CN116645362B - Intelligent quality detection method and system for silicon carbide wafer - Google Patents
Intelligent quality detection method and system for silicon carbide wafer Download PDFInfo
- Publication number
- CN116645362B CN116645362B CN202310778180.5A CN202310778180A CN116645362B CN 116645362 B CN116645362 B CN 116645362B CN 202310778180 A CN202310778180 A CN 202310778180A CN 116645362 B CN116645362 B CN 116645362B
- Authority
- CN
- China
- Prior art keywords
- abnormal
- feature
- result
- identification
- image acquisition
- 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
Links
- 229910010271 silicon carbide Inorganic materials 0.000 title claims abstract description 90
- HBMJWWWQQXIZIP-UHFFFAOYSA-N silicon carbide Chemical compound [Si+]#[C-] HBMJWWWQQXIZIP-UHFFFAOYSA-N 0.000 title claims abstract description 86
- 238000001514 detection method Methods 0.000 title claims abstract description 80
- 230000002159 abnormal effect Effects 0.000 claims abstract description 243
- 230000007547 defect Effects 0.000 claims abstract description 120
- 235000012431 wafers Nutrition 0.000 claims abstract description 94
- 238000000034 method Methods 0.000 claims abstract description 45
- 238000011156 evaluation Methods 0.000 claims description 89
- 238000012549 training Methods 0.000 claims description 37
- 230000007704 transition Effects 0.000 claims description 28
- 238000012937 correction Methods 0.000 claims description 26
- 230000005856 abnormality Effects 0.000 claims description 13
- 230000007246 mechanism Effects 0.000 claims description 12
- 238000012795 verification Methods 0.000 claims description 12
- 238000009826 distribution Methods 0.000 claims description 11
- 238000005457 optimization Methods 0.000 claims description 11
- 238000004364 calculation method Methods 0.000 claims description 10
- 238000004458 analytical method Methods 0.000 claims description 7
- 238000004891 communication Methods 0.000 claims description 7
- 238000010276 construction Methods 0.000 claims description 6
- 230000000694 effects Effects 0.000 abstract description 6
- 238000012544 monitoring process Methods 0.000 abstract description 2
- 230000008569 process Effects 0.000 description 13
- 238000012545 processing Methods 0.000 description 7
- 238000005516 engineering process Methods 0.000 description 6
- 239000000463 material Substances 0.000 description 6
- XUIMIQQOPSSXEZ-UHFFFAOYSA-N Silicon Chemical group [Si] XUIMIQQOPSSXEZ-UHFFFAOYSA-N 0.000 description 3
- 238000012986 modification Methods 0.000 description 3
- 230000004048 modification Effects 0.000 description 3
- OKTJSMMVPCPJKN-UHFFFAOYSA-N Carbon Chemical compound [C] OKTJSMMVPCPJKN-UHFFFAOYSA-N 0.000 description 2
- 229910052799 carbon Inorganic materials 0.000 description 2
- 230000008859 change Effects 0.000 description 2
- 239000013078 crystal Substances 0.000 description 2
- 238000002425 crystallisation Methods 0.000 description 2
- 230000008025 crystallization Effects 0.000 description 2
- 238000011161 development Methods 0.000 description 2
- 238000010606 normalization Methods 0.000 description 2
- 230000003287 optical effect Effects 0.000 description 2
- 229910052710 silicon Inorganic materials 0.000 description 2
- 239000010703 silicon Substances 0.000 description 2
- HMDDXIMCDZRSNE-UHFFFAOYSA-N [C].[Si] Chemical compound [C].[Si] HMDDXIMCDZRSNE-UHFFFAOYSA-N 0.000 description 1
- 125000004432 carbon atom Chemical group C* 0.000 description 1
- 238000012512 characterization method Methods 0.000 description 1
- 238000009833 condensation Methods 0.000 description 1
- 230000005494 condensation Effects 0.000 description 1
- 230000006378 damage Effects 0.000 description 1
- 238000013500 data storage Methods 0.000 description 1
- 238000010586 diagram Methods 0.000 description 1
- 238000007417 hierarchical cluster analysis Methods 0.000 description 1
- 238000007689 inspection Methods 0.000 description 1
- 230000003993 interaction Effects 0.000 description 1
- 150000002500 ions Chemical class 0.000 description 1
- 238000004519 manufacturing process Methods 0.000 description 1
- 230000004660 morphological change Effects 0.000 description 1
- 238000004806 packaging method and process Methods 0.000 description 1
- 239000002245 particle Substances 0.000 description 1
- 238000005498 polishing Methods 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 238000012827 research and development Methods 0.000 description 1
- 238000012552 review Methods 0.000 description 1
- 238000012216 screening Methods 0.000 description 1
- 230000008054 signal transmission Effects 0.000 description 1
- 238000003860 storage Methods 0.000 description 1
- 239000000758 substrate Substances 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/0002—Inspection of images, e.g. flaw detection
- G06T7/0004—Industrial image inspection
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/40—Extraction of image or video features
- G06V10/44—Local feature extraction by analysis of parts of the pattern, e.g. by detecting edges, contours, loops, corners, strokes or intersections; Connectivity analysis, e.g. of connected components
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
- G06V10/74—Image or video pattern matching; Proximity measures in feature spaces
- G06V10/75—Organisation of the matching processes, e.g. simultaneous or sequential comparisons of image or video features; Coarse-fine approaches, e.g. multi-scale approaches; using context analysis; Selection of dictionaries
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
- G06V10/764—Arrangements for image or video recognition or understanding using pattern recognition or machine learning using classification, e.g. of video objects
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
- G06V10/77—Processing image or video features in feature spaces; using data integration or data reduction, e.g. principal component analysis [PCA] or independent component analysis [ICA] or self-organising maps [SOM]; Blind source separation
- G06V10/774—Generating sets of training patterns; Bootstrap methods, e.g. bagging or boosting
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/10—Image acquisition modality
- G06T2207/10056—Microscopic image
- G06T2207/10061—Microscopic image from scanning electron microscope
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/20—Special algorithmic details
- G06T2207/20081—Training; Learning
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/30—Subject of image; Context of image processing
- G06T2207/30108—Industrial image inspection
- G06T2207/30148—Semiconductor; IC; Wafer
-
- 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)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Computer Vision & Pattern Recognition (AREA)
- General Physics & Mathematics (AREA)
- General Health & Medical Sciences (AREA)
- Software Systems (AREA)
- Health & Medical Sciences (AREA)
- Artificial Intelligence (AREA)
- Computing Systems (AREA)
- Multimedia (AREA)
- Evolutionary Computation (AREA)
- Databases & Information Systems (AREA)
- Medical Informatics (AREA)
- Quality & Reliability (AREA)
- Life Sciences & Earth Sciences (AREA)
- Biomedical Technology (AREA)
- Biophysics (AREA)
- Computational Linguistics (AREA)
- Data Mining & Analysis (AREA)
- Molecular Biology (AREA)
- General Engineering & Computer Science (AREA)
- Mathematical Physics (AREA)
- Testing Or Measuring Of Semiconductors Or The Like (AREA)
Abstract
The invention relates to the technical field of intelligent monitoring, and provides a method and a system for intelligently detecting the quality of a silicon carbide wafer, wherein the method comprises the following steps: connecting a positioning carrier, and positioning to finish acquisition of an image acquisition result; constructing a basic comparison feature set, comparing features of the image acquisition result, and determining an abnormal region; scanning an abnormal region to obtain an abnormal region image acquisition result; collecting abnormal characteristics of the same type of wafers, classifying and identifying and constructing an abnormal characteristic identification model; the abnormal region image acquisition result is input into an abnormal feature recognition model, initial recognition is carried out, an abnormal feature set is called, abnormal feature matching is carried out on the abnormal region image acquisition result, and an abnormal detection result is output, so that the technical problem of low defect detection efficiency of the silicon carbide wafer is solved, the defect detection of the silicon carbide wafer is automatically carried out, and the technical effect of improving the defect detection efficiency is achieved.
