Semiconductor wafer defect automatic detection method based on deep learning

Abstract

The invention provides a semiconductor wafer defect automatic detection method based on deep learning, which relates to the technical field of semiconductor detection, and comprises the following steps: interacting quality monitoring units of a target wafer production line, extracting a plurality of wafer defect detection data, carrying out defect feature identification on the plurality of wafer defect detection data, and obtaining K defect feature clusters; and comparing the K defect feature edge intervals with the target wafer quality feature tolerance interval, configuring an automatic detection wheel disc based on the K defect factors to obtain an automatic detection scheme, and carrying out product defect automatic detection on the target wafer production line according to the automatic detection scheme to obtain a defect detection result. The technical problems that in the prior art, semiconductor defect detection efficiency is low, missing detection and false detection are prone to occurring, wafer detection accuracy is not high, and real-time performance is insufficient are solved. The technical effects of automatically and dynamically detecting the wafer defects in real time and improving the wafer defect detection efficiency, accuracy and reliability are achieved.

Description

Technical Field

[0001] The present invention relates to the field of semiconductor detection technology, and in particular to a method for automatic detection of semiconductor wafer defects based on deep learning. Background Art

[0002] In the semiconductor manufacturing industry, wafer quality control is a key link to ensure the quality of the final product. With the continuous advancement of integrated circuit technology and the continuous increase in wafer size, defect detection on wafers has become more and more complicated. Traditional wafer defect detection methods often rely on manual visual inspection or inspection by optical, electron beam and other methods after manufacturing. Although these methods are simple and easy to use, they have problems such as low efficiency, poor reliability, easy missed detection and false detection.

[0003] The existing technology has technical problems such as low efficiency of semiconductor defect detection, easy missed detection and false detection, resulting in low wafer detection accuracy and insufficient real-time performance. Summary of the invention

[0004] The present application provides a method for automatic detection of semiconductor wafer defects based on deep learning, which is used to solve the technical problems in the prior art that semiconductor defect detection is inefficient, easy to miss and misdetect, resulting in low wafer detection accuracy and insufficient real-time performance.

[0005] In view of the above problems, the present application provides a method for automatic detection of semiconductor wafer defects based on deep learning, the method comprising: interacting with the quality monitoring unit of the target wafer production line, extracting multiple wafer defect detection data within a preset detection window, wherein each wafer defect detection data includes a production time mark; performing defect feature identification on the multiple wafer defect detection data to obtain multiple defect feature sets; clustering the multiple defect feature sets with the defect feature type as an index to obtain K defect feature clusters, each defect feature cluster including multiple defect feature values; traversing the K defect feature clusters to perform defect feature value edge identification to generate K defect feature edge intervals; comparing the K defect feature edge intervals with the target wafer quality feature tolerance interval to determine K defect factors; configuring an automatic detection wheel based on the K defect factors, and turning the automatic detection wheel multiple times according to the preset number of detection samples to obtain an automatic detection scheme, wherein the automatic detection scheme includes K detection sample quantities of K defect features; performing automatic detection of product defects of the target wafer production line based on the automatic detection scheme to obtain defect detection results.

[0006] One or more technical solutions provided in this application have at least the following technical effects or advantages: The method provided in the embodiment of the present application extracts multiple wafer defect detection data within a preset detection window through the quality monitoring unit of the interactive target wafer production line, wherein each wafer defect detection data includes a production time mark; the defect feature recognition is performed on the multiple wafer defect detection data to obtain multiple defect feature sets; the multiple defect feature sets are clustered with the defect feature type as the index to obtain K defect feature clusters, each defect feature cluster includes multiple defect feature values; the K defect feature clusters are traversed to perform defect feature value edge recognition to generate K defect feature edge intervals; the K defect feature edge intervals are compared with the target wafer quality feature tolerance interval to determine K defect factors; the automatic detection wheel is configured based on the K defect factors, and the automatic detection wheel is dialed multiple times according to the preset detection sample number to obtain an automatic detection scheme, wherein the automatic detection scheme includes K detection sample quantities of K defect features; the product defects of the target wafer production line are automatically detected based on the automatic detection scheme to obtain defect detection results. The technical effect of automatically and dynamically detecting wafer defects in real time and improving the efficiency, accuracy and reliability of wafer defect detection is achieved. BRIEF DESCRIPTION OF THE DRAWINGS

