise a head removing length, a tail removing length, a minority carrier lifetime value, a head resistor, a tail resistor, a seed crystal thickness, a minority carrier yield, an infrared reject ratio and a final yield;
The mixed data generation module is used for adding the polycrystalline silicon quality test data into the similarity matrix to generate mixed data;
the clustering module is used for clustering the mixed data by adopting a clustering algorithm to generate k mixed data feature classes;
and the raw material proportioning scheme evaluation module is used for evaluating a plurality of raw material proportioning schemes to be tested according to the k mixed data characteristic classes to determine an optimal raw material proportioning scheme.
Optionally, the method further comprises:
and the raw material data numbering module is used for numbering the raw material data in the raw material proportioning scheme, and the name of the raw material data is replaced by the numbered number.
Optionally, the similarity determining module specifically includes:
a similarity determining unit for using the formulaCalculating the similarity p (x) between the ith raw material proportioning scheme and the jth raw material proportioning scheme i ,x j ) The method comprises the steps of carrying out a first treatment on the surface of the Wherein x is i Represents the ith raw material proportioning scheme, x j Represents the j-th raw material proportioning scheme, m represents the dimension of the raw material data, 1 is less than or equal to i, j is less than or equal to n, n is the number of the obtained multiple groups of raw material proportioning schemes, and theta d (x i ，x j ) Representing the orthogonality value of the ith and jth stock solutions in the d-th dimension, wherein ∈ > x id ＝x jd The d attribute value representing the ith raw material proportioning scheme and the jth raw material proportioning scheme is the same, x id ≠x jd The d attribute values of the ith raw material proportioning scheme and the jth raw material proportioning scheme are different, and d is more than or equal to 1 and less than or equal to m; x is x id The d attribute value of the ith raw material proportioning scheme; x is x jd The d attribute value of the j-th raw material proportioning scheme.
Optionally, the clustering module specifically includes:
the original feature center selecting unit is used for selecting the 1 st original feature center in the mixed data by using a minimum maximum distance method;
calculating the distance between each datum in the mixed data and the 1 st original feature center, and selecting the point with the largest distance with the 1 st original feature center as the 2 nd original feature center;
calculating the distance between each data in the mixed data and the 1 st original feature center and the 2 nd original feature center respectively to obtain a first maximum distance farthest from the 1 st original feature center and a second maximum distance farthest from the 2 nd original feature center, and comparing the points corresponding to the minimum value selection values of the first maximum distance and the second maximum distance as the 3 rd original feature center until k original feature centers are selected;
A feature cluster determining unit, configured to allocate the hybrid data to k clusters to which k original feature centers closest to k original feature centers belong;
the updated feature center selecting unit is used for taking the average value of the attributes of the mixed data in the k class clusters as an updated feature center; one of the original feature centers corresponds to only one of the updated feature centers;
a distance determining unit, configured to determine a distance between the original feature center and the updated feature center according to the original feature center and the updated feature center;
the iteration number acquisition unit is used for acquiring the current iteration number;
the judging and analyzing unit is used for judging whether the distance is larger than a distance threshold value or not to obtain a first judging result;
if the first judgment result shows that the distance is larger than a distance threshold, continuing to judge whether the current iteration number is smaller than the iteration number threshold or not, and obtaining a second judgment result;
and if the second judgment result shows that the current iteration number is smaller than an iteration number threshold, returning to the step of distributing the mixed data to k class clusters which are closest to k original feature centers, until the distance is smaller than a distance threshold, or if the current iteration number is larger than the iteration number threshold, outputting the current feature class cluster, and taking the current feature class cluster as the k mixed data feature classes.
According to the specific embodiment provided by the invention, the invention discloses the following technical effects:
the invention discloses a method and a system for determining an optimal raw material proportioning scheme of polycrystalline silicon. The invention can rapidly and efficiently screen out better raw material proportioning combination from hundreds to thousands of raw material proportioning combinations, and improves the production quality and economic benefit of polysilicon.
Drawings
In order to more clearly illustrate the embodiments of the present invention or the technical solutions of the prior art, the drawings that are needed in the embodiments will be briefly described below, it being obvious that the drawings in the following description are only some embodiments of the present invention, and that other drawings may be obtained according to these drawings without inventive effort for a person skilled in the art.
FIG. 1 is a flow chart of a method for determining an optimal raw material proportioning scheme of polycrystalline silicon;
FIG. 2 is a schematic diagram of a raw material data similarity transformation process provided by the invention;
FIG. 3 is a schematic diagram of a system for determining a polysilicon raw material proportioning scheme provided by the invention;
FIG. 4 is a functional flow diagram of a data analysis module according to the present invention;
FIG. 5 is a schematic functional flow diagram of a data monitoring module according to the present invention;
fig. 6 is a schematic diagram of a system for determining an optimal raw material proportioning scheme of polycrystalline silicon.
Detailed Description
The following description of the embodiments of the present invention will be made clearly and completely with reference to the accompanying drawings, in which it is apparent that the embodiments described are only some embodiments of the present invention, but not all embodiments. All other embodiments, which can be made by those skilled in the art based on the embodiments of the invention without making any inventive effort, are intended to be within the scope of the invention.
The invention aims to provide a method and a system for determining an optimal raw material proportioning scheme of polycrystalline silicon, which are used for solving the problems that the product quality cannot be ensured and the production cost is reduced in the conventional actual production process of polycrystalline silicon.
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.
Fig. 1 is a flowchart of a method for determining an optimal raw material proportioning scheme of polycrystalline silicon. As shown in fig. 1, a method for determining an optimal raw material proportioning scheme of polysilicon includes:
step 101: obtaining a plurality of groups of raw material proportioning schemes in the production process of polysilicon; the raw material proportioning scheme comprises a plurality of different raw material data; raw material data are primary polycrystal lump materials, circulating materials, broken polycrystal materials, purified ingot cores, top skins, rim charge materials, tails and the like, and each major class can be divided into a plurality of minor classes according to manufacturer and batch numbers;
