CN117059511B - Method and system for detecting internal stress distribution of wafer material - Google Patents
Method and system for detecting internal stress distribution of wafer material Download PDFInfo
- Publication number
- CN117059511B CN117059511B CN202311312913.2A CN202311312913A CN117059511B CN 117059511 B CN117059511 B CN 117059511B CN 202311312913 A CN202311312913 A CN 202311312913A CN 117059511 B CN117059511 B CN 117059511B
- Authority
- CN
- China
- Prior art keywords
- detection
- wafer
- internal stress
- model
- 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.)
- Active
Links
- 238000009826 distribution Methods 0.000 title claims abstract description 71
- 239000000463 material Substances 0.000 title claims abstract description 39
- 238000000034 method Methods 0.000 title claims abstract description 31
- 238000001514 detection method Methods 0.000 claims abstract description 410
- 238000012545 processing Methods 0.000 claims abstract description 89
- 238000012549 training Methods 0.000 claims abstract description 32
- 230000004927 fusion Effects 0.000 claims abstract description 30
- 238000013135 deep learning Methods 0.000 claims abstract description 8
- 230000007547 defect Effects 0.000 claims description 102
- 230000006870 function Effects 0.000 claims description 17
- 238000012360 testing method Methods 0.000 claims description 12
- 238000000605 extraction Methods 0.000 claims description 10
- 230000009467 reduction Effects 0.000 claims description 9
- 238000010276 construction Methods 0.000 claims description 8
- 238000005516 engineering process Methods 0.000 claims description 8
- 238000000556 factor analysis Methods 0.000 claims description 8
- 238000000513 principal component analysis Methods 0.000 claims description 8
- 238000012795 verification Methods 0.000 claims description 8
- 238000004364 calculation method Methods 0.000 claims description 6
- 230000009977 dual effect Effects 0.000 claims description 6
- 238000004140 cleaning Methods 0.000 claims description 5
- 238000002156 mixing Methods 0.000 claims description 2
- 238000012512 characterization method Methods 0.000 claims 1
- 238000007689 inspection Methods 0.000 claims 1
- 230000000694 effects Effects 0.000 abstract description 8
- 235000012431 wafers Nutrition 0.000 description 191
- 230000008569 process Effects 0.000 description 9
- 238000001914 filtration Methods 0.000 description 6
- 230000003044 adaptive effect Effects 0.000 description 5
- 238000004519 manufacturing process Methods 0.000 description 4
- 238000004458 analytical method Methods 0.000 description 3
- 239000013078 crystal Substances 0.000 description 2
- 230000002411 adverse Effects 0.000 description 1
- 230000002146 bilateral effect Effects 0.000 description 1
- 238000010586 diagram Methods 0.000 description 1
- 230000002708 enhancing effect Effects 0.000 description 1
- 239000012535 impurity Substances 0.000 description 1
- 238000005259 measurement Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000012216 screening Methods 0.000 description 1
- 238000012706 support-vector machine Methods 0.000 description 1
- 239000012780 transparent material Substances 0.000 description 1
Classifications
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01L—MEASURING FORCE, STRESS, TORQUE, WORK, MECHANICAL POWER, MECHANICAL EFFICIENCY, OR FLUID PRESSURE
- G01L1/00—Measuring force or stress, in general
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/25—Fusion techniques
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N20/00—Machine learning
-
- H—ELECTRICITY
- H01—ELECTRIC ELEMENTS
- H01L—SEMICONDUCTOR DEVICES NOT COVERED BY CLASS H10
- H01L22/00—Testing or measuring during manufacture or treatment; Reliability measurements, i.e. testing of parts without further processing to modify the parts as such; Structural arrangements therefor
- H01L22/10—Measuring as part of the manufacturing process
- H01L22/12—Measuring as part of the manufacturing process for structural parameters, e.g. thickness, line width, refractive index, temperature, warp, bond strength, defects, optical inspection, electrical measurement of structural dimensions, metallurgic measurement of diffusions
-
- H—ELECTRICITY
- H01—ELECTRIC ELEMENTS
- H01L—SEMICONDUCTOR DEVICES NOT COVERED BY CLASS H10
- H01L22/00—Testing or measuring during manufacture or treatment; Reliability measurements, i.e. testing of parts without further processing to modify the parts as such; Structural arrangements therefor
- H01L22/20—Sequence of activities consisting of a plurality of measurements, corrections, marking or sorting steps
-
- 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
Abstract
The application relates to the technical field of data processing, and provides a method and a system for detecting internal stress distribution of a wafer material. The method comprises the following steps: obtaining an internal stress detection factor set according to the application standard of the wafer material; classifying the wafer internal stress detection database according to the internal stress detection factor set to obtain a wafer detection factor sample data set; training the wafer detection factor sample data sets respectively by using a deep learning network structure, and fusing the basic internal stress detection branch model sets based on the model fusion coefficient sets to generate a wafer internal stress self-adaptive detection model; and analyzing the processing node detection data flow based on the wafer internal stress self-adaptive detection model to obtain a wafer internal stress distribution detection result. By adopting the method, the intelligent and rapid detection of the internal stress of the wafer can be realized by constructing the self-adaptive detection model, the comprehensiveness of internal stress detection factors is improved, and the technical effects of the internal stress detection precision and the detection efficiency of the wafer are further improved.
Description
Technical Field
The application relates to the technical field of data processing, in particular to a method and a system for detecting internal stress distribution of a wafer material.
