CN116434005A - Wafer defect data enhancement method and device - Google Patents
Wafer defect data enhancement method and device Download PDFInfo
- Publication number
- CN116434005A CN116434005A CN202310321613.4A CN202310321613A CN116434005A CN 116434005 A CN116434005 A CN 116434005A CN 202310321613 A CN202310321613 A CN 202310321613A CN 116434005 A CN116434005 A CN 116434005A
- Authority
- CN
- China
- Prior art keywords
- data
- wafer map
- target
- data set
- map data
- 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.)
- Pending
Links
- 230000007547 defect Effects 0.000 title claims abstract description 87
- 238000000034 method Methods 0.000 title claims abstract description 33
- 238000002372 labelling Methods 0.000 claims abstract description 8
- 239000013598 vector Substances 0.000 claims description 38
- 230000002708 enhancing effect Effects 0.000 claims description 9
- 230000006740 morphological transformation Effects 0.000 claims description 9
- 230000003321 amplification Effects 0.000 claims description 5
- 238000003199 nucleic acid amplification method Methods 0.000 claims description 5
- 238000005520 cutting process Methods 0.000 claims description 3
- 238000009826 distribution Methods 0.000 description 7
- 230000008569 process Effects 0.000 description 5
- 230000009471 action Effects 0.000 description 3
- 238000010586 diagram Methods 0.000 description 3
- 238000004519 manufacturing process Methods 0.000 description 3
- 230000000694 effects Effects 0.000 description 2
- 238000003860 storage Methods 0.000 description 2
- 238000001514 detection method Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 230000003287 optical effect Effects 0.000 description 1
- 239000004065 semiconductor Substances 0.000 description 1
Images
Classifications
-
- 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
-
- 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)
- Evolutionary Computation (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Computing Systems (AREA)
- Databases & Information Systems (AREA)
- Health & Medical Sciences (AREA)
- General Health & Medical Sciences (AREA)
- Medical Informatics (AREA)
- Software Systems (AREA)
- Artificial Intelligence (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Multimedia (AREA)
- Testing Or Measuring Of Semiconductors Or The Like (AREA)
Abstract
The invention relates to a wafer defect data enhancement method and a device, wherein the method comprises the following steps: acquiring a wafer map data set, wherein the wafer map data set comprises wafer map data of a plurality of defect types; determining a plurality of training data sets according to the wafer map data sets, wherein any training data set comprises wafer map data with the same single defect type; training variable self-encoders (VAEs) for target data sets corresponding to target defect types in the plurality of training data sets by using the target data sets to obtain the VAEs corresponding to the target defect types; and generating a result data set by using a decoder of the VAE corresponding to the target defect type, wherein the result data set contains wafer map data with the target defect type as a labeling label of the target defect type.
Description
Technical Field
The present invention relates to the field of semiconductor manufacturing, and in particular, to a method and apparatus for enhancing wafer defect data.
Background
In the field of wafer map defect classification identification, the authenticity of wafer data and the size of valid data are very critical factors. Because the wafer data is closely related to the manufacturing process, and relates to a series of safety problems, a large amount of real wafer data cannot be directly obtained from the disclosed approach at present. Thus, existing studies are generally based on a limited number of real data sets, and some are based on artificially constructed defect data. However, both of these data have their own drawbacks: for a real data set, the marked data are few, the data distribution is unbalanced, the label accuracy is low, and most of defect patterns are not clean enough and cannot be directly used for a target detection task; for artificially constructed wafer data, there may be a large difference from the real data, and the distortion degree is high, so that a model trained based on the artificial data may perform poorly in the face of the real data.
In the prior art, a DCGAN model is also used to generate a defect pattern with a small number of partial samples, but the effect of generating an countermeasure network is poor for defect types with poor spatial symmetry and image consistency such as Loc and Scratch.
Disclosure of Invention
One or more embodiments of the present disclosure describe a method and apparatus for enhancing wafer defect data, based on an existing real dataset, by extracting a small amount of single-feature, clean-patterned wafer map data from the dataset, training a variational self-encoder VAE generation model based on the small amount of data, and then generating a batch of wafer data using the trained generation model, thereby enhancing the wafer data volume.