in practical use, the raw material data are shown in table 1.
Table 1 raw material data example
The "raw material data (major class)" in table 1 refers to several types of raw materials frequently used in the production of polycrystalline silicon, each of which may be provided by a different manufacturer, and the "raw material data (minor class)" refers to lot numbers of raw materials provided by the same manufacturer or different manufacturers.
Step 102: calculating the similarity between each group of the raw material proportioning scheme and each of the plurality of groups of the raw material proportioning schemes;
using the formulaCalculating the similarity p (x) between the ith raw material proportioning scheme and the jth raw material proportioning scheme i ,x j ) The method comprises the steps of carrying out a first treatment on the surface of the Wherein x is i Represents the ith raw material proportioning scheme, x j Represents the j-th raw material proportioning scheme, m represents the dimension of the raw material data, 1 is less than or equal to i, j is less than or equal to n, n is the number of the obtained multiple groups of raw material proportioning schemes, and theta d (x i ，x j ) Representing the orthogonality of the ith and jth stock matching schemes in the d-th dimension, wherein,x id ＝x jd represents the ith raw material proportioning scheme and the jth raw material proportioning schemeThe d attribute values of the raw material proportioning scheme are the same, and x is id ≠x jd The d attribute values of the ith raw material proportioning scheme and the jth raw material proportioning scheme are different, and d is more than or equal to 1 and less than or equal to m; x is x id The d attribute value of the ith raw material proportioning scheme; x is x jd The d attribute value of the j-th raw material proportioning scheme.
In practical application, the calculating the similarity between each group of the raw material proportioning scheme and each group of the raw material proportioning schemes in the multiple groups of the raw material proportioning schemes further comprises:
and numbering each raw material data in the raw material proportioning scheme, and replacing the name of the raw material data with a numbered number. All raw material data used during a single production run were collected, and polysilicon raw material data are shown in table 2, for example, with one line of data in table 2 recording a portion of the raw material used during the production of a polysilicon ingot.
Table 2 polysilicon feedstock data example
The raw material data is numbered, and the invention replaces the raw material name with a number for subsequent processing, and a partial example of the correspondence between the raw material data and the number is shown in table 3. The symbolic data generated by substituting the raw material names in table 3 with numbers are shown in table 4, wherein the blank values in table 4 are filled with 0.
TABLE 3 raw material data and numbering correspondence examples
Raw material name Numbering device
WG190720-XKS-4021 1
WG190720-HJ-4021 2
WG190804-XKS-4021-1 3
WG190804-XKS-4021-2 4
WG190804-XKS-4021-3 5
WG190804-XKS-4021-4 6
WG190804-XKS-4021-5+6 7
…… ……
Table 4 raw material data examples indicated by numerals
Spindle number Raw material 1 Raw material 2 Raw material 3 Raw material 4 Raw material 5
1 1 35 36 41 46
2 1 35 36 41 46
3 1 35 36 41 46
4 2 35 36 41 46
5 2 35 0 41 46
6 2 35 0 41 46
7 2 35 36 41 46
And calculating the similarity of each data and the rest data, wherein the similarity calculation method is shown in a formula (1).
Where m represents the dimension of the data, x i Raw material data representing the ith product, where 1.ltoreq.i, j.ltoreq.n, n is the number of raw material data collected in step 101. θ d (x i ，x j ) The orthogonal value representing the d-th dimension of the product i and the product j is calculated as shown in the formula (2).
X in the above id ＝x jd The d attribute value representing the ith raw material proportioning scheme and the jth raw material proportioning scheme is the same, x id ≠x jd Representing the ith raw material proportioning scheme and the jth raw material proportioningThe d attribute values of the comparison schemes are different, and d is more than or equal to 1 and less than or equal to m; x is x id The d attribute value of the ith raw material proportioning scheme; x is x jd The d attribute value of the j-th raw material proportioning scheme.
Fig. 2 is a schematic diagram of a raw material data similarity transformation process provided by the invention. As shown in fig. 2, after the similarity transformation, the value range of the similarity vector is changed to be between 0 and 1, the smaller the value is, the lower the similarity of the raw materials used by the two products is, the larger the value is, the higher the similarity of the raw materials used by the two products is, and if the value is 1, the highest the similarity of the raw materials used by the two products is, that is, the raw materials used by the two products are identical.
For the raw material data in fig. 2, if a polycrystalline silicon ingot is used for 5 raw materials at most in the secondary production process, the transformed similarity vector has 6 discrete values of 0, 0.2, 0.4, 0.6, 0.8 and 1 respectively, wherein 0 indicates that the raw materials used by the two products are all different, 1 indicates that the raw materials used by the two products are all the same, and 0.2 indicates that only one of the raw materials used by the two products is the same. Similarly, if a silicon ingot is used for 10 raw materials at most in the secondary production process, the value of the similarity vector after transformation is 11 discrete values which are respectively 0, 0.1, 0.2, 0.3, 0.4, 0.5, 0.6, 0.7, 0.8, 0.9 and 1, and the value is similar to the case of 5 raw materials.
Step 103: establishing a similarity matrix according to the similarity;
step 104: determining polycrystalline silicon quality test data according to the raw material proportioning scheme; the polysilicon quality test data comprises a head removal length, a tail removal length, a minority carrier lifetime value, a head resistance, a tail resistance, a seed crystal thickness, a minority carrier yield, an infrared reject ratio and a final yield. Some of the test metrics for the polysilicon product are listed in table 5, each of which indicates how good the polysilicon product is in this regard. If the head removing length and the tail removing length are smaller, the more available parts of the silicon ingot are after being discharged from the furnace; the larger minority carrier lifetime value indicates fewer impurities in the silicon ingot, etc.
Table 5 test data examples
Length of head
Length of tail
Minority carrier lifetime value
Head resistor
Tail resistor
Seed crystal thickness
Rate of minority carrier recovery
Infrared reject ratio
Final yield of material
……
Step 105: adding the polycrystalline silicon quality test data into the similarity matrix to generate mixed data;
in practical application, the similarity vector is transformed to obtain a similarity matrix, and after the transformed similarity matrix is added with the polysilicon quality test data, the embodiment of the invention preferably uses minority carrier lifetime values to obtain mixed data, and as shown in a mixed data example in table 6, the data has n+1 dimensions, wherein n is the number of raw material data.
Table 6 mixed data example
Step 106: clustering the mixed data by adopting a clustering algorithm to generate k mixed data feature classes, wherein the method specifically comprises the following steps:
selecting a 1 st original feature center in the mixed data by using a minimum maximum distance method;
calculating the distance between each datum in the mixed data and the 1 st original feature center, and selecting the point with the largest distance with the 1 st original feature center as the 2 nd original feature center;