Background
The internal stress of a wafer refers to the pressure or tension applied to the inside of the wafer, and during the manufacturing process of the wafer, the internal stress is generated due to the fact that the internal stress is generated by the fact that part of the internal stress remains in the wafer after the internal stress is eliminated due to the effect and influence of factors from various processes and the like. The existence and distribution of the internal stress of the wafer can seriously affect the performance and stability of the wafer, so that the internal stress distribution of the wafer needs to be accurately detected. However, the prior art has low intelligent degree of internal stress detection, and insufficient comprehensiveness of detection factors, so that the detection precision is low.
Disclosure of Invention
Based on the above, it is necessary to provide a method and a system for detecting internal stress distribution of wafer materials, which can realize intelligent and rapid detection of internal stress of a wafer, improve the comprehensiveness of internal stress detection factors, and further improve the detection precision and detection efficiency of internal stress of the wafer.
A method for detecting internal stress distribution of a wafer material, the method comprising: a processing node identification network is laid out and obtained, and a processing node detection data stream of the target wafer is acquired and obtained based on the processing node identification network; obtaining an internal stress detection factor set according to the wafer material application standard, wherein the internal stress detection factor set comprises material characteristics, dimension thickness, surface defects and a processing technology; constructing a wafer internal stress detection database, wherein the wafer internal stress detection database comprises wafer processing detection data and internal stress detection result data; classifying the wafer internal stress detection database according to the internal stress detection factor set to obtain a wafer detection factor sample data set; training the wafer detection factor sample data set by using a deep learning network structure to obtain a basic internal stress detection branch model set; acquiring a model fusion coefficient set, and fusing each branch model in the basic internal stress detection branch model set based on the model fusion coefficient set to generate a wafer internal stress self-adaptive detection model; and analyzing the processing node detection data stream based on the wafer internal stress self-adaptive detection model to obtain a wafer internal stress distribution detection result.
A wafer material internal stress distribution detection system, the system comprising: the detection data stream acquisition module is used for arranging and acquiring a processing node identification network, and acquiring a processing node detection data stream of the target wafer based on the processing node identification network; the detection factor set obtaining module is used for obtaining an internal stress detection factor set according to the application standard of the wafer material, wherein the internal stress detection factor set comprises material characteristics, dimension thickness, surface defects and a processing technology; the wafer internal stress detection database comprises wafer processing detection data and internal stress detection result data; the database classification module is used for classifying the wafer internal stress detection database according to the internal stress detection factor set to obtain a wafer detection factor sample data set; the branch model training obtaining module is used for respectively training the wafer detection factor sample data sets by utilizing a deep learning network structure to obtain a basic internal stress detection branch model set; the self-adaptive detection model generation module is used for acquiring a model fusion coefficient set, and fusing all branch models in the basic internal stress detection branch model set based on the model fusion coefficient set to generate a wafer internal stress self-adaptive detection model; the detection result obtaining module is used for analyzing the processing node detection data stream based on the wafer internal stress self-adaptive detection model to obtain a wafer internal stress distribution detection result.
The method and the system for detecting the internal stress distribution of the wafer material solve the technical problems of low intelligent degree of internal stress detection and low detection precision caused by insufficient comprehensiveness of detection factors in the prior art, and achieve the technical effects of realizing intelligent and rapid detection of the internal stress of the wafer by constructing a self-adaptive detection model, improving comprehensiveness of the internal stress detection factors, and further improving the internal stress detection precision and detection efficiency of the wafer.
The foregoing description is only an overview of the present application, and is intended to be implemented in accordance with the teachings of the present application in order that the same may be more clearly understood and to make the same and other objects, features and advantages of the present application more readily apparent.
Drawings
FIG. 1 is a flow chart illustrating a method for detecting internal stress distribution of a wafer material according to one embodiment;
FIG. 2 is a flow chart of a process node detection data flow for obtaining a target wafer in a wafer material internal stress distribution detection method according to one embodiment;
FIG. 3 is a block diagram of a system for detecting internal stress distribution of a wafer material according to one embodiment;
reference numerals illustrate: the system comprises a detection data flow acquisition module 11, a detection factor set acquisition module 12, a detection database construction module 13, a database classification module 14, a branch model training acquisition module 15, an adaptive detection model generation module 16 and a detection result acquisition module 17.
Detailed Description
The present application will be described in further detail with reference to the drawings and examples, in order to make the objects, technical solutions and advantages of the present application more apparent. It should be understood that the specific embodiments described herein are for purposes of illustration only and are not intended to limit the scope of the application.
As shown in fig. 1, the present application provides a method for detecting internal stress distribution of a wafer material, the method comprising:
step S100: a processing node identification network is laid out and obtained, and a processing node detection data stream of the target wafer is acquired and obtained based on the processing node identification network;
in one embodiment, as shown in fig. 2, the process node detecting data flow for obtaining the target wafer, step S100 of the present application further includes:
step S110: setting a node data acquisition dual channel according to the processing node identification network, wherein the node data acquisition dual channel comprises a structured data identification channel and an image data identification channel;
step S120: respectively acquiring and acquiring a structured detection data stream and a multi-angle image detection data stream of the target wafer based on the structured data identification channel and the image data identification channel;
step S130: performing data cleaning processing on the structured detection data stream to obtain a standard structured detection data stream, and performing denoising pretreatment on the multi-angle image detection data stream to obtain a standard image detection data stream;
step S140: the processing node detection data stream is generated based on the standard structured detection data stream and the standard image detection data stream.