Description
Technical Field
The invention relates to the technical field of intelligent monitoring, in particular to an intelligent quality detection method and system for silicon carbide wafers.
Background
The development of information technology is continuous, the chip is used as the material foundation of the technology, and comprises a CPU for operation processing, a GPU for image processing, a memory chip for storage and a baseband chip for communication, and the development of the integrated circuit manufacturing industry is relatively lagging under the limit of the Watson protocol on the high and new technology.
In general, the detection method can detect most of surface and internal defects by using a high-resolution optical microscope, but the optical microscope is used for point-by-point detection, so that the speed is extremely low, and the defects of the silicon carbide wafer cannot be detected efficiently on the basis of ensuring the detection accuracy.
In summary, the prior art has the technical problem of low defect detection efficiency of silicon carbide wafers.
Disclosure of Invention
The application aims to solve the technical problem of low defect detection efficiency of the silicon carbide wafer in the prior art by providing the intelligent quality detection method and system for the silicon carbide wafer.
In view of the above problems, embodiments of the present application provide a method and a system for intelligently detecting quality of a silicon carbide wafer.
In a first aspect of the disclosure, a method for intelligently detecting quality of a silicon carbide wafer is provided, where the method is applied to a quality intelligent detection system, and the quality intelligent detection system is in communication connection with a positioning stage, an image acquisition device and a scanning electron microscope, and the method includes: the positioning carrier is connected with the positioning signal receiving device; when the received positioning signals are displayed and positioned, the image acquisition device is used for acquiring images of the silicon carbide wafer positioned on the positioning carrier to obtain an image acquisition result; constructing a basic comparison feature set based on the silicon carbide wafer, and comparing the features of the image acquisition result through the basic comparison feature set to determine an abnormal region; acquiring an image of the abnormal region through the scanning electron microscope to obtain an image acquisition result of the abnormal region; collecting abnormal characteristics of the same type of wafers of the silicon carbide wafers, carrying out classification identification, and constructing an abnormal characteristic identification model according to the classification identification result; after the abnormal region image acquisition result is input into the abnormal feature recognition model, initially recognizing the abnormal region image acquisition result through a feature calling evaluation unit, and calling an abnormal feature set; and executing abnormal feature matching on the abnormal region image acquisition result through the abnormal feature set, and outputting an abnormal detection result.
In another aspect of the disclosure, a system for intelligently detecting quality of a silicon carbide wafer is provided, wherein the system comprises: the positioning signal receiving module is used for connecting with the positioning carrier and receiving the positioning signal of the positioning carrier; the first image acquisition module is used for acquiring images of the silicon carbide wafer positioned on the positioning carrier through the image acquisition device when the received positioning signals are displayed and positioned completely, so that an image acquisition result is obtained; the feature comparison module is used for constructing a basic comparison feature set based on the silicon carbide wafer, and carrying out feature comparison on the image acquisition result through the basic comparison feature set to determine an abnormal region; the second image acquisition module is used for acquiring images of the abnormal areas through a scanning electron microscope to obtain an abnormal area image acquisition result; the classification identification module is used for collecting abnormal characteristics of the silicon carbide wafers of the same type, carrying out classification identification, and constructing an abnormal characteristic identification model according to the classification identification result; the initial recognition module is used for inputting the abnormal region image acquisition result into the abnormal feature recognition model, initially recognizing the abnormal region image acquisition result through a feature calling evaluation unit, and calling an abnormal feature set; and the abnormal feature matching module is used for executing abnormal feature matching on the abnormal region image acquisition result through the abnormal feature set and outputting an abnormal detection result.
One or more technical solutions provided in the present application have at least the following technical effects or advantages:
because of adopting the connection positioning carrier, receive the positioning signal; when the display positioning is completed, acquiring an image acquisition result; constructing a basic comparison feature set based on the silicon carbide wafer, comparing features of the image acquisition result, and determining an abnormal region; scanning an abnormal region to obtain an abnormal region image acquisition result; collecting abnormal characteristics of the same type of wafers, classifying and identifying and constructing an abnormal characteristic identification model; the abnormal region image acquisition result is input into an abnormal feature recognition model, initial recognition is carried out, an abnormal feature set is called, abnormal feature matching is carried out on the abnormal region image acquisition result, and an abnormal detection result is output, so that the defect detection of the silicon carbide wafer is automatically carried out, and the technical effect of improving the defect detection efficiency is achieved.
The foregoing description is only an overview of the technical solutions of the present application, and may be implemented according to the content of the specification in order to make the technical means of the present application more clearly understood, and in order to make the above-mentioned and other objects, features and advantages of the present application more clearly understood, the following detailed description of the present application will be given.
Drawings
Fig. 1 is a schematic flow chart of a possible method for intelligently detecting quality of a silicon carbide wafer according to an embodiment of the present application;
fig. 2 is a schematic flow chart of a possible sequential call in an intelligent quality detection method for silicon carbide wafers according to an embodiment of the present application;
fig. 3 is a schematic flow chart of a possible construction of an abnormal feature recognition model in an intelligent quality detection method of a silicon carbide wafer according to an embodiment of the present application;
fig. 4 is a schematic diagram of a possible structure of an intelligent detection system for quality of a silicon carbide wafer according to an embodiment of the present application.
Reference numerals illustrate: the system comprises a positioning signal receiving module 100, a first image acquisition module 200, a feature comparison module 300, a second image acquisition module 400, a classification identification module 500, an initial identification module 600 and an abnormal feature matching module 700.
Detailed Description
The embodiment of the application provides a quality intelligent detection method and a system for a silicon carbide wafer, which solve the technical problem of low defect detection efficiency of the silicon carbide wafer, realize the automatic defect detection of the silicon carbide wafer and improve the technical effect of the defect detection efficiency.