calculating the distance between each data in the mixed data and the 1 st original feature center and the 2 nd original feature center respectively to obtain a first maximum distance farthest from the 1 st original feature center and a second maximum distance farthest from the 2 nd original feature center, and comparing the points corresponding to the minimum value selection values of the first maximum distance and the second maximum distance as the 3 rd original feature center until k original feature centers are selected;
distributing the mixed data to k class clusters which are closest to k original feature centers;
taking the average value of the attributes of the mixed data in the k class clusters as an updated feature center; one of the original feature centers corresponds to only one of the updated feature centers;
Determining a distance between the original feature center and the updated feature center according to the original feature center and the updated feature center;
acquiring the current iteration times;
judging whether the distance is greater than a distance threshold value or not to obtain a first judgment result;
if the first judgment result shows that the distance is larger than a distance threshold, continuing to judge whether the current iteration number is smaller than the iteration number threshold or not, and obtaining a second judgment result;
and if the second judgment result shows that the current iteration number is smaller than an iteration number threshold, returning to the step of distributing the mixed data to k class clusters which are closest to k original feature centers, until the distance is smaller than a distance threshold, or if the current iteration number is larger than the iteration number threshold, outputting the current feature class cluster, and taking the current feature class cluster as the k mixed data feature classes.
In practical applications, the specific applications are as follows,
s1, initializing k original feature centers by using a minimum maximum distance method, and randomly selecting one data from the mixed data obtained in the step 5 as a 1 st original feature center; calculating the distance between the rest data and the 1 st original feature center, wherein the distance calculation method is shown in a formula (3), and selecting the data with the largest distance from the first original feature center as the 2 nd original feature center; the selection method of the remaining k-2 original feature centers is that the point with the largest distance from the existing original feature centers in the remaining data is calculated, the distance value is recorded, the point corresponding to the smallest distance value is selected as the new original feature center, and the selection is finished until the k original feature centers are selected.
X in the above i Represents the ith data, i is more than or equal to 1 and less than or equal to n, C j Represents the j-th feature center, k, x id The d attribute value, C, representing the i-th data jd The d attribute value of the j feature center is equal to or greater than 1 and equal to or less than n+1.
And S2, respectively calculating the distances between the data and k original feature centers, and selecting the class with the smallest distance from the original feature center as the own feature class to obtain k feature class clusters k.
And S3, after all the data are iterated for one time, calculating an updated characteristic center, wherein the calculation method of the updated characteristic center is shown in a formula (4).
C in the above t Represents the center of the t feature class cluster, k, |S t I represents the number of data in the t-th feature class, x i Representing the ith data in the t-th feature class, wherein i is more than or equal to 1 and less than or equal to |S t |。
S4: after updating all the updated feature centers, calculating the distance between the original feature centers and the updated feature centers, wherein the calculation method is shown in a formula (5). If the distance is greater than a threshold value minDis (which can be abbreviated as a distance threshold value) and the iteration number is less than an item (which can be abbreviated as an iteration number threshold value), returning to S2 to continue iteration; and stopping iteration if the distances between all original and updated feature centers are smaller than a threshold value minDis or the iteration times are larger than iter, wherein the threshold value minDis is used for judging the distances between the feature centers after two iterations.
C in the above it Representing the result after the ith iteration of the ith feature center, k, t is more than or equal to 1 and less than or equal to item, C itd The d attribute value after the t iteration of the ith feature center is represented, and d is more than or equal to 1 and less than or equal to n+1.
And obtaining k feature classes after the iteration is finished, wherein the feature classes are k mixed data feature classes.
Step 107: evaluating a plurality of raw material proportioning schemes to be tested according to the k mixed data characteristic classes to determine an optimal raw material proportioning scheme, and specifically comprising the following steps:
reducing the similarity matrix in the k mixed data feature classes into a raw material proportioning scheme to obtain k raw material feature classes; one of the feedstock characteristics includes a plurality of feedstock proportioning schemes;
calculating the average value of the polycrystalline silicon quality test data in k raw material characteristic classes, and taking the average value of the polycrystalline silicon quality test data as k identification values of k raw material characteristic classes;
dividing the grades of k raw material characteristic classes according to k identification values; wherein the raw material characteristic class rating comprises 3 kinds of normal class, poor class and abnormal class;
evaluating a plurality of to-be-tested polycrystalline silicon raw material proportioning schemes according to the grades, and determining an optimal raw material proportioning scheme;
acquiring a raw material proportioning scheme to be tested;
Calculating the similarity of the to-be-detected raw material proportioning scheme and each raw material proportioning scheme in each raw material characteristic class, and determining the average similarity of the to-be-detected raw material proportioning scheme and each raw material characteristic class; wherein the calculation of the similarity is performed according to the similarity determination unit 2, and the specific calculation is as follows:
a similarity determination unit 2 for applying a formulaCalculating the similarity p (x) between the ith raw material proportioning scheme and the jth raw material proportioning scheme i ,x j ) The method comprises the steps of carrying out a first treatment on the surface of the Wherein x is i Represents the ith raw material proportioning scheme, x j Represents the j-th raw material proportioning scheme, m represents the dimension of the raw material data, i is more than or equal to 1, and j is more than or equal to Num r ，Num r Is the number of raw material proportioning schemes in the (r) mixed data feature class after clustering, wherein k and theta d (x i ，x j ) Representing the orthogonality value of the ith and jth stock solutions in the d-th dimension, wherein ∈>x id ＝x jd The d attribute value representing the ith raw material proportioning scheme and the jth raw material proportioning scheme is the same, x id ≠x jd The d attribute values of the ith raw material proportioning scheme and the jth raw material proportioning scheme are different, and d is more than or equal to 1 and less than or equal to m; x is x id The d attribute value of the ith raw material proportioning scheme; x is x jd The d attribute value of the j-th raw material proportioning scheme.