Specifically, the internal stress of the wafer refers to the pressure or tension applied to the wafer, and during the process of manufacturing the wafer, the internal stress is generated due to the effect and influence of factors such as various processes, and when the factors disappear, part of the effect and influence remain in the wafer. The existence and distribution of the internal stress of the wafer can seriously affect the performance and stability of the wafer, so that the internal stress distribution of the wafer needs to be accurately detected. In order to realize comprehensive acquisition of wafer processing data, a processing node identification network is arranged on a wafer processing flow node, wherein the processing node identification network is a sensor data detection network and is composed of various sensors, including a temperature sensor, a humidity sensor, an image acquisition sensor, a thickness sensor and the like.
And acquiring processing flow node data of a target wafer based on the processing node identification network, wherein the target wafer can be one or the same batch of wafers to be detected. Firstly, in order to improve data acquisition efficiency, setting node data acquisition double channels according to the processing node identification network, wherein the node data acquisition double channels are classified according to data acquisition format types, and specifically comprise a structured data identification channel and an image data identification channel. The structured data identification channel is in a structured data format acquired by a sensor, and comprises temperature, humidity, thickness, processing technology data and the like, and the image data identification channel is in a wafer image format and is monitored and acquired by the image acquisition sensor.
And respectively acquiring and acquiring a structural detection data stream and a multi-angle image detection data stream of the target wafer in a processing flow node based on the structural data identification channel and the image data identification channel. And meanwhile, in order to improve the availability of the acquired data, carrying out data cleaning processing on the structured detection data stream, and carrying out cleaning processing on repeated data, error data and the like in the detection data stream to obtain a corresponding standard structured detection data stream. And denoising the multi-angle image detection data stream, wherein an image filtering algorithm can be adopted to filter and denoise noise existing in the image, and common filtering algorithms comprise self-adaptive median filtering, gaussian filtering, bilateral filtering, guide filtering and the like, so that the standard image detection data stream is obtained through denoising, and the image quality is improved. And integrating and generating a processing node detection data stream based on the standard structured detection data stream and the standard image detection data stream to serve as wafer processing flow detection data, and providing a data basis for subsequent wafer internal stress detection analysis. The intelligent data acquisition is realized, and the comprehensiveness of data acquisition and the standardization of data availability are improved.
Step S200: obtaining an internal stress detection factor set according to the wafer material application standard, wherein the internal stress detection factor set comprises material characteristics, dimension thickness, surface defects and a processing technology;
specifically, the wafer material application standard of the target wafer is obtained, namely the internal stress distribution requirement of the applicable standard of the wafer is met, so that the internal stress of the wafer is ensured not to exceed the application limit. And determining an internal stress detection factor set meeting the application standard according to the application standard of the wafer material, wherein the internal stress detection factor set is influence factor information related to the internal stress of the wafer, and comprises material characteristics, dimension thickness, surface defects, processing technology and the like, so that the consideration of the influence factors of the internal stress of the wafer is improved.
Step S300: constructing a wafer internal stress detection database, wherein the wafer internal stress detection database comprises wafer processing detection data and internal stress detection result data;
in one embodiment, the step S300 of the present application further includes:
step S310: performing data acquisition on the wafer processing flow based on the internal stress detection factor set to obtain a wafer structured detection data stream and a wafer image detection data stream;
step S320: carrying out surface defect identification and feature marking on the wafer image detection data stream to obtain a wafer surface defect feature information set;
step S330: determining the wafer processing detection data according to the wafer structural detection data flow and the wafer surface defect characteristic information set;
step S340: acquiring the internal stress detection result data through an internal stress detector, and constructing the wafer internal stress detection database based on the wafer processing detection data and the internal stress detection result data.
In one embodiment, the step S320 of obtaining the wafer surface defect characteristic information set further includes:
step S321: carrying out surface defect identification on the wafer image detection data stream based on a wafer defect identifier to obtain a wafer surface defect information set;
step S322: clustering and marking the wafer surface defect information set according to the wafer defect label set to obtain a surface defect clustering and marking result;
step S323: respectively extracting defect point areas from the wafer surface defect information set to obtain a surface defect characteristic extraction value set;
step S324: and determining the wafer surface defect characteristic information set based on the surface defect clustering marking result and the surface defect characteristic extraction value set.
Specifically, a wafer internal stress detection database is constructed, wherein the wafer internal stress detection database is a historical wafer internal stress processing detection data set and comprises wafer processing detection data and internal stress detection result data. The specific construction process is as follows: and carrying out data acquisition on the wafer processing flow based on the internal stress detection factor set, namely carrying out detection acquisition on the node data of the wafer processing flow through the processing node identification network to obtain corresponding wafer structural detection data flow and wafer image detection data flow, wherein the detection data flow corresponds to the detection data of the internal stress detection factor set. And carrying out surface defect identification and defect characteristic marking on the wafer image detection data stream, and firstly training and constructing a wafer defect identifier through historical wafer defect data, wherein the wafer defect identifier is a defect identification support vector machine and is used for quickly identifying the surface defects of the wafer.
And carrying out surface defect identification on the wafer image detection data stream based on the wafer defect identifier, and outputting the wafer detection data stream for acquiring a wafer surface defect information set, namely the surface defect. And then a wafer defect label set is formulated according to the actual production of the wafer, wherein the wafer defect label set is of a wafer surface defect type and comprises oxide film defects, impurity pollution, mechanical defects and the like. And clustering and marking the wafer surface defect information set according to the wafer defect label set, gathering the surface defect information with the same defect label into one type, and dividing to obtain a surface defect clustering and marking result. And meanwhile, respectively extracting defect point areas from the wafer surface defect information set, namely, extracting and calculating the area size of the surface defect, namely, extracting and calculating the area size of each defect area through calculus, and taking the calculated defect area as a defect characteristic extraction value to obtain a surface defect characteristic extraction value set corresponding to the wafer surface defect information set. And based on the surface defect clustering marking result and the surface defect feature extraction value set, determining a wafer surface defect feature information set in a combined mode, namely, the wafer surface defect feature information set comprises wafer surface defect feature types and defect area sizes.