Having described the basic principles of the present application, various non-limiting embodiments of the present application will now be described in detail with reference to the accompanying drawings.
Embodiment one:
as shown in fig. 1, an embodiment of the present application provides a method for intelligently detecting quality of a silicon carbide wafer, where the method is applied to a quality intelligent detection system, and the quality intelligent detection system is communicatively connected with a positioning carrier, an image acquisition device and a scanning electron microscope, and the method includes:
s10: the positioning carrier is connected with the positioning signal receiving device;
s20: when the received positioning signals are displayed and positioned, the image acquisition device is used for acquiring images of the silicon carbide wafer positioned on the positioning carrier to obtain an image acquisition result;
specifically, the quality intelligent detection system is in communication connection with the positioning carrier, the image acquisition device and the scanning electron microscope, and the communication connection is simply through signal transmission interaction, and a communication network is formed among the quality intelligent detection system, the positioning carrier, the image acquisition device and the scanning electron microscope, so that hardware support is provided for quality intelligent detection;
based on the communication connection between the quality intelligent detection system and the positioning carrier, a positioning signal of the positioning carrier is received, a plurality of carrier positioning points are added in the positioning signal of the positioning carrier, a plurality of carrier positioning points can be randomly arranged (the randomly arranged positioning points are the prior art), when the shape and the position of a silicon carbide wafer of the positioning carrier can be uniquely determined by randomly arranging the plurality of carrier positioning points (if the shape or the position of the silicon carbide wafer is uncertain, the carrier positioning points need to be continuously arranged until the shape and the position of the silicon carbide wafer can be uniquely determined), the received positioning signal is displayed and positioned, when the received positioning signal is displayed and positioned, the image acquisition device is used for carrying out image acquisition on the silicon carbide wafer positioned on the positioning carrier, so that an image acquisition result is obtained, and a data basis is provided for subsequent detection.
S30: constructing a basic comparison feature set based on the silicon carbide wafer, and comparing the features of the image acquisition result through the basic comparison feature set to determine an abnormal region;
as shown in fig. 2, step S30 includes the steps of:
s31: performing feature recognition according to the feature comparison result, and generating a first sequential feature evaluation value of the abnormal type feature, wherein the first sequential feature evaluation value is provided with a position mark;
s32: obtaining historical detection data of the silicon carbide wafer, carrying out abnormal characteristic recurrence frequency analysis on the historical detection data based on time, and generating a second sequence characteristic evaluation value according to an analysis result;
s33: calculating to obtain feature calling sequence constraint with position identification through the first sequence feature evaluation value and the second sequence feature evaluation value;
s34: before the feature call evaluation unit performs feature recognition on the abnormal region image acquisition result, sequentially calling the abnormal features based on the position identification through the feature call sequence constraint.
Specifically, a basic comparison feature set is constructed based on the silicon carbide wafer, elements in the basic comparison feature set can be size features, slicing process features (the processing process of the silicon carbide substrate mainly comprises slicing, thinning and polishing, the slicing process features can correspond to process features defined by related technologies such as a cold split technology for dividing the silicon carbide wafer, the slice material utilization rate of the cold split technology is generally more than 80%), and other related features, feature comparison is performed on the image acquisition result through the basic comparison feature set (the basic comparison feature set can be divided into surface detection, ultrasonic detection, size comparison and slice material utilization rate of the silicon carbide wafer), and feature comparison results are obtained, wherein the feature comparison results correspond to abnormal areas with abnormal type features, the abnormal type features comprise but are not limited to external defects, internal bubble defects and size defects, and the abnormal areas are scratch defect areas with any type of features;
Performing feature recognition according to a feature comparison result (feature recognition: surface detection is used for detecting and recognizing external scratch defects, ultrasonic detection is used for detecting and recognizing internal bubble defects, size comparison is used for recognizing size defects), and if defects are recognized and determined to exist, generating first sequence feature evaluation values of corresponding abnormal type features, wherein the first sequence feature evaluation values have position marks (generally, a certain repeatability rule exists on defects of silicon carbide wafers produced by the same enterprise, the first sequence feature evaluation values can correspond to any type of features), and the first sequence feature evaluation values of the first point positions of the silicon carbide wafers are exemplified by that the size defects and the external scratch defects exist on the first point positions, namely, the first point positions are used as position marks, the size defects and the external scratch defects exist on the first point positions are subjected to standardized treatment, and standardized treatment results of the size defects and the external scratch defects exist on the first point positions are used as first sequence feature evaluation values of the first point positions;
performing data retrieval in a data storage unit of the quality intelligent detection system to obtain historical detection data of the silicon carbide wafer in the last natural month, counting abnormal feature reproduction based on the historical detection data, calculating abnormal feature reproduction frequency, wherein the abnormal feature reproduction frequency comprises external scratch defect reproduction frequency, internal bubble defect reproduction frequency and size defect reproduction frequency, the abnormal feature reproduction frequency corresponds to the abnormal type features one by one (combined with the above illustration, the probability of occurrence of the external scratch defect in the historical detection data is as high as 85.7%, namely that the slicing process of an enterprise is poor, the external scratch defect is high in discovery, and the slicing process setting correlation degree with the first point is low), and taking the abnormal feature reproduction frequency as a second sequence feature evaluation value;
Before the feature call evaluation unit performs feature recognition on the abnormal region image acquisition result, sequential call on abnormal features is required: calculating to obtain a feature calling sequence constraint (feature calling sequence constraint: adjusting detection sequence constraints of external scratch defects, internal bubble defects and size defects) with a position mark (in combination with the above-mentioned illustration, the position mark may be a first point position of a silicon carbide wafer in the size comparison process) through the first sequence feature evaluation value and the second sequence feature evaluation value, and in combination with the above-mentioned illustration, the slicing process of an enterprise is not good, namely the external scratch defect detection with the highest occurrence probability in the historical detection data corresponds to the first feature calling sequence constraint); and sequentially calling the abnormal features based on the position identification through the feature calling sequence constraint, performing feature comparison on the image acquisition result through the basic comparison feature set, and performing rapid inspection comparison from the angles of external scratches, internal bubbles and size defects to determine an abnormal region on the silicon carbide wafer, so that the defect detection efficiency of the silicon carbide wafer is accelerated.