And comparing the k average similarity, determining the raw material characteristic class corresponding to the maximum average similarity, acquiring the rating of the raw material characteristic class corresponding to the maximum average similarity, and taking the rating of the raw material characteristic class corresponding to the maximum average similarity as the rating of the to-be-detected polycrystalline silicon raw material proportioning scheme.
In practical application, the similarity matrix in the mixed data feature class is reduced to the original batching data to obtain k raw material feature classes.
And calculating an average value of minority carrier lifetime values in the raw material feature class, and taking the average value as an identification value of the raw material feature class. The class with the identification value larger than the threshold value 1 is a normal class, the class with the identification value between the threshold value 1 and the threshold value 2 is a poor class, and the class with the identification value smaller than the threshold value 2 is an abnormal class. If the average minority carrier lifetime value of a feature class is higher, the product produced according to the raw material proportion in the feature class has higher minority carrier lifetime value, which indicates that the raw material proportion of the product is reasonable; if the average minority carrier lifetime value is lower, the product produced according to the raw material ratio of the characteristic class has a lower minority carrier lifetime value, which indicates that the raw material ratio of the product is unreasonable; if the future proportioning scheme is the same as or has higher similarity with the proportioning scheme with lower minority carrier lifetime value, a warning is sent out to prompt that the proportioning scheme is not selectable.
And (3) inputting a to-be-detected raw material proportioning scheme, and calculating the average similarity of the to-be-detected raw material proportioning scheme and each raw material characteristic class, wherein the calculation method of the average similarity is shown in a formula (6).
Wherein x is j ∈S i (6)
X in formula (6) test Represents the proportioning scheme of the raw materials to be measured, averSimi (S i ) Representing the proportioning scheme of the raw materials to be tested and the characteristic class S of the raw materials i Average similarity of S i I indicates the number of data in the feature class, p (x test ，x j ) Representing data x test And x j Is a similarity of (3).
Dividing the proportioning scheme of the raw materials to be detected into the feature class with the maximum average similarity. If the characteristic identification value belongs to the poor class, giving a warning, and using the raw material ratio is more likely to produce a polysilicon product with lower quality; if the identification value of the belonging characteristic class belongs to the abnormal class, a warning is sent out, and the raw material proportioning scheme is greatly influenced by abnormal factors and is cautiously used.
Fig. 3 is a schematic diagram of a system for determining a polysilicon raw material proportioning scheme provided by the invention. As shown in FIG. 3, the invention is used for analyzing the raw batching data of the polysilicon, and is divided into 3 sub-modules, namely a data storage module, a data analysis module and a data monitoring module (which can be simply called as a storage module, an analysis module and a monitoring module). The storage module is used for storing the original data and the analysis result of the data analysis module. Raw data refer to the batching data and quality testing data of the produced products, and analysis results refer to different feature classes divided by an analysis module.
Fig. 4 is a functional flow diagram of a data analysis module provided by the present invention. As shown in fig. 4, the data analysis module (may be simply referred to as an analysis module) is configured to analyze raw ingredient data, divide the raw ingredient data into different feature classes, and store the feature classes in the storage module; the analysis results can be classified into 3 major classes as a whole, including normal class, poor class, and abnormal class. The characteristic class identification value is larger than the threshold value 1 and is a normal class, the identification value is between the threshold value 1 and the threshold value 2 and is a poor class, and the identification value is smaller than the threshold value 2 and is an abnormal class. The data analysis module comprises a raw material proportioning scheme acquisition module, a similarity determination module, a similarity matrix establishment module, a polysilicon quality test data determination module, a mixed data generation module and a clustering module.
Fig. 5 is a functional flow diagram of a data monitoring module provided by the present invention. As shown in fig. 5, the monitoring module monitors the data to be detected according to the output result of the data analysis module, and sends out early warning according to the quality of the monitoring result. The data to be measured refers to known raw ingredient data, while the quality test data is data (e.g., minority carrier lifetime value, resistance value, etc.) unknown. The monitoring module only gives an early warning to quality evaluation data such as minority carrier lifetime value or resistance value according to the original batching data. The data monitoring module comprises a raw material proportioning scheme evaluation module.
In order to ensure that the feature class difference divided by the analysis module is large enough, the class data k can be set to be 20; the iteration number threshold value is generally 100 or integral multiple thereof; the threshold minDis is typically set to 0.0001; the product threshold 1 and the product threshold 2 are flexibly valued according to different quality test data, and the valued range (experience value) of part of the quality test data is shown in table 7.
Table 7 quality test data value range example
If a conventional data organization mode is used for polysilicon raw material data, namely each raw material is used as a single attribute, when the variety of raw materials used by the product is more, the dimension of the data is increased, the raw material data needs to be filled with 0 filling positions more, and the reliability and stability of analysis results are reduced due to dimension disaster by using the data in the format. The data organization mode of the invention can accommodate more raw material data with lower dimensionality, and can effectively reduce the occurrence of dimension disasters.
When initializing the feature center, if the feature center is initialized randomly, the analysis result is easier to fall into local optimum; the invention uses the minimum maximum distance method to initialize the feature center, and disperses each special center as far as possible in the initial stage, thereby effectively reducing the probability of local optimum and improving the reliability of analysis results.
The similarity transformation processing raw material data used in the invention converts symbol data which are difficult to compare originally into a similarity matrix with a calculated distance, and each similarity vector in the matrix can reflect the similarity of one piece of sample data and all other samples. The more the same raw materials are used for two pieces of sample data in the raw material data, the higher the similarity is, and the reliability of the analysis result can be greatly improved.
Fig. 6 is a schematic diagram of a system for determining an optimal raw material proportioning scheme of polycrystalline silicon. As shown in fig. 6, a system for determining an optimal raw material proportioning scheme for polysilicon includes:
a raw material proportioning scheme acquisition module 601, configured to acquire a plurality of groups of raw material proportioning schemes in the polysilicon production process; the raw material proportioning scheme comprises a plurality of raw material data; the raw material data comprise raw polycrystal blocks, circulating materials, broken polycrystal blocks, purified ingot cores, top skins, rim charge and tails;