And integrating and determining wafer processing detection data according to the wafer structural detection data flow and the wafer surface defect characteristic information set, wherein the wafer processing detection data comprises sensor detection data and surface defect characteristic analysis data of wafer processing flow nodes, and the sensor detection data and the surface defect characteristic analysis data correspond to the internal stress detection factor set. And acquiring and obtaining the internal stress detection result data through an internal stress detector, wherein the internal stress detector has detection functions of internal stress distribution measurement, defect screening and the like of the transparent material, and the detection principle is that the internal stress distribution of the wafer material is detected based on the polarized light stress birefringence effect. When the stress concentration exists in the crystal material due to the internal defects, the stress birefringence effect is caused, the polarized light can be modulated in a polarized state when transmitted through the crystal material, and the stress retardation of the material can be calculated by measuring the Stokes vector of the transmitted light, so that the stress distribution data of the material is obtained. And integrating and constructing a wafer internal stress detection database based on the wafer processing detection data and the internal stress detection result data, and taking the wafer internal stress detection database as a model training data basis. The data integrity of the wafer internal stress detection database is improved, and the training data integrity and accuracy of the subsequent internal stress detection model are further improved.
Step S400: classifying the wafer internal stress detection database according to the internal stress detection factor set to obtain a wafer detection factor sample data set;
specifically, the wafer internal stress detection database is classified according to the internal stress detection factor set, namely, detection data in the wafer internal stress detection database are classified and integrated according to the internal stress detection factor set. And taking the classified wafer internal stress detection data as a wafer detection factor sample data set to be used as a training sample of a subsequent internal stress detection model.
Step S500: training the wafer detection factor sample data set by using a deep learning network structure to obtain a basic internal stress detection branch model set;
in one embodiment, the obtaining the basic internal stress detection branch model set, the step S500 of the present application further includes:
step S510: randomly selecting N groups of model sample data from the wafer detection factor sample data set, and respectively performing equal distribution weight layer training on the N groups of model sample data to obtain N internal stress detection branch models;
step S520: model verification is carried out on the N internal stress detection branch models respectively, and N model accuracy rates and a prediction loss value set of the N groups of model sample data are obtained;
step S530: and iteratively updating weight distribution of the N groups of model sample data based on the N model accuracy rates and the prediction loss value sets, and obtaining the basic internal stress detection branch model set.
In one embodiment, the updating the weight distribution of the N groups of model sample data, the applying step S530 further includes:
step S531: a sample weight updating function is constructed, wherein the sample weight updating function specifically comprises the following steps:the method comprises the steps of carrying out a first treatment on the surface of the Wherein (1)>Indicating updated weights for the model sample data,indicating the weight of the ith sample of the Nth test branch model,/for the test branch model>Model accuracy for Nth detection branch model, < ->Predictive value for the nth test model for the ith sample, +.>A correct output value for the ith sample;
step S532: and iteratively updating the weight distribution of the N groups of model sample data based on the calculation result of the weight updating function.
Specifically, training the wafer detection factor sample data sets by using a deep learning network structure, firstly randomly selecting N groups of model sample data from the wafer detection factor sample data sets, and respectively performing equal distribution weight layer training on the N groups of model sample data, namely, the N groups of model training sample data have the same initial weight, the weight of the sample data in each group of model training data is 1/N, and training to obtain N corresponding internal stress detection branch models. The N internal stress detection branch models obtained through training are weak detection branch models, and the model detection accuracy is low, so that model verification can be performed on the N internal stress detection branch models respectively by using a model verification set to obtain corresponding N model accuracy and a prediction loss value set of the N groups of model sample data, wherein the prediction loss value set is a difference value between an actual value of the model sample data and an output prediction value of the model.
And iteratively updating weight distribution of the N groups of model sample data based on the N model accuracy rates and the prediction loss value set so as to improve the output accuracy rate of the weak detection branch model. The specific process of the iterative updating of the weight value is that firstly, a sample weight value updating function is constructed, and the sample weight value updating function is a calculating function of the sample weight value updating function, and is specificThe method comprises the following steps:wherein->Indicating updated weights for the model sample data,indicating the weight of the ith sample of the Nth test branch model,/for the test branch model>Model accuracy for Nth detection branch model, < ->Predictive value for the nth test model for the ith sample, +.>The correct output value for the i-th sample, i.e. the actual value of the sample.
Based on the calculation result of the weight updating function, iteratively updating the weight distribution of the N groups of model sample data, giving larger weight to training data individuals with failed classification through the function calculation, paying more attention to the training individuals during the next iterative operation, and enhancing the sample weight of the previous weak detection branch model with wrong classification for the next iteration. And iterating until the preset error rate or the specified maximum iteration times are reached, training and obtaining a basic internal stress detection branch model set corresponding to the wafer detection factor sample data set, and improving the output accuracy of the detection branch model.