In a first aspect, a method for enhancing wafer defect data is provided, including:
acquiring a wafer map data set, wherein the wafer map data set comprises wafer map data of a plurality of defect types;
determining a plurality of training data sets according to the wafer map data sets, wherein any training data set comprises wafer map data with the same single defect type;
training variable self-encoders (VAEs) for target data sets corresponding to target defect types in the plurality of training data sets by using the target data sets to obtain the VAEs corresponding to the target defect types;
and generating a result data set by using a decoder of the VAE corresponding to the target defect type, wherein the result data set contains wafer map data with the target defect type as a labeling label of the target defect type.
In one possible embodiment, before training the variable self-encoder VAE using the target data set, the method further comprises:
and carrying out morphological transformation operation on any wafer map data in the target data set to obtain a plurality of amplified wafer map data, and placing the amplified wafer map data in the target data set.
In one possible embodiment, the morphological transformation operation includes at least: reversing, rotating, amplifying and cutting.
In one possible implementation, training the variable self-encoder VAE using the target data set includes:
inputting the wafer map data in the target data set into an encoder of the VAE for encoding to obtain hidden space vectors;
inputting the hidden space vector into a decoder of the VAE for decoding to obtain decoded data;
values of parameters in the VAE are adjusted by minimizing errors in the decoded data and the wafer map data.
In one possible implementation, the error is a sum of a reconstruction loss and a KL divergence between the decoded data and the wafer map data.
In one possible implementation, generating a result data set using a decoder of a VAE corresponding to the target defect type includes:
acquiring a vector data set, wherein the dimension of a vector in the vector data set is the same as the dimension of a hidden space vector generated during training of the VAE;
and inputting the vector data set into a decoder of the VAE corresponding to the target defect type to obtain a result data set.
In one possible implementation, the acquiring the vector dataset includes:
a number of vectors having dimensions that are dimensions of the hidden space vector are randomly generated and organized into a vector dataset.
In one possible implementation, the wafer map dataset is a WM-811K dataset.
In a second aspect, there is provided a wafer defect data enhancement apparatus, comprising:
a data acquisition unit configured to acquire a wafer map data set including wafer map data of a plurality of defect types;
a data determining unit configured to determine a plurality of training data sets according to the wafer map data sets, wherein any one of the training data sets contains wafer map data having the same single defect type;
the model training unit is configured to train the variable self-encoder VAE for the target data set corresponding to the target defect type in the plurality of training data sets by using the target data set to obtain the VAE corresponding to the target defect type;
and a result generation unit configured to generate a result data set using a decoder of the VAE corresponding to the target defect type, the result data set including wafer map data having the target defect type as its labeling label.
In one possible embodiment, the apparatus further comprises:
and the data amplification unit is configured to perform morphological transformation operation on any wafer map data in the target data set to obtain a plurality of amplified wafer map data, and put the amplified wafer map data into the target data set.
One or more embodiments of the present disclosure describe a method and apparatus for enhancing wafer defect data, where a VAE generation model is selected, and since a wafer map is actually two-dimensional data, the VAE has advantages in generating such data, and learns the generation process and data distribution of input data at the same time, so that samples with a relatively high diversity can be better generated.
Drawings
In order to more clearly illustrate the technical solutions of the embodiments disclosed in the present specification, the drawings required for the description of the embodiments will be briefly described below, and it is apparent that the drawings in the following description are only examples of the embodiments disclosed in the present specification, and other drawings may be obtained according to these drawings without inventive effort for a person skilled in the art.
FIG. 1 is a block diagram of a method for enhancing wafer defect data according to an embodiment of the present invention;
FIG. 2 is a flow chart of a method for enhancing wafer defect data according to an embodiment of the present invention;
fig. 3 is a schematic block diagram of a wafer defect data enhancement device according to an embodiment of the present invention.
Detailed Description
For the purpose of making the objects, technical solutions and advantages of the embodiments of the present invention more apparent, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention, and it is apparent that the described embodiments are some embodiments of the present invention, but not all embodiments of the present invention. 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.
Fig. 1 illustrates a framework for a wafer defect data enhancement method, according to one embodiment. As shown in fig. 1, the framework used in the method mainly comprises two parts: model training and defect wafer map generation. The model training part is used for training a VAE (variable value) generation model according to the existing real wafer map data sets, specifically, classifying and selecting data in the wafer map data sets according to defect types to generate a plurality of single defect type data sets, and then performing data amplification on the wafer map in each data set to obtain a plurality of amplified data sets. Multiple VAEs are trained sequentially using multiple amplification data sets, resulting in multiple VAEs that can generate a single defect type. The defect wafer map generation portion is configured to generate a wafer map dataset of a plurality of single defect types using a plurality of trained decoders of the VAEs.