[0007] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0008] Figure 1 A schematic diagram of the process flow of the semiconductor wafer defect automatic detection method based on deep learning provided in this application; Figure 2 A schematic diagram of a process for generating K defect feature edge intervals in a semiconductor wafer defect automatic detection method based on deep learning provided in the present application; Figure 3 A schematic diagram of the process of configuring an automatic detection wheel in the deep learning-based automatic detection method for semiconductor wafer defects provided in this application. DETAILED DESCRIPTION

[0009] This application provides a method for automatic detection of semiconductor wafer defects based on deep learning, which is used to solve the technical problems in the prior art that semiconductor defect detection is inefficient, easy to miss and misdetect, resulting in low wafer detection accuracy and insufficient real-time performance. Through intelligent data analysis of deep learning, automatic real-time dynamic detection of wafer defects is achieved, and the technical effect of improving the efficiency, accuracy and reliability of wafer defect detection is achieved.

[0010] Below, the technical solutions in the present invention will be clearly and completely described with reference to the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, rather than all of the embodiments of the present invention. It should be understood that the present invention is not limited to the example embodiments described herein. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention. It should also be noted that, for the convenience of description, only the parts related to the present invention are shown in the accompanying drawings, rather than all of them.

[0011] like Figure 1 As shown, the present application provides a method for automatic detection of semiconductor wafer defects based on deep learning, the method comprising: The quality monitoring unit of the interactive target wafer production line extracts multiple wafer defect detection data within a preset detection window, wherein each wafer defect detection data includes a production time mark.

[0012] Specifically, the quality monitoring unit is a key component of the wafer production line, which is used to monitor the wafer production process in real time and collect relevant quality data. The monitoring data includes images of the wafer surface during the generation process, dimension measurements, electrical performance test results, etc. The preset detection window is a time range or condition range pre-set based on production batches, production time, product types or other relevant factors, which is used to limit the range of wafer defect detection data extracted from the quality monitoring unit. The quality monitoring unit of the interactive target wafer production line extracts multiple wafer defect detection data within the preset detection window, which means extracting multiple wafer defect detection data within the preset detection window range from the quality monitoring unit of the target wafer production line. The wafer defect detection data includes wafer images, defect locations, defect types, and defect sizes, and each wafer defect detection data contains a production time stamp to record when the data was generated, providing strong support for subsequent quality analysis and automatic detection.

[0013] Defect feature recognition is performed on the plurality of wafer defect detection data to obtain a plurality of defect feature sets.

[0014] Specifically, a neural network model is used to identify defect features of multiple extracted wafer defect detection data to obtain multiple defect feature sets, wherein the neural network model for defect feature identification is a model that has been trained by including multiple normal and multiple historical wafer detection data containing various types of defects. The trained neural network model can achieve efficient and accurate defect feature identification. Each defect feature set includes defect feature type, geometric features, texture features, etc., providing data support for subsequent quality assessment and automatic detection.

[0015] The plurality of defect feature sets are clustered with the defect feature type as an index to obtain K defect feature clusters, each defect feature cluster including a plurality of defect feature values.

[0016] Specifically, clustering algorithms such as K-means clustering and hierarchical clustering are used to take defect feature types as indexes and cluster multiple defect feature sets, that is, defect feature values ​​with similar feature values ​​and the same feature type are clustered together according to the similarity between the defect features, and different defect feature types are assigned to different clusters to form K defect feature clusters, where K is a positive integer ≥ 1, and each cluster contains multiple defect feature values ​​with similarity and belonging to the same type of defect features. Cluster analysis can better understand the type characteristics of defects and provide guidance for subsequent automatic detection of defects.

[0017] The K defect feature clusters are traversed to perform defect feature value edge recognition to generate K defect feature edge intervals.

[0018] Specifically, the characteristic values ​​in each of the K defect characteristic clusters are analyzed to determine the boundaries of their characteristic values. The boundaries can be the natural boundaries of the characteristic values ​​within the cluster, such as the minimum and maximum values, or artificially set thresholds to distinguish samples within and outside the cluster. Then, based on the edge recognition method, K defect characteristic edge intervals are generated for each defect characteristic cluster according to the identified boundaries. The defect characteristic edge intervals are numerical intervals or probability distributions. Obtaining defect characteristic edge intervals through identification helps to achieve more accurate defect detection and classification in the wafer production process.

[0019] The K defect feature edge intervals are compared with the target wafer quality feature tolerance interval to determine K defect factors.