Step S600: acquiring a model fusion coefficient set, and fusing each branch model in the basic internal stress detection branch model set based on the model fusion coefficient set to generate a wafer internal stress self-adaptive detection model;
in one embodiment, the obtaining the model fusion coefficient set, step S600 of the present application further includes:
step S610: obtaining detection attribute information of the internal stress detection factor set, and performing principal component analysis on the detection attribute information to obtain dimension reduction detection attribute information;
step S620: based on factor analysis on the dimension reduction detection attribute information, obtaining a detection factor weight distribution result;
step S630: according to the detection factor weight distribution result, giving influence to each detection factor in the stress detection factor set, and obtaining a detection factor influence coefficient;
step S640: and carrying out parameter blending on the N model accuracy rates by using the detection factor influence coefficient to obtain the model fusion coefficient set.
Specifically, in order to improve the fusion accuracy of the detection branch models, the fusion coefficients of the detection branch models need to be determined. The process is specifically as follows: firstly, acquiring detection attribute information in the internal stress detection factor set, wherein the detection attribute information is a plurality of proposed variables related to internal stress detection factors, including specific parameters of detection factors such as process parameters, material elastic modulus, processing equipment and the like, in order to comprehensively analyze the internal stress distribution detection accuracy. And carrying out principal component analysis on the detection attribute information, namely carrying out dimension reduction treatment on the detection attribute, removing noise mingled with the detection attribute information, and clearly displaying important attribute characteristics so as to obtain attribute information with strong relevance with internal stress distribution detection, namely dimension reduction detection attribute information.
And performing factor analysis on the dimension reduction detection attribute information, namely extracting common features in all detection attributes, so as to classify the attribute information with the same nature into one attribute information. The attribute information with more common factors has larger weight correspondence and the attribute information with less common factors has smaller weight correspondence, so that the weight distribution result of the attribute factors is obtained, namely the weight distribution result of the detection factors is obtained. The system calculation complexity is reduced by utilizing principal component analysis to reduce the dimension, and the weight of each attribute factor is distributed by factor analysis, so that the accuracy and reliability of the weight distribution result are improved. And giving influence to each detection factor in the stress detection factor set according to the detection factor weight distribution result, namely taking the weight distribution value of each detection factor as the influence coefficient of the detection factor, thereby obtaining the corresponding detection factor influence coefficient of each detection factor.
And carrying out parameter fusion on the N model accuracy rates by using the influence coefficient of the detection factors, namely taking the product of the influence coefficient and the model accuracy rate as the fusion coefficient of each detection branch model, wherein the more important the detection factors are, the higher the model training accuracy rate is, and the larger the corresponding model fusion coefficient is. The weight of the weak detection branch model with high detection accuracy and great importance is increased, so that the weak detection branch model plays a great role in voting; the weight of the weak detection branch model with small detection accuracy and small importance is reduced, so that the weak detection branch model plays a small role in voting, each branch model in the basic internal stress detection branch model set is subjected to weighted fusion based on the model fusion coefficient set, a wafer internal stress self-adaptive detection model is generated, namely a strong wafer internal stress self-adaptive detection model formed by the weak detection branch models, and the wafer internal stress distribution detection accuracy is improved.
Step S700: and analyzing the processing node detection data stream based on the wafer internal stress self-adaptive detection model to obtain a wafer internal stress distribution detection result.
Specifically, the processing node detection data flow acquired by identifying the target wafer is analyzed based on the wafer internal stress self-adaptive detection model, and a wafer internal stress distribution detection result is output and obtained, wherein the wafer internal stress distribution detection result can show the internal stress distribution condition of the target wafer. The intelligent rapid detection of the internal stress of the wafer is realized by constructing the self-adaptive detection model, the comprehensiveness of the internal stress detection factors is improved, and the internal stress detection precision and the internal stress detection efficiency of the wafer are further improved. If the internal stress distribution detection result of the wafer exceeds the internal stress distribution application standard of the wafer, adverse effects can be generated on the application performance of the wafer, the internal stress of the wafer is improved and eliminated, and the unqualified mark of the application quality is required, so that the production quality of the wafer is ensured.
In one embodiment, as shown in FIG. 3, a wafer material internal stress distribution detection system is provided, comprising: the system comprises a detection data flow acquisition module 11, a detection factor set acquisition module 12, a detection database construction module 13, a database classification module 14, a branch model training acquisition module 15, an adaptive detection model generation module 16 and a detection result acquisition module 17, wherein:
the detection data stream acquisition module 11 is used for arranging and acquiring a processing node identification network, and acquiring a processing node detection data stream of the target wafer based on the processing node identification network;
a detection factor set obtaining module 12, configured to obtain a set of internal stress detection factors according to the wafer material application standard, where the set of internal stress detection factors includes material characteristics, dimension thickness, surface defects, and a processing technology;
a detection database construction module 13, configured to construct a wafer internal stress detection database, where the wafer internal stress detection database includes wafer processing detection data and internal stress detection result data;
the database classification module 14 is configured to classify the wafer internal stress detection database according to the internal stress detection factor set, and obtain a wafer detection factor sample data set;
the branch model training obtaining module 15 is configured to respectively train the wafer detection factor sample data sets by using a deep learning network structure to obtain a basic internal stress detection branch model set;
the adaptive detection model generation module 16 is configured to obtain a model fusion coefficient set, fuse each branch model in the basic internal stress detection branch model set based on the model fusion coefficient set, and generate a wafer internal stress adaptive detection model;
and the detection result obtaining module 17 is configured to analyze the processing node detection data stream based on the wafer internal stress adaptive detection model, so as to obtain a wafer internal stress distribution detection result.