A similarity determining module 602, configured to calculate a similarity between each of the raw material proportioning schemes and each of the plurality of groups of raw material proportioning schemes;
a similarity matrix establishing module 603, configured to establish a similarity matrix according to the similarity;
a polysilicon quality test data determining module 604, configured to determine polysilicon quality test data according to the raw material proportioning scheme; the polycrystalline silicon quality test data comprise a head removing length, a tail removing length, a minority carrier lifetime value, a head resistor, a tail resistor, a seed crystal thickness, a minority carrier yield, an infrared reject ratio and a final yield;
the hybrid data generating module 605 is configured to add the polysilicon quality test data to the similarity matrix to generate hybrid data;
the clustering module 606 is configured to cluster the hybrid data by using a clustering algorithm, so as to generate k hybrid data feature classes;
and the raw material proportioning scheme evaluation module 607 is used for evaluating a plurality of raw material proportioning schemes to be tested according to the k mixed data feature classes to determine an optimal raw material proportioning scheme.
The system further comprises:
and the raw material data numbering module is used for numbering the raw material data in the raw material proportioning scheme, and the name of the raw material data is replaced by the numbered number.
The similarity determining module specifically comprises:
a similarity determining unit for using the formulaCalculating the similarity p (x) between the ith raw material proportioning scheme and the jth raw material proportioning scheme i ,x j ) The method comprises the steps of carrying out a first treatment on the surface of the Wherein x is i Represents the ith raw material proportioning scheme, x j Represents the j-th raw material proportioning scheme, m represents the dimension of the raw material data, 1 is less than or equal to i, j is less than or equal to n, n is the number of the obtained multiple groups of raw material proportioning schemes, and theta d (x i ，x j ) Representing the orthogonality value of the ith and jth stock solutions in the d-th dimension, wherein ∈> x id ＝x jd The d attribute value representing the ith raw material proportioning scheme and the jth raw material proportioning scheme is the same, x id ≠x jd The d attribute values of the ith raw material proportioning scheme and the jth raw material proportioning scheme are different, and d is more than or equal to 1 and less than or equal to m; x is x id The d attribute value of the ith raw material proportioning scheme; x is x jd The d attribute value of the j-th raw material proportioning scheme.
The clustering module specifically comprises:
the original feature center selecting unit is used for selecting the 1 st original feature center in the mixed data by using a minimum maximum distance method;
calculating the distance between each datum in the mixed data and the 1 st original feature center, and selecting the point with the largest distance with the 1 st original feature center as the 2 nd original feature center;
Calculating the distance between each data in the mixed data and the 1 st original feature center and the 2 nd original feature center respectively to obtain a first maximum distance farthest from the 1 st original feature center and a second maximum distance farthest from the 2 nd original feature center, and comparing the points corresponding to the minimum value selection values of the first maximum distance and the second maximum distance as the 3 rd original feature center until k original feature centers are selected;
a feature cluster determining unit, configured to allocate the hybrid data to k clusters to which k original feature centers closest to k original feature centers belong;
the updated feature center selecting unit is used for taking the average value of the attributes of the mixed data in the k class clusters as an updated feature center; one of the original feature centers corresponds to only one of the updated feature centers;
a distance determining unit, configured to determine a distance between the original feature center and the updated feature center according to the original feature center and the updated feature center;
the iteration number acquisition unit is used for acquiring the current iteration number;
The judging and analyzing unit is used for judging whether the distance is larger than a distance threshold value or not to obtain a first judging result;
if the first judgment result shows that the distance is larger than a distance threshold, continuing to judge whether the current iteration number is smaller than the iteration number threshold or not, and obtaining a second judgment result;
and if the second judgment result shows that the current iteration number is smaller than an iteration number threshold, returning to the step of distributing the mixed data to k class clusters which are closest to k original feature centers, until the distance is smaller than a distance threshold, or if the current iteration number is larger than the iteration number threshold, outputting the current feature class cluster, and taking the current feature class cluster as the k mixed data feature classes.
The invention discloses a method and a system for determining an optimal raw material proportioning scheme of polycrystalline silicon, which are used for analyzing raw material proportioning data of polycrystalline silicon to divide data with higher similarity in the raw material proportioning data into one type, divide raw material proportioning data with lower similarity into different types, analyze common characteristics of the raw material proportioning data of the same type of products according to results, and remarkably distinguish raw material proportioning data of different types of products. And storing the original batching data of the abnormal products into an abnormal database, and storing the original batching data of the quality normal products into a normal database. The original proportioning data in the normal database and the abnormal database are further subjected to comparison analysis of each raw material proportioning scheme, factors causing abnormal proportioning of the abnormal raw materials are found, and in actual production, the average similarity between the proportioning data to be measured and various proportioning data in the database is calculated first, feedback is given, and poor proportioning combination of the raw materials is avoided. And (3) giving a warning to the raw material proportioning scheme with higher similarity to the abnormal data, and prompting that the formula is unqualified so as to reselect the proportioning scheme. The invention can rapidly and efficiently screen out better raw material proportioning combination from hundreds to thousands of raw material proportioning combinations, and improves the production quality of polysilicon.
In the present specification, each embodiment is described in a progressive manner, and each embodiment is mainly described in a different point from other embodiments, and identical and similar parts between the embodiments are all enough to refer to each other. For the system disclosed in the embodiment, since it corresponds to the method disclosed in the embodiment, the description is relatively simple, and the relevant points refer to the description of the method section.
The principles and embodiments of the present invention have been described herein with reference to specific examples, the description of which is intended only to assist in understanding the methods of the present invention and the core ideas thereof; also, it is within the scope of the present invention to be modified by those of ordinary skill in the art in light of the present teachings. In view of the foregoing, this description should not be construed as limiting the invention.