The following description will proceed with reference being made to the drawings, which are not intended to limit the scope of embodiments of the invention.
Fig. 2 is a flowchart of a method for enhancing wafer defect data according to an embodiment of the present invention. As shown in fig. 2, the method at least includes: step 201, obtaining a wafer map data set, wherein the wafer map data set comprises wafer map data of a plurality of defect types; step 202, determining a plurality of training data sets according to the wafer map data sets, wherein any training data set contains wafer map data with the same single defect type; step 204, training the variable self-encoder VAE for the target data set corresponding to the target defect type in the plurality of training data sets by using the target data set to obtain the VAE corresponding to the target defect type; step 205, generating a result data set by using the decoder of the VAE corresponding to the target defect type, wherein the result data set contains wafer map data with the target defect type as a labeling label thereof.
In step 201, a wafer map dataset is obtained, the wafer map dataset comprising wafer map data for a number of defect types.
In one embodiment, the wafer map dataset is the WM-811K dataset disclosed by Taiwan integrated circuit manufacturing Inc. (Taiwan electric).
In step 202, a plurality of training data sets are determined from the wafer map data sets, wherein any one training data set contains wafer map data having the same single defect type.
In one embodiment, the plurality of training data sets includes wafer maps with clean and clear images and distinct features.
In step 204, training the variable self-encoder VAE using the target data set for a target data set of the plurality of training data sets corresponding to the target defect type to obtain a VAE corresponding to the target defect type.
Specifically, inputting wafer map data in the target data set into an encoder of the VAE for encoding to obtain a hidden space vector z; inputting the hidden space vector into a decoder of the VAE for decoding to obtain decoded data; values of parameters in the VAE are adjusted by minimizing errors in the decoded data and the wafer map data. Wherein the error is a sum of a reconstruction loss between the decoded data and the wafer map data and a KL divergence (Kullback-Leibler Divergence).
The reconstruction loss is used to measure the error between the output generated by the VAE decoder and the original input. As the reconstruction error, a mean square error (Mean Squared Error, MSE) or Cross Entropy (Cross Entropy) may be used. When using the mean square error as the reconstruction error, for one sample x, the reconstruction error thereof can be expressed as shown in the formula (1):
wherein,,
is the output of the VAE decoder, |·|| 2 Representing the square norm.
When cross entropy is used as the reconstruction error, the reconstruction error can be expressed as shown in formula (2):
where n is the dimension of the input data, x i And
respectively representing the i-th element of the input data and the i-th element of the decoder output.
The KL divergence measures the distance between the distribution of the hidden space vector z and the standard normal distribution, and the specific form is shown in a formula (3):
where k is the dimension of the hidden space vector z, μ i Sum sigma i The mean and variance of the ith dimension of the hidden space vector z, respectively. Here, it is assumed that the blank isThe distribution of the inter-vector z is a gaussian distribution, so a specific form of KL-divergence loss can be derived from the KL-divergence formula between two gaussian distributions.
In step 205, a decoder using the VAE corresponding to the target defect type generates a result dataset containing wafer map data having the target defect type as its labeling.
Specifically, a vector data set is obtained, wherein the dimension of a vector in the vector data set is the same as the dimension of a hidden space vector z generated when training the VAE; and inputting the vector data set into a decoder of the VAE corresponding to the target defect type to obtain a result data set.
It should be appreciated that for any of a number of defect types included in the wafer map data set, an independent VAE generation model is trained, and a plurality of VAE generation models are in one-to-one correspondence with a plurality of defect types in the wafer map data set. After the vector data set is input to the decoder of the VAE generation model corresponding to a certain defect type, the decoder generates wafer map data for that defect type.
In one embodiment, the acquiring the vector dataset includes:
a plurality of vectors with dimensions being the dimensions of the hidden space vector z are randomly generated and formed into a vector data set.
In some possible embodiments, prior to step 204, the method further comprises: and 203, performing morphological transformation operation on any wafer map data in the target data set to obtain a plurality of amplified wafer map data, and placing the amplified wafer map data in the target data set.
In one embodiment, the morphological transformation operation includes at least: reversing, rotating, amplifying and cutting. Wherein the rotation may be any angle or random angle rotation.
Through step 203, the data set can be amplified under the condition that the data amount in the original target data set is smaller, so as to obtain better training effect in the subsequent training process of the model.