The embodiment of the application further comprises:
S35: setting a multi-level transition level set of the first sequential feature evaluation values;
s36: inputting the first sequential characteristic evaluation value into the multi-level transition level set to obtain a coefficient allocation adjustment value;
s37: and carrying out coefficient adjustment on the initial distribution coefficient by the coefficient distribution adjustment value, calculating to obtain the calling value of each feature, and determining the feature calling sequence constraint according to the calling value sequence ordering result, wherein the calculation formula is as follows:wherein K is a calling value, ++>For the first order feature evaluation value, +.>For the second sequential characteristic evaluation value, T is the adjusted second sequential characteristic evaluation value influence coefficient, and +.>Wherein a is the initial coefficient of the first sequential feature evaluation value, b is the coefficient assignment adjustment value, +.>To adjust the influence constant, and->The value is +.>，/>The value at transition level m is +.>。
Specifically, by the first sequential characteristic evaluation value and the first sequential characteristic evaluation valueThe calculating of the two-order feature evaluation value to obtain feature calling order constraint with position identification comprises the following steps: the silicon carbide wafer has a plurality of crystallization states, namely a transition, namely a cross-grade morphology change, the multi-grade transition is essentially the morphology change in particles (carbon atoms, silicon atoms or other ions and molecules) of a silicon carbide material, the multi-grade transition grade corresponds to the plurality of crystallization states (silicon layers and carbon layers with repeated positions can appear due to four coordination requirements of carbon silicon), common several can be 3C-SiC, 2H-SiC, 4H-SiC and 15R-SiC, the multi-grade transition grade set of the first order characteristic evaluation value is set, and the multi-grade transition grade set is usually in a mode of # ，/>，/>，……，/>… …) form characterization;
inputting the first order characteristic evaluation value into the multi-stage transition grade set, taking the multi-stage transition grade set as a nominal number set, carrying out dimension normalization processing on the first order characteristic evaluation value (the dimension normalization processing is carried out to be consistent on different types of data types, and numerical calculation can be directly carried out after the data types are converted into the same type of data), and calculating to obtain a coefficient allocation adjustment value corresponding to the first order characteristic evaluation value (the coefficient allocation adjustment value corresponding to the first order characteristic evaluation value is consistent with the data types of the multi-stage transition grade set);
coefficient adjustment is carried out on an initial distribution coefficient (a preset parameter index, in the initial condition, the initial distribution coefficient can be set to be 0.5) through the coefficient distribution adjustment value, the calling value of each feature is obtained through calculation, the feature calling sequence constraint is determined according to the sequence sorting result of the calling values from large to small, and a calculation formula is as followsThe following steps:wherein K is a calling value, ++>For the first order feature evaluation value, +.>For the second order feature evaluation value, +.>To adjust the influence constant, and->The value is +. >，/>The value at transition level m is +.>(multiple transition levels are used to reflect the crystal state of the silicon carbide wafer, and the coefficients are assigned in the set of multiple transition levels, and if the crystal state of the silicon carbide wafer corresponds to the first transition level,/-the set of multiple transition levels is the first transition level>=/>The method comprises the steps of carrying out a first treatment on the surface of the If the crystalline state of the silicon carbide wafer corresponds to the mth transition level,=/>) T is the adjusted second order characteristic evaluation value influence coefficient, and +.>And b is a coefficient distribution adjustment value, and is used for setting a multilevel transition grade against the morphological change of a carbon atomic layer and a silicon atomic layer of the silicon carbide material so as to be suitable for angles of the silicon carbide material, accurately calculating a calling value and providing support for setting reasonable sequential calling for abnormal characteristics.
S40: acquiring an image of the abnormal region through the scanning electron microscope to obtain an image acquisition result of the abnormal region;
s50: collecting abnormal characteristics of the same type of wafers of the silicon carbide wafers, carrying out classification identification, and constructing an abnormal characteristic identification model according to the classification identification result;
specifically, image acquisition is performed on the abnormal region by the scanning electron microscope (taking the defect can be clearly observed as a standard, and the magnification of the microscope is adapted to the defect), and the image information of the abnormal region is used as an abnormal region image acquisition result; and taking the silicon carbide wafers with the same index specification as the wafers of the same type, collecting the abnormal characteristics of the wafers of the same type, carrying out classification identification (classification identification is carried out according to defect types, generally comprises but is not limited to external scratch defect identification, internal bubble defect identification and size defect identification), obtaining classification identification results, taking data obtained by the classification identification results as training data, constructing an abnormal characteristic identification model, taking the abnormal characteristic collection data of the wafers of the same type as reference data, and providing a data basis for constructing the abnormal characteristic identification model.
Step S50 includes the steps of:
s51: acquiring a public abnormal data set of the same type of wafers, and carrying out classification identification on the public abnormal data set to obtain a first classification identification result;
s52: performing data coverage evaluation on the first classification identification result to generate a coverage identification result;
s53: and training an initial abnormal feature recognition model based on the first classification identification result, and constructing the abnormal feature recognition model through the coverage identification and the coverage identification result.
Specifically, a public abnormal data set is disclosed, namely, an abnormal data set disclosed in an enterprise (the enterprise is internally disclosed, the public abnormal data set is only available for internal personnel to review, and the enterprise interests are related), the public abnormal data set of the same type of wafer is acquired and obtained, and the public abnormal data set is classified and identified by using an external scratch defect identifier, an internal bubble defect identifier and a size defect identifier to obtain a first classification identification result;
taking an external scratch defect mark as an example, if the area with the external scratch defect mark accounts for 0.16% of the total area of the silicon carbide wafer, the area without the external scratch defect mark accounts for 63.21% of the total area of the silicon carbide wafer, and 0.16% +63.21% = 63.37%, namely, the coverage of the external scratch defect mark is 63.37%), calculating coverage, and generating a coverage mark result, wherein the coverage mark result comprises the coverage of the external scratch defect mark, the coverage of the internal bubble defect mark and the coverage of the size defect mark;
Based on the first classification identification result as a data basis, taking a BP network model as a model basis, and taking an external scratch defect identification existing in the public abnormal data set as a first training set; the identification of the internal bubble defects existing in the public abnormal data set is used as a second training set; taking the size defect mark existing in the public abnormal data set as a third training set; and after the output tends to be stable (the output is stable: the output is consistent with the expected result), determining an initial abnormal characteristic identification model (the initial abnormal characteristic identification model comprises an external scratch defect identification layer, an internal bubble defect identification layer and a size defect identification layer), and marking the external scratch defect identification layer, the internal bubble defect identification layer and the size defect identification layer respectively through the coverage identification and the coverage identification result to obtain an abnormal characteristic identification model, thereby providing model support for defect detection identification of the silicon carbide wafer.
As shown in fig. 3, the embodiment of the present application further includes:
s54: encrypting the data of the public abnormal data set and generating a decryption key;
s55: transmitting the encrypted public abnormal data set, the decryption key, the initial abnormal feature identification model and the coverage identification result to a first mechanism;
s56: obtaining a compensation data set through the coverage identification result matching, wherein the compensation data set is an encrypted data set in the first mechanism;
s57: and carrying out model optimization training on the initial abnormal feature recognition model through the compensation data set, and completing construction of the abnormal feature recognition model according to an optimization training result.