In one embodiment, the system further comprises:
the data acquisition dual-channel setting unit is used for setting node data acquisition dual channels according to the processing node identification network, wherein the node data acquisition dual channels comprise a structured data identification channel and an image data identification channel;
the data stream acquisition unit is used for respectively acquiring and acquiring a structured detection data stream and a multi-angle image detection data stream of the target wafer based on the structured data identification channel and the image data identification channel;
the data stream processing unit is used for carrying out data cleaning processing on the structured detection data stream to obtain a standard structured detection data stream, and carrying out denoising pretreatment on the multi-angle image detection data stream to obtain a standard image detection data stream;
and the node detection data stream generating unit is used for generating the processing node detection data stream based on the standard structured detection data stream and the standard image detection data stream.
In one embodiment, the system further comprises:
the branch model training unit is used for randomly selecting N groups of model sample data from the wafer detection factor sample data set, and respectively carrying out equal distribution weight layer training on the N groups of model sample data to obtain N internal stress detection branch models;
the branch model verification unit is used for respectively carrying out model verification on the N internal stress detection branch models to obtain N model accuracy rates and a prediction loss value set of the N groups of model sample data;
and the weight distribution updating unit is used for iteratively updating the weight distribution of the N groups of model sample data based on the N model accuracy rates and the prediction loss value sets, and acquiring the basic internal stress detection branch model set.
In one embodiment, the system further comprises:
the weight updating function construction unit is used for constructing a sample weight updating function, and the sample weight updating function specifically comprises the following steps:the method comprises the steps of carrying out a first treatment on the surface of the Wherein (1)>Updated weights indicating model sample data, +.>Indicating the weight of the ith sample of the Nth test branch model,/for the test branch model>Model accuracy for Nth detection branch model, < ->Predictive value for the nth test model for the ith sample, +.>A correct output value for the ith sample;
and the weight iterative updating unit is used for iteratively updating the weight distribution of the N groups of model sample data based on the calculation result of the weight updating function.
In one embodiment, the system further comprises:
the principal component analysis unit is used for obtaining the detection attribute information of the internal stress detection factor set, and carrying out principal component analysis on the detection attribute information to obtain dimension-reduction detection attribute information;
the factor analysis unit is used for carrying out factor analysis on the dimension reduction detection attribute information to obtain a detection factor weight distribution result;
the influence degree giving unit is used for giving influence degree to each detection factor in the stress detection factor set according to the detection factor weight distribution result, and obtaining a detection factor influence degree coefficient;
and the parameter fusion unit is used for carrying out parameter fusion on the N model accuracy rates by using the detection factor influence coefficient to obtain the model fusion coefficient set.
In one embodiment, the system further comprises:
the detection data acquisition unit is used for carrying out data acquisition on the wafer processing flow based on the internal stress detection factor set to obtain a wafer structured detection data stream and a wafer image detection data stream;
the defect characteristic obtaining unit is used for carrying out surface defect identification and characteristic marking on the wafer image detection data stream to obtain a wafer surface defect characteristic information set;
the processing detection data determining unit is used for determining the wafer processing detection data according to the wafer structural detection data flow and the wafer surface defect characteristic information set;
the database construction unit is used for acquiring the internal stress detection result data through an internal stress detector and constructing the wafer internal stress detection database based on the wafer processing detection data and the internal stress detection result data.
In one embodiment, the system further comprises:
the surface defect identification unit is used for carrying out surface defect identification on the wafer image detection data stream based on the wafer defect identifier to obtain a wafer surface defect information set;
the clustering marking unit is used for carrying out clustering marking on the wafer surface defect information set according to the wafer defect label set to obtain a surface defect clustering marking result;
the defect point region extraction unit is used for respectively extracting defect point regions from the wafer surface defect information set to obtain a surface defect characteristic extraction value set;
and the surface defect feature determining unit is used for determining the wafer surface defect feature information set based on the surface defect cluster marking result and the surface defect feature extraction value set.
For an embodiment of a system for detecting internal stress distribution of a wafer material, reference may be made to the above embodiment of a method for detecting internal stress distribution of a wafer material, which is not described herein. The above examples illustrate only a few embodiments of the application, which are described in detail and are not to be construed as limiting the scope of the application. It should be noted that it will be apparent to those skilled in the art that several variations and modifications can be made without departing from the spirit of the application, which are all within the scope of the application. Accordingly, the scope of protection of the present application is to be determined by the appended claims.