Fig. 3 is a schematic block diagram of a wafer defect data enhancement device according to an embodiment of the present invention. The apparatus 300 includes:
a data acquisition unit 301 configured to acquire a wafer map data set, the wafer map data set including wafer map data of a plurality of defect types; a data determining unit 302 configured to determine a plurality of training data sets according to the wafer map data sets, wherein any one training data set contains wafer map data having the same single defect type; a model training unit 304 configured to, for a target data set corresponding to a target defect type in the plurality of training data sets, train the variable self-encoder VAE using the target data set to obtain a VAE corresponding to the target defect type; the result generation unit 305 is configured to generate a result data set using the decoder of the VAE corresponding to the target defect type, the result data set containing wafer map data having the target defect type as its labeling label.
In one embodiment, the apparatus further comprises:
the data amplification unit 303 is configured to perform morphological transformation operation on any wafer map data in the target data set to obtain a plurality of amplified wafer map data, and put the plurality of amplified wafer map data in the target data set.
It is noted that relational terms such as first and second, and the like are used solely to distinguish one entity or action from another entity or action without necessarily requiring or implying any actual such relationship or order between such entities or actions. Moreover, the terms "comprises," "comprising," or any other variation thereof, are intended to cover a non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements does not include only those elements but may include other elements not expressly listed or inherent to such process, method, article, or apparatus. Without further limitation, an element defined by the phrase "comprising one … …" does not exclude the presence of other like elements in a process, method, article, or apparatus that comprises an element.
It will be understood by those skilled in the art that all or part of the steps for implementing the above embodiments may be implemented by hardware, or may be implemented by a program for instructing relevant hardware, and the program may be stored in a computer readable storage medium, where the storage medium may be a read-only memory, a magnetic disk or an optical disk, etc.
The foregoing description of the embodiments has been provided for the purpose of illustrating the general principles of the invention, and is not meant to limit the scope of the invention, but to limit the invention to the particular embodiments, and any modifications, equivalents, improvements, etc. that fall within the spirit and principles of the invention are intended to be included within the scope of the invention.
Claims (10)
1. A method of enhancing wafer defect data, comprising:
acquiring a wafer map data set, wherein the wafer map data set comprises wafer map data of a plurality of defect types;
determining a plurality of training data sets according to the wafer map data sets, wherein any training data set comprises wafer map data with the same single defect type;
training variable self-encoders (VAEs) for target data sets corresponding to target defect types in the plurality of training data sets by using the target data sets to obtain the VAEs corresponding to the target defect types;
and generating a result data set by using a decoder of the VAE corresponding to the target defect type, wherein the result data set contains wafer map data with the target defect type as a labeling label of the target defect type.
2. The method of claim 1, wherein prior to training the variable self-encoder VAE using the target data set, the method further comprises:
and carrying out morphological transformation operation on any wafer map data in the target data set to obtain a plurality of amplified wafer map data, and placing the amplified wafer map data in the target data set.
3. The method of claim 2, wherein the morphological transformation operation comprises at least: reversing, rotating, amplifying and cutting.
4. The method of claim 1, wherein training a variable self-encoder VAE using the target data set comprises:
inputting the wafer map data in the target data set into an encoder of the VAE for encoding to obtain hidden space vectors;
inputting the hidden space vector into a decoder of the VAE for decoding to obtain decoded data;
values of parameters in the VAE are adjusted by minimizing errors in the decoded data and the wafer map data.
5. The method of claim 4, wherein the error is a sum of a reconstruction loss and a KL divergence between the decoded data and the wafer map data.
6. The method of claim 1, wherein generating a result dataset using a decoder of a VAE corresponding to the target defect type, comprises:
acquiring a vector data set, wherein the dimension of a vector in the vector data set is the same as the dimension of a hidden space vector generated during training of the VAE;
and inputting the vector data set into a decoder of the VAE corresponding to the target defect type to obtain a result data set.
7. The method of claim 6, wherein the acquiring the vector dataset comprises:
a number of vectors having dimensions that are dimensions of the hidden space vector are randomly generated and organized into a vector dataset.
8. The method of claim 1, wherein the wafer map dataset is a WM-811K dataset.
9. A wafer defect data enhancement apparatus comprising:
a data acquisition unit configured to acquire a wafer map data set including wafer map data of a plurality of defect types;
a data determining unit configured to determine a plurality of training data sets according to the wafer map data sets, wherein any one of the training data sets contains wafer map data having the same single defect type;
the model training unit is configured to train the variable self-encoder VAE for the target data set corresponding to the target defect type in the plurality of training data sets by using the target data set to obtain the VAE corresponding to the target defect type;
and a result generation unit configured to generate a result data set using a decoder of the VAE corresponding to the target defect type, the result data set including wafer map data having the target defect type as its labeling label.