Specifically, in the quality detection stage of the silicon carbide wafer, the model is easy to be fitted by only relying on the public abnormal data set in the enterprise, and generally, the public abnormal data sets in multiple enterprises in the same industry are required to be used as a training basis, so that the comprehensiveness of the data is improved, and the method is specific: performing data encryption on the public abnormal data set (encryption operation is performed by enterprises corresponding to the abnormal data set, data decryption needs to be performed by identity authentication, if the identity authentication is not passed, data destruction operation is automatically started, a basis is provided for ensuring data security), and a decryption key is generated, wherein the decryption key is generally disposable; the encrypted public abnormal data set, the decryption key, the initial abnormal feature identification model and the coverage identification result are subjected to data packaging, the names of the data packets are named as corresponding enterprise names, and then the data packets are sent to a first institution (generally a trusted third party institution);
Matching to obtain a compensation data set (in combination with the above description, the coverage of the external scratch defect mark is 63.37%, in other data in enterprises, if the coverage of the external scratch defect mark is more than 63.37%, it indicates that the first training set can be compensated, the coverage of the external scratch defect mark is more than 63.37% as a reference, the first training set is compared with the first training set, and if the coverage of the external scratch defect mark is not consistent with the first training set, the first training set is added to the compensation data set), wherein the compensation data set is an encrypted data set in the first mechanism, and the compensation data set is divided into an external scratch defect compensation data set, an internal bubble defect compensation data set and a size defect compensation data set; and carrying out model optimization training on the initial abnormal feature recognition model through the compensation data set, completing the construction of the abnormal feature recognition model according to an optimization training result by adopting an incremental learning (incremental learning: a common means in the model training process), providing support for avoiding model overfitting, and expanding training sample data under the condition of ensuring data safety so as to realize model optimization.
And compared with any enterprise, the method has the advantages that the data in the enterprise has limitation due to research and development investment tendency (more investment in the aspect of internal bubble defects and more complete data in the aspect of corresponding internal bubble defects and more complete data in the aspect of size defects), the training sample data is expanded, the accuracy of abnormal feature identification can be improved, and the accuracy of silicon carbide wafer defect detection is further improved.
S60: after the abnormal region image acquisition result is input into the abnormal feature recognition model, initially recognizing the abnormal region image acquisition result through a feature calling evaluation unit, and calling an abnormal feature set;
s70: and executing abnormal feature matching on the abnormal region image acquisition result through the abnormal feature set, and outputting an abnormal detection result.
Specifically, the abnormal region image acquisition result is taken as an object of abnormal feature identification, after the abnormal feature identification model is input, the abnormal region image acquisition result is initially identified (initial identification, namely, first identification, namely, after identification is finished, the feature call evaluation unit is required to be initialized, and related data leakage in the feature call evaluation unit is avoided) through a feature call evaluation unit (the feature call evaluation unit is an external scratch defect identification layer, an internal bubble defect identification layer and a size defect identification layer are selected as a functional unit, in essence, if the external scratch defect identification layer is selected, a selection request header is added to the external scratch defect identification layer, and the selection request header is added and then the external scratch defect identification layer is called to the feature call evaluation unit);
And performing abnormal feature matching on the abnormal region image acquisition result through the abnormal feature set (if the matching is consistent, the defect can be an untapped technical difficulty in the field, and the defect essence exists is verified, if the matching is inconsistent, the defect can be an untapped technical difficulty in the enterprise, and the defect essence exists, and meanwhile, the defect needs to be paid attention to and improved in time), marking the abnormal feature matching result as an abnormal detection result, outputting the abnormal detection result, and ensuring the effectiveness and the accuracy of the quality detection of the silicon carbide wafer.
The embodiment of the application further comprises:
s81: performing sequential evaluation of abnormal feature call based on the abnormal detection result to generate an abnormal call speed evaluation result;
s82: generating a correction weight according to the abnormal calling speed evaluation result;
s83: and compensating the correction weight to the calling value.
Specifically, after the quality detection of the silicon carbide wafer is finished, performing sequential evaluation of abnormal feature call (if the external scratch defect identification layer is selected firstly, if the external scratch defect identification layer is matched to be consistent for a plurality of times, executing the abnormal feature matching process for a plurality of times to confirm that the external scratch defect is an untapped technical difficulty in the field, namely, the sequential evaluation of the abnormal feature call of the selected external scratch defect identification layer is low-level, and secondly, selecting the internal bubble defect identification layer, if the matching is inconsistent, the defect is probably an untapped technical difficulty in the enterprise, namely, the sequential evaluation of the abnormal feature call of the selected internal bubble defect identification layer is high-level), and marking the sequential evaluation of the abnormal feature call as an abnormal call speed evaluation result;
Carrying out standardization processing on the abnormal call speed evaluation result and the call value, and carrying out weighted calculation on each result obtained by the standardization processing by utilizing a variation coefficient method, wherein the method specifically comprises the following steps: the coefficient of variation method is an objective weighting method, the information contained in each result obtained by the standardization process is directly utilized, the correction weight obtained by the standardization process is obtained through calculation, after the correction weight is determined, the call value of the standardization process is subjected to weighted calculation, the correction weight is compensated to the call value through weight adjustment, an abnormal detection result is adopted, and the optimization compensation is carried out on sequential call in a mode of compensating to the call value, so that support is provided for rapidly and efficiently checking out abnormal silicon carbide wafers.
The embodiment of the application further comprises:
s84: performing abnormality verification on the abnormality detection result, and performing abnormality recognition feature clustering based on the abnormality verification result;
s85: transmitting the abnormal recognition clustering result to the first mechanism for data matching to obtain a corrected data matching result;
s86: and carrying out correction training on the abnormal feature recognition model through the correction data matching result.
Specifically, performing anomaly verification (anomaly verification: performing anomaly consistency verification for the technical difficulty of unaddressed in the enterprise) on the anomaly detection result to obtain an anomaly verification result, wherein the anomaly verification result comprises a first type unaddressed defect set, a second type unaddressed defect set and a Z type unaddressed defect set, and performing anomaly identification feature clustering (by a K-media (center point) algorithm, namely simply selecting the position center in the first type unaddressed defect set/the second type unaddressed defect set/the Z type unaddressed defect set as a reference point, performing self-bottom upward condensation hierarchical clustering analysis, and iterating until the distribution of the first type unaddressed defect set/the second type unaddressed defect set/the Z type unaddressed defect set is unchanged), thereby obtaining an anomaly identification clustering result;
transmitting the abnormal recognition clustering result to the first mechanism for data matching (if the abnormal recognition clustering result corresponds to a size defect, the largest coverage of the size defect mark is used as a correction data matching result, and a large amount of research investment is carried out on the aspect of indicating the size defect when the largest coverage of the size defect mark is the largest, so that data resources are the most abundant), and a correction data matching result is obtained (in the first mechanism, the correction data matching result is only used as model correction training under the condition that user reading access is not carried out); and carrying out correction training on the abnormal characteristic recognition model by adopting an incremental learning mode through the correction data matching result, screening out data which is most suitable for correction training, and providing support for accelerating the incremental learning efficiency (generally, if a training sample comprises various types of data of various indexes, the learning efficiency is low).