[0020] Specifically, according to the application requirements, historical data and industry standards of the target wafer, the target wafer quality feature tolerance interval is set, and the target wafer quality feature tolerance interval refers to the acceptable range of the target wafer in various features, and the tolerance interval can be represented by a numerical interval. Then, each defect feature cluster edge interval of the K defect feature edge intervals is compared with the target wafer quality feature tolerance interval to check whether the defect feature edge interval is completely within the target wafer quality feature tolerance interval, and K defect factors are obtained according to the comparison results. The defect factor is used to reflect the severity of the defect or the potential impact on the wafer quality. For example, when the defect feature edge interval is completely within the target wafer quality feature tolerance interval, the defect factor can be set to 0, indicating that the defect feature of the cluster is completely within the acceptable range. If the defect feature edge interval overlaps with the tolerance interval, but partially exceeds the tolerance interval, the defect factor is calculated according to the degree of excess. For example, the proportion of the excess portion to the entire interval can be used as the defect factor, indicating that the defect has a certain impact on the wafer quality. If the defect feature edge interval is completely outside the tolerance interval, the defect factor should be set to a larger value, indicating that the defect feature of the cluster has a serious impact on the wafer quality.

[0021] An automatic detection wheel is configured based on the K defect factors, and the automatic detection wheel is turned multiple times according to a preset number of detection samples to obtain an automatic detection plan, wherein the automatic detection plan includes K detection sample quantities of K defect features.

[0022] Specifically, the size or severity of each defect factor of the K defect factors is analyzed, and a detection priority is assigned to each defect feature. The higher the defect factor, the higher the detection priority. Then, according to the detection priority, each defect factor is mapped to the corresponding area of ​​the automatic detection wheel. The larger the defect factor, the larger the area allocated in the wheel. When the corresponding subsequent fluctuation wheel is performed, the greater the probability of being selected for the detection, and the configuration of the automatic detection wheel is realized. Each area of ​​the automatic detection wheel represents a specific defect feature, which is used to guide the detection order and frequency of multiple defect features in the automatic detection of wafer defects. According to the preset number of detection samples, that is, the number of samples that need to be detected in the automatic detection process for each defect feature determined according to the defect history data, defect severity and wafer detection goals, the automatic detection wheel is run multiple times to obtain an automatic detection scheme with K detection sample quantities containing K defect features, which helps to achieve efficient and accurate detection of key defects in the wafer production process.

[0023] Based on the automatic detection scheme, automatic detection of product defects of the target wafer production line is performed to obtain defect detection results.

[0024] Specifically, the target wafer production line products are automatically inspected according to the defect features and corresponding inspection sample quantities in the automatic inspection scheme to identify the defect features of the target wafer production line products. According to the number of K inspection samples in the automatic inspection scheme, the wafer products are continuously and highly accurately inspected, including steps such as image capture, feature extraction, and defect identification. The inspection results of each defect feature are recorded in real time to obtain defect inspection results, including information such as defect type, quantity, location, and severity. This not only improves the inspection efficiency and ensures the accuracy and reliability of the inspection results, but also realizes dynamic real-time inspection of wafer defects, thereby improving the wafer quality of the target wafer production line.

[0025] In one embodiment, Figure 2 As shown, traversing the K defect feature clusters to perform defect feature value edge recognition and generate K defect feature edge intervals, including: extracting a first defect feature cluster from the K defect feature clusters, wherein the first defect feature cluster includes multiple first defect feature values; randomly extracting M first defect feature values ​​from the multiple first defect feature values ​​as M transition particles, wherein the M transition particles have M transition densities; the M transition particles move in the first defect feature cluster according to a preset step length to obtain M stage transition particles, wherein the M stage transition particles have M stage transition densities; iteratively updating the M transition particles according to the M stage transition particles to obtain M updated transition particles; determining a first defect feature edge interval of the first defect feature cluster based on the M updated transition particles; performing defect feature value edge recognition on the remaining K-1 defect feature clusters to generate K-1 defect feature edge intervals; generating the K defect feature edge intervals according to the first defect feature edge interval and the K-1 defect feature edge intervals.