Claims (5)
1. A method for detecting internal stress distribution of a wafer material, the method comprising:
a processing node identification network is laid out and obtained, and a processing node detection data stream of the target wafer is acquired and obtained based on the processing node identification network;
obtaining an internal stress detection factor set according to the wafer material application standard, wherein the internal stress detection factor set comprises material characteristics, dimension thickness, surface defects and a processing technology;
constructing a wafer internal stress detection database, wherein the wafer internal stress detection database comprises wafer processing detection data and internal stress detection result data;
the construction of the wafer internal stress detection database comprises the following steps:
performing data acquisition on the wafer processing flow based on the internal stress detection factor set to obtain a wafer structured detection data stream and a wafer image detection data stream;
carrying out surface defect identification and feature marking on the wafer image detection data stream to obtain a wafer surface defect feature information set;
determining the wafer processing detection data according to the wafer structural detection data flow and the wafer surface defect characteristic information set;
acquiring the internal stress detection result data through an internal stress detector, and constructing the wafer internal stress detection database based on the wafer processing detection data and the internal stress detection result data;
classifying the wafer internal stress detection database according to the internal stress detection factor set to obtain a wafer detection factor sample data set;
training the wafer detection factor sample data sets by using a deep learning network structure to obtain a basic internal stress detection branch model set, wherein the training comprises the following steps:
randomly selecting N groups of model sample data from the wafer detection factor sample data set, and respectively performing equal distribution weight layer training on the N groups of model sample data to obtain N internal stress detection branch models;
model verification is carried out on the N internal stress detection branch models respectively, and N model accuracy rates and a prediction loss value set of the N groups of model sample data are obtained;
iteratively updating weight distribution of the N groups of model sample data based on the N model accuracy rates and the prediction loss value sets to obtain the basic internal stress detection branch model set;
acquiring a model fusion coefficient set, and fusing each branch model in the basic internal stress detection branch model set based on the model fusion coefficient set to generate a wafer internal stress self-adaptive detection model;
the obtaining the model fusion coefficient set includes:
obtaining detection attribute information of the internal stress detection factor set, and performing principal component analysis on the detection attribute information to obtain dimension reduction detection attribute information;
based on factor analysis on the dimension reduction detection attribute information, obtaining a detection factor weight distribution result;
according to the detection factor weight distribution result, giving influence to each detection factor in the stress detection factor set, and obtaining a detection factor influence coefficient;
carrying out parameter blending on the N model accuracy rates by using the detection factor influence coefficient to obtain the model fusion coefficient set;
and analyzing the processing node detection data stream based on the wafer internal stress self-adaptive detection model to obtain a wafer internal stress distribution detection result.
2. The method of claim 1, wherein the obtaining the processing node inspection data stream for the target wafer comprises:
setting a node data acquisition dual channel according to the processing node identification network, wherein the node data acquisition dual channel comprises a structured data identification channel and an image data identification channel;
respectively acquiring and acquiring a structured detection data stream and a multi-angle image detection data stream of the target wafer based on the structured data identification channel and the image data identification channel;
performing data cleaning processing on the structured detection data stream to obtain a standard structured detection data stream, and performing denoising pretreatment on the multi-angle image detection data stream to obtain a standard image detection data stream;
the processing node detection data stream is generated based on the standard structured detection data stream and the standard image detection data stream.
3. The method of claim 1, wherein the updating the weight distribution of the N sets of model sample data comprises:
a sample weight updating function is constructed, wherein the sample weight updating function specifically comprises the following steps:
；
wherein,updated weights indicating model sample data, +.>Indicating the weight of the ith sample of the Nth test branch model,/for the test branch model>Model accuracy for Nth detection branch model, < ->Predictive value for the nth test model for the ith sample, +.>A correct output value for the ith sample;
and iteratively updating the weight distribution of the N groups of model sample data based on the calculation result of the weight updating function.
4. The method of claim 1, wherein the obtaining a wafer surface defect characterization information set comprises:
carrying out surface defect identification on the wafer image detection data stream based on a wafer defect identifier to obtain a wafer surface defect information set;
clustering and marking the wafer surface defect information set according to the wafer defect label set to obtain a surface defect clustering and marking result;
respectively extracting defect point areas from the wafer surface defect information set to obtain a surface defect characteristic extraction value set;
and determining the wafer surface defect characteristic information set based on the surface defect clustering marking result and the surface defect characteristic extraction value set.
5. A system for detecting internal stress distribution of a wafer material, the system comprising:
the detection data stream acquisition module is used for arranging and acquiring a processing node identification network, and acquiring a processing node detection data stream of the target wafer based on the processing node identification network;
the detection factor set obtaining module is used for obtaining an internal stress detection factor set according to the application standard of the wafer material, wherein the internal stress detection factor set comprises material characteristics, dimension thickness, surface defects and a processing technology;
the wafer internal stress detection database comprises wafer processing detection data and internal stress detection result data;
the database classification module is used for classifying the wafer internal stress detection database according to the internal stress detection factor set to obtain a wafer detection factor sample data set;
the branch model training obtaining module is used for respectively training the wafer detection factor sample data sets by utilizing a deep learning network structure to obtain a basic internal stress detection branch model set;
the self-adaptive detection model generation module is used for acquiring a model fusion coefficient set, and fusing all branch models in the basic internal stress detection branch model set based on the model fusion coefficient set to generate a wafer internal stress self-adaptive detection model;
the detection result obtaining module is used for analyzing the processing node detection data stream based on the wafer internal stress self-adaptive detection model to obtain a wafer internal stress distribution detection result;