10. The apparatus of claim 9, wherein the apparatus further comprises:
and the data amplification unit is configured to perform morphological transformation operation on any wafer map data in the target data set to obtain a plurality of amplified wafer map data, and put the amplified wafer map data into the target data set.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310321613.4A CN116434005A (en) | 2023-03-29 | 2023-03-29 | Wafer defect data enhancement method and device |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310321613.4A CN116434005A (en) | 2023-03-29 | 2023-03-29 | Wafer defect data enhancement method and device |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN116434005A true CN116434005A (en) | 2023-07-14 |
Family
ID=87093648
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202310321613.4A Pending CN116434005A (en) | 2023-03-29 | 2023-03-29 | Wafer defect data enhancement method and device |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN116434005A (en) |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111258992A (en) * | 2020-01-09 | 2020-06-09 | 电子科技大学 | Seismic data expansion method based on variational self-encoder |
| WO2021225741A1 (en) * | 2020-05-07 | 2021-11-11 | Microsoft Technology Licensing, Llc | Variational auto encoder for mixed data types |
| CN115115921A (en) * | 2022-07-05 | 2022-09-27 | 广东利扬芯片测试股份有限公司 | Wafer defect mode identification method based on residual error network |
-
2023
- 2023-03-29 CN CN202310321613.4A patent/CN116434005A/en active Pending
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111258992A (en) * | 2020-01-09 | 2020-06-09 | 电子科技大学 | Seismic data expansion method based on variational self-encoder |
| WO2021225741A1 (en) * | 2020-05-07 | 2021-11-11 | Microsoft Technology Licensing, Llc | Variational auto encoder for mixed data types |
| CN115115921A (en) * | 2022-07-05 | 2022-09-27 | 广东利扬芯片测试股份有限公司 | Wafer defect mode identification method based on residual error network |
Non-Patent Citations (1)
| Title |
|---|
| 郭业才 等: "《深度学习与信号处理：原理与实践》", 30 June 2022, 机械工业出版社, pages: 269 - 271 * |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN105825243A (en) | Method and device for certificate image detection | |
| CN107784288A (en) | A kind of iteration positioning formula method for detecting human face based on deep neural network | |
| CN110555399A (en) | Finger vein identification method and device, computer equipment and readable storage medium | |
| CN114155244B (en) | Defect detection method, device, equipment and storage medium | |
| CN112884721B (en) | Abnormality detection method, abnormality detection system and computer-readable storage medium | |
| US20230401691A1 (en) | Image defect detection method, electronic device and readable storage medium | |
| CN113221987A (en) | Small sample target detection method based on cross attention mechanism | |
| CN111371611B (en) | Weighted network community discovery method and device based on deep learning | |
| CN113052577A (en) | Method and system for estimating category of virtual address of block chain digital currency | |
| CN117710378B (en) | Wafer surface defect detection method, system and storage medium based on deep learning | |
| CN114863209A (en) | Class proportion guided unsupervised domain adaptive modeling method, system, device and medium | |
| Kumari et al. | Sylvester Matrix‐Based Similarity Estimation Method for Automation of Defect Detection in Textile Fabrics | |
| CN112967251B (en) | Picture detection method, training method and device of picture detection model | |
| Ren et al. | Barcode detection and decoding method based on deep learning | |
| CN114155524A (en) | Single-stage 3D point cloud target detection method and device, computer equipment and medium | |
| CN113557528A (en) | Generating point cloud completion network and processing point cloud data | |
| CN108122223A (en) | Ferrite defect deep learning identification method based on Fisher criterion | |
| CN116434005A (en) | Wafer defect data enhancement method and device | |
| CN116912144A (en) | Data enhancement method based on discipline algorithm and channel attention mechanism | |
| CN117132564A (en) | YOLOv 3-based sapphire substrate surface defect detection method and system | |
| CN116958809A (en) | Remote sensing small sample target detection method for feature library migration | |
| CN116385778A (en) | Defect classification method of small sample unbalanced wafer data based on deep learning | |
| CN113269137B (en) | Non-matching face recognition method combining PCANet and shielding positioning | |
| CN109145132B (en) | Method and device for extracting hash code from image and image retrieval method and device | |
| CN115797334B (en) | Method and device for rapidly judging quality of large-scale light spots |
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 |