[0026] Specifically, first, a defect feature cluster is extracted from K defect feature clusters as the first defect feature cluster, and the first defect feature cluster contains multiple first defect feature values. Then, M values ​​are randomly selected from multiple first defect feature values ​​as M transition particles, each transition particle represents a defect feature value, M is a positive integer ≥ 1, and the transition density is the ratio of the number of feature values ​​in the area constructed with the transition particle as the center and the preset step length as the radius to the interval length. The interval length is two preset step lengths. The transition density reflects the distribution density of the feature values ​​gathered around the transition particles. M transition particles have M transition densities. Then, the M transition particles are moved within the range of the first defect feature cluster according to the preset step length. The preset step length determines the distance that the particle moves each time in the feature space, which can be set according to actual needs. During the movement, each transition particle will be adjusted according to the distribution of the characteristic values ​​at its position, and M stage transition particles will be formed according to the distribution of the characteristic values ​​at its adjusted position, and their stage transition density will be calculated according to the new position of the stage transition particle. The stage transition density reflects the degree of aggregation of particles in the characteristic space, and each stage transition particle has a corresponding stage transition density. Afterwards, the initial M transition particles are iteratively updated based on the density information of the stage transition particles. The iterative update includes adjusting the position, speed or density of the particles. Through the iterative update, the transition particles can more accurately reflect the distribution of the characteristic values ​​in a specific interval, thereby obtaining M updated transition particles. Then, by analyzing the density changes, clustering boundaries or other statistical methods of the M updated transition particles, the defect feature edge interval of the first defect feature cluster is identified and determined. The defect feature edge interval is used to define the abnormal range of the defect feature value in the cluster, that is, it represents the left and right intervals of the defect feature value corresponding to the defective wafer. Repeat the above steps to perform defect feature value edge recognition on the remaining K-1 defect feature clusters to generate corresponding K-1 defect feature edge intervals. Finally, the defect feature edge interval of the first defect feature cluster and the remaining K-1 defect feature edge intervals are merged to form a complete K defect feature edge intervals, which can realize comprehensive defect feature value edge recognition for the K defect feature clusters of the wafer and generate accurate defect feature edge intervals, providing an important reference basis for subsequent product defect detection of the target wafer production line, ensuring that the detection process can accurately identify various defects and improve the accuracy and reliability of detection.

[0027] Furthermore, the M transition particles are iteratively updated according to the M stage transition particles to obtain M updated transition particles, including: comparing the M transition densities with the M stage transition densities, and when the M stage transition densities are greater than the M transition densities, updating the M transition particles using the M stage transition particles to obtain M stage updated transition particles; and taking the M stage updated transition particles as the starting point, moving in the first defect feature cluster according to the M first transition directions according to a preset step size, and iteratively updating according to the movement results until a preset number of iterations is met to obtain the M updated transition particles.

[0028] Specifically, after obtaining M phase transition particles and their corresponding M phase transition densities, compare the transition densities of the M phase transition particles with the phase transition densities of the M phase transition particles formed after they are moved. If the phase transition density of a phase transition particle is greater than the transition density of its corresponding initial transition particle, it means that the position of the phase transition particle may be closer to or located in the high-density area of ​​the data. In this case, use M phase transition particles to update the M original transition particles to form M phase update transition particles, that is, the position of the particle is transferred from the original position to the new position of the phase transition particle. Then, take the M phase update transition particles as the new starting point, and move in the first defect feature cluster again according to the preset step length. According to the result of this movement, iteratively update the particles again, so that the transition particles are closer to the real characteristic value distribution. Repeat this step until the preset number of iterations is met. When the preset maximum number of iterations is reached, M updated transition particles after multiple iterations are obtained, which provides an important basis for the determination of the subsequent defect feature edge interval, thereby improving the accuracy and reliability of subsequent wafer defect detection.

[0029] Furthermore, the M transition particles are iteratively updated according to the M stage transition particles to obtain M updated transition particles, and it also includes: when the M stage transition densities are less than or equal to the M transition densities, M first transition directions are added to a taboo table, wherein the taboo table has a taboo iteration number; again taking the M transition particles as the starting point, moving in the first defect feature cluster according to a preset step size in a direction other than the transition direction stored in the taboo table, and iteratively updating according to the movement result until the preset iteration parameter is met to obtain the M updated transition particles.

[0030] Specifically, when the comparison of the M transition densities and the M stage transition densities shows that the M stage transition densities are less than or equal to the M transition densities, it indicates that the position of the stage transition particle is not better than the position of the original transition particle, or no better solution has been found. At this time, the M first transition directions are added to the taboo table. The taboo table is one of the core components of the taboo search algorithm, which is used to record solutions or search directions that should be avoided during the search process. The taboo table can be a list or an array, which stores the transition directions that have been marked as taboo, and the taboo table has a taboo iteration number. For example, if the taboo iteration number is set to 3 times, the corresponding direction cannot be used within 3 times during the iteration process to avoid falling into an inferior solution.