Claims (8)

1. The method for determining the optimal raw material proportioning scheme of the polycrystalline silicon is characterized by comprising the following steps of:
obtaining a plurality of groups of raw material proportioning schemes in the production process of polysilicon; the raw material proportioning scheme comprises a plurality of different raw material data; the raw material data comprise raw polycrystal lump materials, circulating materials, broken polycrystal materials, purified ingot cores, top skins, rim charge and tailing materials;
Calculating the similarity between each group of the raw material proportioning scheme and each of the plurality of groups of the raw material proportioning schemes;
establishing a similarity matrix according to the similarity;
determining polycrystalline silicon quality test data according to the raw material proportioning scheme; the polycrystalline silicon quality test data comprise a head removing length, a tail removing length, a minority carrier lifetime value, a head resistor, a tail resistor, a seed crystal thickness, a minority carrier yield, an infrared reject ratio and a final yield;
adding the polycrystalline silicon quality test data into the similarity matrix to generate mixed data;
clustering the mixed data by adopting a clustering algorithm to generate k mixed data feature classes, wherein the method specifically comprises the following steps: selecting a 1 st original feature center in the mixed data by using a minimum maximum distance method; calculating the distance between each datum in the mixed data and the 1 st original feature center, and selecting the point with the largest distance with the 1 st original feature center as the 2 nd original feature center; calculating the distance between each data in the mixed data and the 1 st original feature center and the 2 nd original feature center respectively to obtain a first maximum distance farthest from the 1 st original feature center and a second maximum distance farthest from the 2 nd original feature center, and comparing the points corresponding to the minimum value selection values of the first maximum distance and the second maximum distance as the 3 rd original feature center until k original feature centers are selected; distributing the mixed data to k class clusters which are closest to k original feature centers; taking the average value of the attributes of the mixed data in the k class clusters as an updated feature center; one of the original feature centers corresponds to only one of the updated feature centers; determining a distance between the original feature center and the updated feature center according to the original feature center and the updated feature center; acquiring the current iteration times; judging whether the distance is greater than a distance threshold value or not to obtain a first judgment result; if the first judgment result shows that the distance is larger than a distance threshold, continuing to judge whether the current iteration number is smaller than the iteration number threshold or not, and obtaining a second judgment result; if the second judgment result shows that the current iteration number is smaller than an iteration number threshold, returning to the step of distributing the mixed data to k class clusters which are closest to k original feature centers and belong to the k original feature centers until the distance is smaller than a distance threshold or the current iteration number is larger than the iteration number threshold, outputting the current feature class cluster, and taking the current feature class cluster as the k mixed data feature classes;
And evaluating a plurality of raw material proportioning schemes to be tested according to the k mixed data characteristic classes, and determining an optimal raw material proportioning scheme.
2. The method for determining an optimal raw material formulation for polycrystalline silicon according to claim 1, wherein calculating the similarity between each of the raw material formulation and each of the raw material formulation in the plurality of raw material formulation sets comprises:
and numbering each raw material data in the raw material proportioning scheme, and replacing the name of the raw material data with a numbered number.
3. The method for determining an optimal raw material proportioning scheme for polysilicon according to claim 1, wherein the calculating of the similarity between each raw material proportioning scheme and each of the plurality of groups of raw material proportioning schemes specifically comprises:
using the formulaCalculating the similarity p (x) between the ith raw material proportioning scheme and the jth raw material proportioning scheme i ,x j ) The method comprises the steps of carrying out a first treatment on the surface of the Wherein xi represents the ith raw material proportioning scheme, x j Represents the j-th raw material proportioning scheme, m represents the dimension of the raw material data, i is not less than 1, n is not less than j, n is the number of the obtained multiple groups of raw material proportioning schemes,θ d (x i ,x j ) Representing the orthogonality value of the ith and jth stock solutions in the d-th dimension, wherein +. >x id ＝x jd The d attribute value representing the ith raw material proportioning scheme and the jth raw material proportioning scheme is the same, x id ≠x jd The d attribute values of the ith raw material proportioning scheme and the jth raw material proportioning scheme are different, and d is more than or equal to 1 and less than or equal to m; x is x id The d attribute value of the ith raw material proportioning scheme; x is x jd The d attribute value of the j-th raw material proportioning scheme.
4. The method for determining an optimal raw material proportioning scheme for polysilicon according to claim 1, wherein the method for determining an optimal raw material proportioning scheme by evaluating a plurality of raw material proportioning schemes to be tested according to the k mixed data feature classes specifically comprises:
reducing the similarity matrix in the k mixed data feature classes into a raw material proportioning scheme to obtain k raw material feature classes; one of the feedstock characteristics includes a plurality of feedstock proportioning schemes;
calculating the average value of the polycrystalline silicon quality test data in k raw material characteristic classes, and taking the average value of the polycrystalline silicon quality test data as k identification values of k raw material characteristic classes;
dividing the grades of k raw material characteristic classes according to k identification values; the raw material feature class rating includes: normal class, worse class, and abnormal class;
And evaluating a plurality of to-be-tested polycrystalline silicon raw material proportioning schemes according to the grades, and determining an optimal raw material proportioning scheme.
5. The method for determining an optimal raw material proportioning scheme for polysilicon according to claim 4, wherein the step of evaluating a plurality of raw material proportioning schemes to be tested according to the evaluation, the method specifically comprises:
acquiring a raw material proportioning scheme to be tested;