In summary, the method and the system for intelligently detecting the quality of the silicon carbide wafer provided by the embodiment of the application have the following technical effects:
1. because of adopting the connection positioning carrier, receive the positioning signal; when the display positioning is completed, acquiring an image acquisition result; constructing a basic comparison feature set based on the silicon carbide wafer, comparing features of the image acquisition result, and determining an abnormal region; scanning an abnormal region to obtain an abnormal region image acquisition result; collecting abnormal characteristics of the same type of wafers, classifying and identifying and constructing an abnormal characteristic identification model; the method and the system for intelligently detecting the quality of the silicon carbide wafer have the technical effects that the defect detection of the silicon carbide wafer is automatically carried out, and the defect detection efficiency is improved.
2. Because the sequence evaluation of the abnormal feature call is performed based on the abnormal detection result, an abnormal call speed evaluation result is generated; generating a correction weight; and compensating the correction weight to the calling value, and optimizing and compensating the sequential calling in a mode of compensating to the calling value, so that support is provided for rapidly and efficiently checking the abnormal silicon carbide wafer.
Embodiment two:
based on the same inventive concept as the intelligent quality detection method of a silicon carbide wafer in the foregoing embodiment, as shown in fig. 4, an embodiment of the present application provides an intelligent quality detection system of a silicon carbide wafer, where the system includes:
a positioning signal receiving module 100, configured to connect to a positioning carrier and receive a positioning signal of the positioning carrier;
the first image acquisition module 200 is configured to acquire an image of the silicon carbide wafer located on the positioning stage by using an image acquisition device when the received positioning signal indicates that positioning is completed, so as to obtain an image acquisition result;
the feature comparison module 300 is configured to construct a basic comparison feature set based on the silicon carbide wafer, and perform feature comparison on the image acquisition result through the basic comparison feature set to determine an abnormal region;
the second image acquisition module 400 is configured to acquire an image of the abnormal region by using a scanning electron microscope, so as to obtain an image acquisition result of the abnormal region;
the classification identification module 500 is used for collecting abnormal characteristics of the silicon carbide wafers of the same type, performing classification identification, and constructing an abnormal characteristic identification model according to the classification identification result;
The initial recognition module 600 is configured to input the abnormal region image acquisition result into the abnormal feature recognition model, perform initial recognition on the abnormal region image acquisition result through a feature call evaluation unit, and call an abnormal feature set;
the abnormal feature matching module 700 is configured to perform abnormal feature matching on the abnormal region image acquisition result through the abnormal feature set, and output an abnormal detection result.
Further, the system includes:
the first classification identification result acquisition module is used for acquiring a public abnormal data set of the same type of wafers, classifying and identifying the public abnormal data set, and acquiring a first classification identification result;
the coverage identification result generation module is used for carrying out data coverage evaluation on the first classification identification result and generating a coverage identification result;
the abnormal feature recognition model construction module is used for training an initial abnormal feature recognition model based on the first classification identification result and constructing the abnormal feature recognition model through the coverage identification and the coverage identification result.
Further, the system includes:
the decryption key generation module is used for carrying out data encryption on the public abnormal data set and generating a decryption key;
The data sending module is used for sending the encrypted public abnormal data set, the decryption key, the initial abnormal characteristic identification model and the coverage identification result to a first mechanism;
a compensation data set obtaining module, configured to obtain a compensation data set through the coverage identifier result matching, where the compensation data set is an encrypted data set in the first mechanism;
and the optimization training module is used for carrying out model optimization training on the initial abnormal feature recognition model through the compensation data set, and completing construction of the abnormal feature recognition model according to an optimization training result.
Further, the system includes:
the first sequential feature evaluation value generation module is used for carrying out feature recognition according to the feature comparison result to generate a first sequential feature evaluation value of the abnormal type feature, wherein the first sequential feature evaluation value is provided with a position mark;
the characteristic recurrence frequency analysis module is used for obtaining historical detection data of the silicon carbide wafer, carrying out abnormal characteristic recurrence frequency analysis on the historical detection data based on time, and generating a second sequence characteristic evaluation value according to an analysis result;
The feature calling sequence constraint module is used for obtaining feature calling sequence constraint with a position identifier through calculation of the first sequence feature evaluation value and the second sequence feature evaluation value;
and the sequence calling module is used for sequentially calling the abnormal features based on the position identification through the feature calling sequence constraint before the feature calling evaluation unit performs feature recognition on the abnormal region image acquisition result.
Further, the system includes:
the multi-level transition level set setting module is used for setting a multi-level transition level set of the first sequential characteristic evaluation values;
the coefficient allocation adjustment value acquisition module is used for inputting the first sequential characteristic evaluation value into the multi-level transition grade set to obtain a coefficient allocation adjustment value;
the feature calling sequence constraint module is used for carrying out coefficient adjustment on the initial distribution coefficient through the coefficient distribution adjustment value, calculating to obtain the calling value of each feature, determining the feature calling sequence constraint according to the calling value sequence ordering result, and the calculation formula is as follows:wherein K is a calling value, ++>For the first order feature evaluation value, +.>For the second sequential characteristic evaluation value, T is the adjusted second sequential characteristic evaluation value influence coefficient, and +. >Wherein a is the initial coefficient of the first sequential feature evaluation value, b is the coefficient assignment adjustment value, +.>To adjust the influence constant, and->The value is +.>，/>The value at transition level m is +.>。
Further, the system includes:
the abnormality verification module is used for carrying out abnormality verification on the abnormality detection result and carrying out abnormality recognition feature clustering based on the abnormality verification result;
the data matching module is used for sending the abnormal recognition clustering result to the first mechanism for data matching to obtain a corrected data matching result;
and the correction training module is used for carrying out correction training on the abnormal characteristic recognition model through the correction data matching result.
Further, the system includes:
the sequence evaluation module is used for carrying out sequence evaluation of abnormal feature call based on the abnormal detection result and generating an abnormal call speed evaluation result;
the correction weight generation module is used for generating correction weights according to the abnormal calling speed evaluation results;
and the call value compensation module is used for compensating the correction weight to the call value.
Any of the steps of the methods described above may be stored as computer instructions or programs in a non-limiting computer memory and may be called by a non-limiting computer processor to identify any of the methods to implement embodiments of the present application, without unnecessary limitations.
Further, the first or second element may not only represent a sequential relationship, but may also represent a particular concept, and/or may be selected individually or in whole among a plurality of elements. It will be apparent to those skilled in the art that various modifications and variations can be made in the present application without departing from the scope of the application. Thus, if such modifications and variations of the present application fall within the scope of the present application and the equivalents thereof, the present application is intended to cover such modifications and variations.