the system further comprises:
the branch model training unit is used for randomly selecting N groups of model sample data from the wafer detection factor sample data set, and respectively carrying out equal distribution weight layer training on the N groups of model sample data to obtain N internal stress detection branch models;
the branch model verification unit is used for respectively carrying out model verification on the N internal stress detection branch models to obtain N model accuracy rates and a prediction loss value set of the N groups of model sample data;
the weight distribution updating unit is used for iteratively updating the weight distribution of the N groups of model sample data based on the N model accuracy rates and the prediction loss value sets to obtain the basic internal stress detection branch model set;
the principal component analysis unit is used for obtaining the detection attribute information of the internal stress detection factor set, and carrying out principal component analysis on the detection attribute information to obtain dimension-reduction detection attribute information;
the factor analysis unit is used for carrying out factor analysis on the dimension reduction detection attribute information to obtain a detection factor weight distribution result;
the influence degree giving unit is used for giving influence degree to each detection factor in the stress detection factor set according to the detection factor weight distribution result, and obtaining a detection factor influence degree coefficient;
the parameter fusion unit is used for carrying out parameter fusion on the N model accuracy rates by using the detection factor influence coefficient to obtain the model fusion coefficient set;
the detection data acquisition unit is used for carrying out data acquisition on the wafer processing flow based on the internal stress detection factor set to obtain a wafer structured detection data stream and a wafer image detection data stream;
the defect characteristic obtaining unit is used for carrying out surface defect identification and characteristic marking on the wafer image detection data stream to obtain a wafer surface defect characteristic information set;
the processing detection data determining unit is used for determining the wafer processing detection data according to the wafer structural detection data flow and the wafer surface defect characteristic information set;
the database construction unit is used for acquiring the internal stress detection result data through an internal stress detector and constructing the wafer internal stress detection database based on the wafer processing detection data and the internal stress detection result data.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202311312913.2A CN117059511B (en) | 2023-10-11 | 2023-10-11 | Method and system for detecting internal stress distribution of wafer material |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202311312913.2A CN117059511B (en) | 2023-10-11 | 2023-10-11 | Method and system for detecting internal stress distribution of wafer material |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN117059511A CN117059511A (en) | 2023-11-14 |
| CN117059511B true CN117059511B (en) | 2023-12-15 |
Family
ID=88655784
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202311312913.2A Active CN117059511B (en) | 2023-10-11 | 2023-10-11 | Method and system for detecting internal stress distribution of wafer material |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN117059511B (en) |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN1858665A (en) * | 2005-05-03 | 2006-11-08 | 台湾积体电路制造股份有限公司 | Method and system for monitoring the processing tools in semiconductor production device |
| CN109977808A (en) * | 2019-03-11 | 2019-07-05 | 北京工业大学 | A kind of wafer surface defects mode detection and analysis method |
| CN113094893A (en) * | 2021-04-06 | 2021-07-09 | 江苏大学 | Wafer quality virtual measurement method and device, computer equipment and storage medium |
| CN113770555A (en) * | 2021-09-27 | 2021-12-10 | 北京大学 | Involute laser-assisted wafer cutting processing method and system and stress detection method |
| CN116168033A (en) * | 2023-04-25 | 2023-05-26 | 厦门福信光电集成有限公司 | Wafer lattice dislocation image detection method and system based on deep learning |
-
2023
- 2023-10-11 CN CN202311312913.2A patent/CN117059511B/en active Active
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN1858665A (en) * | 2005-05-03 | 2006-11-08 | 台湾积体电路制造股份有限公司 | Method and system for monitoring the processing tools in semiconductor production device |
| CN109977808A (en) * | 2019-03-11 | 2019-07-05 | 北京工业大学 | A kind of wafer surface defects mode detection and analysis method |
| CN113094893A (en) * | 2021-04-06 | 2021-07-09 | 江苏大学 | Wafer quality virtual measurement method and device, computer equipment and storage medium |
| CN113770555A (en) * | 2021-09-27 | 2021-12-10 | 北京大学 | Involute laser-assisted wafer cutting processing method and system and stress detection method |
| CN116168033A (en) * | 2023-04-25 | 2023-05-26 | 厦门福信光电集成有限公司 | Wafer lattice dislocation image detection method and system based on deep learning |
Also Published As
| Publication number | Publication date |
|---|---|
| CN117059511A (en) | 2023-11-14 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| JP5735499B2 (en) | Method and apparatus for automatically detecting and classifying fault conditions during substrate processing | |
| JP2020126601A (en) | Fault detecting method and fault detecting device | |
| CN109285791B (en) | Design layout-based rapid online defect diagnosis, classification and sampling method and system | |
| CN111652098B (en) | Product surface defect detection method and device | |
| JP2021190716A (en) | Detection of failure in semiconductor specimen using weak labeling | |
| CN112000081B (en) | Fault monitoring method and system based on multi-block information extraction and Mahalanobis distance | |
| CN112529109A (en) | Unsupervised multi-model-based anomaly detection method and system | |
| CN116113942A (en) | Predicting equipment failure modes based on process traces | |
| CN116861369A (en) | Air compressor health assessment method and system for industrial internet heterogeneous data fusion | |
| CN116308963A (en) | Government affair data analysis method and system | |
| CN117059511B (en) | Method and system for detecting internal stress distribution of wafer material | |
| CN116627116B (en) | Process industry fault positioning method and system and electronic equipment | |
| CN112418353A (en) | Neural network training method for battery diaphragm abnormity detection | |
| TWI400619B (en) | Product quality fault detection method and real metrolgy data evaluation method | |
| CN115063385A (en) | Machine vision method for wafer detection | |
| TWI801973B (en) | Device and method for estimating error factors | |
| CN107067034B (en) | Method and system for rapidly identifying infrared spectrum data classification | |
| JP5566616B2 (en) | Characteristic analysis apparatus, program for causing computer to execute characteristic analysis method, and computer-readable recording medium recording the program | |
| US8000905B1 (en) | Computer-implemented methods, carrier media, and systems for determining sizes of defects detected on a wafer | |
| JP4538205B2 (en) | Inspection data analysis program, inspection data analyzer | |
| CN117235650B (en) | Method, device, equipment and medium for detecting high-altitude operation state | |
| CN111369508A (en) | Defect detection method and system for metal three-dimensional lattice structure | |
| CN111724352B (en) | Patch LED flaw labeling method based on kernel density estimation | |
| US20220392187A1 (en) | Image recognition system | |
| CN116020756A (en) | Fingerprint chip detection and good product identification 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 |