calculating the similarity of the to-be-detected raw material proportioning scheme and each raw material proportioning scheme in each raw material characteristic class, and determining the average similarity of the to-be-detected raw material proportioning scheme and each raw material characteristic class;
and comparing the k average similarity, determining the raw material characteristic class corresponding to the maximum average similarity, acquiring the rating of the raw material characteristic class corresponding to the maximum average similarity, and taking the rating of the raw material characteristic class corresponding to the maximum average similarity as the rating of the to-be-detected polycrystalline silicon raw material proportioning scheme.
6. The system for determining the optimal raw material proportioning scheme of the polycrystalline silicon is characterized by comprising the following components:
the raw material proportioning scheme acquisition module is used for acquiring a plurality of groups of raw material proportioning schemes in the production process of the polysilicon; the raw material proportioning scheme comprises a plurality of raw material data; the raw material data comprise raw polycrystal blocks, circulating materials, broken polycrystal blocks, purified ingot cores, top skins, rim charge and tails;
The similarity determining module is used for calculating the similarity between each raw material proportioning scheme and each raw material proportioning scheme in the plurality of groups of raw material proportioning schemes;
the similarity matrix establishing module is used for establishing a similarity matrix according to the similarity;
the polycrystalline silicon quality test data determining module is used for determining polycrystalline silicon quality test data according to the raw material proportioning scheme; the polycrystalline silicon quality test data comprise a head removing length, a tail removing length, a minority carrier lifetime value, a head resistor, a tail resistor, a seed crystal thickness, a minority carrier yield, an infrared reject ratio and a final yield;
the mixed data generation module is used for adding the polycrystalline silicon quality test data into the similarity matrix to generate mixed data;
the clustering module is used for clustering the mixed data by adopting a clustering algorithm to generate k mixed data feature classes, and specifically comprises the following steps:
the original feature center selecting unit is used for selecting the 1 st original feature center in the mixed data by using a minimum maximum distance method;
calculating the distance between each datum in the mixed data and the 1 st original feature center, and selecting the point with the largest distance with the 1 st original feature center as the 2 nd original feature center;
Calculating the distance between each data in the mixed data and the 1 st original feature center and the 2 nd original feature center respectively to obtain a first maximum distance farthest from the 1 st original feature center and a second maximum distance farthest from the 2 nd original feature center, and comparing the points corresponding to the minimum value selection values of the first maximum distance and the second maximum distance as the 3 rd original feature center until k original feature centers are selected;
a feature cluster determining unit, configured to allocate the hybrid data to k clusters to which k original feature centers closest to k original feature centers belong;
the updated feature center selecting unit is used for taking the average value of the attributes of the mixed data in the k class clusters as an updated feature center; one of the original feature centers corresponds to only one of the updated feature centers;
a distance determining unit, configured to determine a distance between the original feature center and the updated feature center according to the original feature center and the updated feature center;
the iteration number acquisition unit is used for acquiring the current iteration number;
The judging and analyzing unit is used for judging whether the distance is larger than a distance threshold value or not to obtain a first judging result;
if the first judgment result shows that the distance is larger than a distance threshold, continuing to judge whether the current iteration number is smaller than the iteration number threshold or not, and obtaining a second judgment result;
if the second judgment result shows that the current iteration number is smaller than an iteration number threshold, returning to the step of distributing the mixed data to k class clusters which are closest to k original feature centers and belong to the k original feature centers until the distance is smaller than a distance threshold or the current iteration number is larger than the iteration number threshold, outputting the current feature class cluster, and taking the current feature class cluster as the k mixed data feature classes;
and the raw material proportioning scheme evaluation module is used for evaluating a plurality of raw material proportioning schemes to be tested according to the k mixed data characteristic classes to determine an optimal raw material proportioning scheme.
7. The polysilicon optimal raw material proportioning scheme determination system as set forth in claim 6, further comprising:
and the raw material data numbering module is used for numbering the raw material data in the raw material proportioning scheme, and the name of the raw material data is replaced by the numbered number.
8. The system for determining the optimal raw material proportioning scheme for the polycrystalline silicon according to claim 6, wherein the similarity determining module specifically comprises:
a similarity determining unit for using the formulaCalculating the similarity p (x) between the ith raw material proportioning scheme and the jth raw material proportioning scheme i ,x j ) The method comprises the steps of carrying out a first treatment on the surface of the Wherein xi represents the ith raw material proportioning scheme, x j Represents the j-th raw material proportioning scheme, m represents the dimension of the raw material data, 1 is less than or equal to i, j is less than or equal to n, n is the number of the obtained multiple groups of raw material proportioning schemes, and theta d (x i ,x j ) Representing the orthogonality value of the ith and jth stock solutions in the d-th dimension, wherein ∈> x id ＝x jd The d attribute value representing the ith raw material proportioning scheme and the jth raw material proportioning scheme is the same, x id ≠x jd The d attribute values of the ith raw material proportioning scheme and the jth raw material proportioning scheme are different, and d is more than or equal to 1 and less than or equal to m; x is x id The d attribute value of the ith raw material proportioning scheme; x is x jd The d attribute value of the j-th raw material proportioning scheme.
CN202010639708.7A 2020-07-06 2020-07-06 Method and system for determining optimal raw material proportioning scheme of polycrystalline silicon Active CN111755079B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202010639708.7A CN111755079B (en) 2020-07-06 2020-07-06 Method and system for determining optimal raw material proportioning scheme of polycrystalline silicon