Claims (7)
1. The intelligent quality detection method for the silicon carbide wafer is characterized by being applied to an intelligent quality detection system, wherein the intelligent quality detection system is in communication connection with a positioning carrier, an image acquisition device and a scanning electron microscope, and the method comprises the following steps:
the positioning carrier is connected with the positioning signal receiving device;
when the received positioning signals are displayed and positioned, the image acquisition device is used for acquiring images of the silicon carbide wafer positioned on the positioning carrier to obtain an image acquisition result;
constructing a basic comparison feature set based on the silicon carbide wafer, and comparing the features of the image acquisition result through the basic comparison feature set to determine an abnormal region;
Acquiring an image of the abnormal region through the scanning electron microscope to obtain an image acquisition result of the abnormal region;
collecting abnormal characteristics of the same type of wafers of the silicon carbide wafers, carrying out classification identification, and constructing an abnormal characteristic identification model according to the classification identification result;
after the abnormal region image acquisition result is input into the abnormal feature recognition model, initially recognizing the abnormal region image acquisition result through a feature calling evaluation unit, and calling an abnormal feature set;
performing abnormal feature matching on the abnormal region image acquisition result through the abnormal feature set, and outputting an abnormal detection result;
acquiring a public abnormal data set of the same type of wafers, and carrying out classification identification on the public abnormal data set to obtain a first classification identification result;
performing data coverage evaluation on the first classification identification result, calculating coverage, and generating a coverage identification result, wherein the coverage identification result comprises coverage of an external scratch defect identification, coverage of an internal bubble defect identification and coverage of a size defect identification;
training an initial abnormal feature recognition model based on the first classification identification result, and constructing the abnormal feature recognition model through the coverage identification and the coverage identification result, wherein the method comprises the following steps: based on the first classification identification result as a data base, taking a BP network model as a model base, taking an external scratch defect identification existing in a public abnormal data set as a first training set, taking an internal bubble defect identification existing in the public abnormal data set as a second training set, taking a size defect identification existing in the public abnormal data set as a third training set, performing training supervision until output reaches convergence, and obtaining an initial abnormal characteristic identification model, wherein the initial abnormal characteristic identification model comprises an external scratch defect identification layer, an internal bubble defect identification layer and a size defect identification layer, and marking the external scratch defect identification layer, the internal bubble defect identification layer and the size defect identification layer through the coverage identification and the coverage identification result respectively to obtain an abnormal characteristic identification model.
2. The method of claim 1, wherein the method comprises:
encrypting the data of the public abnormal data set and generating a decryption key;
transmitting the encrypted public abnormal data set, the decryption key, the initial abnormal feature identification model and the coverage identification result to a first mechanism;
obtaining a compensation data set through the coverage identification result matching, wherein the compensation data set is an encrypted data set in the first mechanism;
and carrying out model optimization training on the initial abnormal feature recognition model through the compensation data set, and completing construction of the abnormal feature recognition model according to an optimization training result.
3. The method of claim 1, wherein the feature comparison of the image acquisition results by the base comparison feature set further comprises:
performing feature recognition according to the feature comparison result, and generating a first sequential feature evaluation value of the abnormal type feature, wherein the first sequential feature evaluation value is provided with a position mark;
obtaining historical detection data of the silicon carbide wafer, carrying out abnormal characteristic recurrence frequency analysis on the historical detection data based on time, and generating a second sequence characteristic evaluation value according to an analysis result;
Calculating to obtain feature calling sequence constraint with position identification through the first sequence feature evaluation value and the second sequence feature evaluation value;
before the feature call evaluation unit performs feature recognition on the abnormal region image acquisition result, sequentially calling the abnormal features based on the position identification through the feature call sequence constraint.
4. A method as claimed in claim 3, wherein the method further comprises:
setting a multi-level transition level set of the first sequential feature evaluation values;
inputting the first sequential characteristic evaluation value into the multi-level transition level set to obtain a coefficient allocation adjustment value;
and carrying out coefficient adjustment on the initial distribution coefficient by the coefficient distribution adjustment value, calculating to obtain the calling value of each feature, and determining the feature calling sequence constraint according to the calling value sequence ordering result, wherein the calculation formula is as follows:
；
wherein, K is a calling value,for the first order feature evaluation value, +.>For the second sequential characteristic evaluation value, T is the adjusted second sequential characteristic evaluation value influence coefficient, and +.>Wherein a is the initial coefficient of the first sequential feature evaluation value, b is the coefficient assignment adjustment value, +. >To adjust the influence constant, and->The value is +.>Indicating that the crystalline state of the silicon carbide wafer corresponds to a first transition level, +.>The value at transition level m is +.>Indicating that the crystalline state of the silicon carbide wafer corresponds to the mth transition level.
5. The method according to claim 2, wherein the method comprises:
performing abnormality verification on the abnormality detection result, and performing abnormality recognition feature clustering based on the abnormality verification result;
transmitting the abnormal recognition clustering result to the first mechanism for data matching to obtain a corrected data matching result;
and carrying out correction training on the abnormal feature recognition model through the correction data matching result.
6. The method of claim 4, wherein the method comprises:
performing sequential evaluation of abnormal feature call based on the abnormal detection result to generate an abnormal call speed evaluation result;
generating a correction weight according to the abnormal calling speed evaluation result;
and compensating the correction weight to the calling value.