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202010639708.7A CN111755079B (en) 2020-07-06 2020-07-06 Method and system for determining optimal raw material proportioning scheme of polycrystalline silicon

Publications (2)

Publication Number Publication Date
CN111755079A CN111755079A (en) 2020-10-09
CN111755079B true CN111755079B (en) 2024-03-19

Family

ID=72679334

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202010639708.7A Active CN111755079B (en) 2020-07-06 2020-07-06 Method and system for determining optimal raw material proportioning scheme of polycrystalline silicon

Country Status (1)

Country Link
CN (1) CN111755079B (en)

Citations (13)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
AU2001271892A1 (en) * 2000-07-07 2002-01-21 Physiome Sciences, Inc. Method and system for modeling biological systems
CN102609784A (en) * 2012-01-16 2012-07-25 东北大学 Collocation method of iron ores of steel raw material field
CN103193234A (en) * 2013-04-12 2013-07-10 太原理工大学 Method for preparing magnesium silicide-based powder thermoelectric material by utilizing polysilicon byproduct
CN103524284A (en) * 2013-10-14 2014-01-22 中国石油化工股份有限公司 Forecasting and optimizing method for ethylene cracking material configuration
CN103729550A (en) * 2013-12-18 2014-04-16 河海大学 Multi-model integrated flood forecasting method based on propagation time clustering analysis
CN104408206A (en) * 2014-12-23 2015-03-11 许昌学院 Distributed support vector clustering method and system
US9009156B1 (en) * 2009-11-10 2015-04-14 Hrl Laboratories, Llc System for automatic data clustering utilizing bio-inspired computing models
CN105023026A (en) * 2015-08-18 2015-11-04 苏州大学张家港工业技术研究院 Semi-supervised clustering method and semi-supervised clustering system based on nonnegative matrix factorization
CN105069521A (en) * 2015-07-24 2015-11-18 许继集团有限公司 Photovoltaic power plant output power prediction method based on weighted FCM clustering algorithm
CN106777535A (en) * 2016-11-25 2017-05-31 深圳大学 Alkali-activated carbonatite mixture ratio design of recycled aggregate concrete method and system
CN108549789A (en) * 2018-04-19 2018-09-18 中南大学 A kind of integrated modelling approach of tertiary cathode material feed proportioning system
CN109523060A (en) * 2018-10-22 2019-03-26 上海交通大学 Ratio optimization method of the high proportion renewable energy under transmission and distribution network collaboration access
WO2019209723A1 (en) * 2018-04-26 2019-10-31 Becton, Dickinson And Company Characterization and sorting for particle analyzers

Family Cites Families (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20170220284A1 (en) * 2016-01-29 2017-08-03 Netapp, Inc. Block-level internal fragmentation reduction using a heuristic-based approach to allocate fine-grained blocks

Patent Citations (13)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
AU2001271892A1 (en) * 2000-07-07 2002-01-21 Physiome Sciences, Inc. Method and system for modeling biological systems
US9009156B1 (en) * 2009-11-10 2015-04-14 Hrl Laboratories, Llc System for automatic data clustering utilizing bio-inspired computing models
CN102609784A (en) * 2012-01-16 2012-07-25 东北大学 Collocation method of iron ores of steel raw material field
CN103193234A (en) * 2013-04-12 2013-07-10 太原理工大学 Method for preparing magnesium silicide-based powder thermoelectric material by utilizing polysilicon byproduct
CN103524284A (en) * 2013-10-14 2014-01-22 中国石油化工股份有限公司 Forecasting and optimizing method for ethylene cracking material configuration
CN103729550A (en) * 2013-12-18 2014-04-16 河海大学 Multi-model integrated flood forecasting method based on propagation time clustering analysis
CN104408206A (en) * 2014-12-23 2015-03-11 许昌学院 Distributed support vector clustering method and system
CN105069521A (en) * 2015-07-24 2015-11-18 许继集团有限公司 Photovoltaic power plant output power prediction method based on weighted FCM clustering algorithm
CN105023026A (en) * 2015-08-18 2015-11-04 苏州大学张家港工业技术研究院 Semi-supervised clustering method and semi-supervised clustering system based on nonnegative matrix factorization
CN106777535A (en) * 2016-11-25 2017-05-31 深圳大学 Alkali-activated carbonatite mixture ratio design of recycled aggregate concrete method and system
CN108549789A (en) * 2018-04-19 2018-09-18 中南大学 A kind of integrated modelling approach of tertiary cathode material feed proportioning system
WO2019209723A1 (en) * 2018-04-26 2019-10-31 Becton, Dickinson And Company Characterization and sorting for particle analyzers
CN109523060A (en) * 2018-10-22 2019-03-26 上海交通大学 Ratio optimization method of the high proportion renewable energy under transmission and distribution network collaboration access

Non-Patent Citations (3)

* Cited by examiner, † Cited by third party
Title
半熔高效多晶硅铸锭中籽晶厚度自动测量系统的研究;郭丽;;现代工业经济和信息化(06);全文 *
基于模糊聚类算法的矿业城市竞争力研究;王丽;彭新光;;科技情报开发与经济(06);全文 *
结合最优灰色选择原理的聚类分析;李桂芝;许君妮;;知识经济(10);全文 *

Also Published As

Publication number Publication date
CN111755079A (en) 2020-10-09

Similar Documents

Publication Publication Date Title
CN109409628B (en) Acquisition terminal manufacturer evaluation method based on metering big data clustering model
CN109523518B (en) Tire X-ray defect detection method
CN113092981B (en) Wafer data detection method and system, storage medium and test parameter adjustment method
CN116825169B (en) Abnormal memory chip detection method based on test equipment
US7174281B2 (en) Method for analyzing manufacturing data
CN108197647B (en) Rapid clustering method for automobile starter endurance test data
CN115021679B (en) Photovoltaic equipment fault detection method based on multi-dimensional outlier detection
CN110659682A (en) Data classification method based on MCWD-KSMOTE-AdaBoost-DenseNet algorithm
CN110544047A (en) Bad data identification method
CN112098361A (en) Corn seed identification method based on near infrared spectrum
CN112837754B (en) Method and device for automatically classifying single cells based on characteristic genes
CN108537249B (en) Industrial process data clustering method for density peak clustering
CN112906740B (en) Method for removing batch-to-batch differences aiming at tissue mass spectrum imaging result
CN111755079B (en) Method and system for determining optimal raw material proportioning scheme of polycrystalline silicon
CN117592656A (en) Carbon footprint monitoring method and system based on carbon data accounting
US6615157B1 (en) System method and computer program product for automatically assessing experiment results
US6944561B2 (en) Method for detection of manufacture defects
CN110910021A (en) Method for monitoring online defects based on support vector machine
CN107957944B (en) User data coverage rate oriented test case automatic generation method
CN112397404B (en) Semiconductor process detection method and system
CN108089136B (en) Automatic slicing method for fuel cell stack polarization curve test data
CN113610350A (en) Complex working condition fault diagnosis method, equipment, storage medium and device
CN107992886B (en) Method for predicting failure trend of gas turbine compressor equipment
CN116452581B (en) Intelligent voltage source state detection system and method based on machine vision
CN112711665B (en) Log anomaly detection method based on density weighted integration rule

Legal Events

Date Code Title Description
PB01 Publication
PB01 Publication
SE01 Entry into force of request for substantive examination
SE01 Entry into force of request for substantive examination
GR01 Patent grant
GR01 Patent grant