7. A system for intelligently detecting the quality of a silicon carbide wafer, which is used for implementing the intelligent detection method for the quality of the silicon carbide wafer according to any one of claims 1 to 6, comprising:
The positioning signal receiving module is used for connecting with the positioning carrier and receiving the positioning signal of the positioning carrier;
the first image acquisition module is used for acquiring images of the silicon carbide wafer positioned on the positioning carrier through the image acquisition device when the received positioning signals are displayed and positioned completely, so that an image acquisition result is obtained;
the feature comparison module is used for constructing a basic comparison feature set based on the silicon carbide wafer, and carrying out feature comparison on the image acquisition result through the basic comparison feature set to determine an abnormal region;
the second image acquisition module is used for acquiring images of the abnormal areas through a scanning electron microscope to obtain an abnormal area image acquisition result;
the classification identification module is used for collecting abnormal characteristics of the silicon carbide wafers of the same type, carrying out classification identification, and constructing an abnormal characteristic identification model according to the classification identification result;
the initial recognition module is used for inputting the abnormal region image acquisition result into the abnormal feature recognition model, initially recognizing the abnormal region image acquisition result through a feature calling evaluation unit, and calling an abnormal feature set;
And the abnormal feature matching module is used for executing abnormal feature matching on the abnormal region image acquisition result through the abnormal feature set and outputting an abnormal detection result.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310778180.5A CN116645362B (en) | 2023-06-29 | 2023-06-29 | Intelligent quality detection method and system for silicon carbide wafer |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310778180.5A CN116645362B (en) | 2023-06-29 | 2023-06-29 | Intelligent quality detection method and system for silicon carbide wafer |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN116645362A CN116645362A (en) | 2023-08-25 |
| CN116645362B true CN116645362B (en) | 2024-01-23 |
Family
ID=87643583
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202310778180.5A Active CN116645362B (en) | 2023-06-29 | 2023-06-29 | Intelligent quality detection method and system for silicon carbide wafer |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN116645362B (en) |
Families Citing this family (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117094989B (en) * | 2023-10-16 | 2024-01-26 | 南通蓬盛机械有限公司 | Lens quality management method and system for optical sighting telescope |
| CN117096068B (en) * | 2023-10-18 | 2023-12-26 | 江苏卓玉智能科技有限公司 | Wafer film coating processing control method and system in IC package |
| CN117269197B (en) * | 2023-11-21 | 2024-02-02 | 昆明理工大学 | Method and device for detecting surface quality of polycrystalline silicon raw material |
| CN117554407B (en) * | 2024-01-10 | 2024-05-24 | 南通纳科达聚氨酯科技有限公司 | On-line detection method and system for photonic crystal coating defects |
| CN118150584B (en) * | 2024-05-10 | 2024-08-02 | 太仓点石航空动力有限公司 | Turbine casing surface quality detection method based on image recognition |
Citations (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN114627089A (en) * | 2022-03-21 | 2022-06-14 | 成都数之联科技股份有限公司 | Defect identification method, defect identification device, computer equipment and computer readable storage medium |
| CN115272329A (en) * | 2022-09-28 | 2022-11-01 | 中科卓芯半导体科技(苏州)有限公司 | Mask production optimization method and system based on image detection |
| CN115409789A (en) * | 2022-08-17 | 2022-11-29 | 广东电网有限责任公司 | Power transmission line engineering defect detection method based on image semantic segmentation |
| CN115797314A (en) * | 2022-12-16 | 2023-03-14 | 哈尔滨耐是智能科技有限公司 | Part surface defect detection method, system, equipment and storage medium |
| CN116130377A (en) * | 2023-04-17 | 2023-05-16 | 西安奕斯伟材料科技股份有限公司 | Method, device and system for detecting defects of epitaxial wafer and manufacturing method thereof |
| CN116188475A (en) * | 2023-05-05 | 2023-05-30 | 德中(深圳)激光智能科技有限公司 | Intelligent control method, system and medium for automatic optical detection of appearance defects |
| CN116223519A (en) * | 2022-12-09 | 2023-06-06 | 视睿（杭州）信息科技有限公司 | Wafer detection equipment, detection method, system and readable storage medium |
| CN116245810A (en) * | 2022-12-30 | 2023-06-09 | 阿丘机器人科技(苏州)有限公司 | Flat cable welding point defect detection method, equipment, storage medium and device |
-
2023
- 2023-06-29 CN CN202310778180.5A patent/CN116645362B/en active Active
Patent Citations (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN114627089A (en) * | 2022-03-21 | 2022-06-14 | 成都数之联科技股份有限公司 | Defect identification method, defect identification device, computer equipment and computer readable storage medium |
| CN115409789A (en) * | 2022-08-17 | 2022-11-29 | 广东电网有限责任公司 | Power transmission line engineering defect detection method based on image semantic segmentation |
| CN115272329A (en) * | 2022-09-28 | 2022-11-01 | 中科卓芯半导体科技(苏州)有限公司 | Mask production optimization method and system based on image detection |
| CN116223519A (en) * | 2022-12-09 | 2023-06-06 | 视睿（杭州）信息科技有限公司 | Wafer detection equipment, detection method, system and readable storage medium |
| CN115797314A (en) * | 2022-12-16 | 2023-03-14 | 哈尔滨耐是智能科技有限公司 | Part surface defect detection method, system, equipment and storage medium |
| CN116245810A (en) * | 2022-12-30 | 2023-06-09 | 阿丘机器人科技(苏州)有限公司 | Flat cable welding point defect detection method, equipment, storage medium and device |
| CN116130377A (en) * | 2023-04-17 | 2023-05-16 | 西安奕斯伟材料科技股份有限公司 | Method, device and system for detecting defects of epitaxial wafer and manufacturing method thereof |
| CN116188475A (en) * | 2023-05-05 | 2023-05-30 | 德中(深圳)激光智能科技有限公司 | Intelligent control method, system and medium for automatic optical detection of appearance defects |
Also Published As
| Publication number | Publication date |
|---|---|
| CN116645362A (en) | 2023-08-25 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN116645362B (en) | Intelligent quality detection method and system for silicon carbide wafer | |
| US20230141886A1 (en) | Method for assessing hazard on flood sensitivity based on ensemble learning | |
| TW483080B (en) | Defect reference system automatic pattern classification | |
| US10290088B2 (en) | Wafer and lot based hierarchical method combining customized metrics with a global classification methodology to monitor process tool condition at extremely high throughput | |
| CN106290378B (en) | Defect classification method and defect inspecting system | |
| Thilker et al. | H ii regions and diffuse ionized gas in 11 nearby spiral galaxies | |
| CN110045015B (en) | Concrete structure internal defect detection method based on deep learning | |
| Banerji et al. | Combining dark energy survey science verification data with near-infrared data from the ESO VISTA hemisphere survey | |
| CN107407648B (en) | Adaptively create disturbances to filter | |
| TW201350836A (en) | Optimization of unknown defect rejection for automatic defect classification | |
| TW201407154A (en) | Integration of automatic and manual defect classification | |
| CN104169945A (en) | Two stage categorization of objects in images | |
| CN105917455B (en) | It fails and finds by using the pattern of the alternately nominal characteristic of failure mode | |
| CN105184225B (en) | A kind of multinational banknote image recognition methods and device | |
| CN113256636A (en) | Bottom-up parasite species development stage and image pixel classification method | |
| CN109919908A (en) | The method and apparatus of light-emitting diode chip for backlight unit defects detection | |
| TW535207B (en) | Method for automatically controlling defect-specification in manufacturing process of semiconductors | |
| CN112948937B (en) | Intelligent pre-judging method and device for concrete strength | |
| CN110308182A (en) | Wafer defect detection method and device | |
| CN108268901A (en) | A kind of algorithm that environmental monitoring abnormal data is found based on dynamic time warping distance | |
| CN117173640B (en) | Beef cattle temperature anomaly detection method | |
| CN102101112A (en) | Light emitting diode wafer sorting method | |
| CN110350993A (en) | A kind of black broadcast automatic discovering method based on networking monitoring under big data scene | |
| CN118174276A (en) | Automatic phase sequence identification method and device for low-voltage photovoltaic platform region | |
| US20050075841A1 (en) | Automated defect classification system and method |
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 |