[0031] Next, the M transition particles are used as the starting point again, and they are moved in the first defect feature cluster according to the preset step length. However, this movement will exclude the transition direction stored in the taboo table, that is, only move in the non-taboo direction according to the preset step length, and avoid repeated selection of known bad directions during the search process, thereby improving the search efficiency and the quality of the solution. In each iteration, their eigenvalue distribution information is updated according to the movement results of the current transition particles. When the transition density of a certain stage is greater than the corresponding transition density, the transition particles of this stage are used to update the corresponding transition particles, and the iterative update is continued. Each iteration checks whether the direction in the taboo table is used until the preset iteration parameters are met. Finally, when the iterative process ends, the M updated transition particles obtained represent the optimized exploration results after considering the taboo search strategy, avoiding repeated entry into the search area that previously led to unsatisfactory results, thereby increasing the possibility of finding the global optimal solution, helping to improve the accuracy of defect feature value edge recognition, and thus improving the accuracy and reliability of wafer defect detection.

[0032] Furthermore, the first defect characteristic edge interval of the first defect characteristic cluster is determined based on the M updated transition particles, including: taking the updated transition particle corresponding to the maximum value of the transition density among the M updated transition particles as the target transition particle; constructing a first partition interval with the target transition particle as the midpoint of the interval and the preset step size as the interval radius; taking the maximum value of multiple first defect characteristic values ​​in the first defect characteristic cluster located in the first partition interval as the first right endpoint; taking the minimum value of multiple first defect characteristic values ​​in the first defect characteristic cluster located in the first partition interval as the first left endpoint; and constructing the first defect characteristic edge interval based on the first left endpoint and the first right endpoint.

[0033] Specifically, from the M updated transition particles, the updated transition particle corresponding to the maximum transition density is selected as the target transition particle. The target transition particle refers to the position that is most likely to contain the defect feature edge in the current search process. Then, the position of the target transition particle is used as the midpoint of the interval, and the preset step size is used as the radius of the interval. The first partition interval is constructed around the midpoint. The first partition interval is a neighborhood centered on the target transition particle, and its size is determined by the step size. In the first defect feature cluster, all the first defect feature values ​​located in the first partition interval are screened out, and the maximum value is selected from the screened feature values, which is used as the first right endpoint, and the minimum value is selected from the screened feature values, which is used as the first left endpoint. Finally, using the determined first left endpoint and the first right endpoint, a closed interval is constructed, namely the first defect feature edge interval. The first defect feature edge interval contains the feature value range that may represent the defect feature edge within the current exploration range. By effectively utilizing the information provided by the updated transition particles, a relatively accurate defect feature edge interval is determined in the first defect feature cluster, which can better identify defect features and improve defect detection efficiency and accuracy.

[0034] In one embodiment, Figure 3 As shown, configuring an automatic detection wheel based on the K defect factors includes: collecting a preset wheel area; respectively calculating the ratios of the K defect factors to the sum of the K defect factors, and using the calculation results as K wheel coefficients; using the K wheel coefficients as weight ratios, dividing the preset wheel area according to the weight ratios, and marking K defect features in corresponding divided areas of the wheel to generate the automatic detection wheel.

[0035] Specifically, the total area of ​​the preset roulette wheel is determined according to actual needs. Then, for K known defect factors, the ratio of each defect factor to the sum of the K defect factors is calculated respectively. This ratio reflects the importance or influence weight of the defect factor among all factors. And these ratio results are used as K roulette wheel coefficients. The size of the roulette wheel coefficient represents the relative importance of the corresponding defect factor, and each roulette wheel coefficient corresponds to a defect factor. Use the calculated K roulette wheel coefficients as the weight ratio to divide the preset roulette wheel area. Each defect factor is assigned to a specific area on the roulette wheel according to its roulette wheel coefficient, and the area of ​​this area is proportional to the weight of the factor. Then, the K defect features are marked on the corresponding divided areas of the roulette wheel, and the defect feature mark is used to indicate the corresponding defect factor in each area. After completing the marking of all defect features, an automatic detection roulette wheel is generated. By configuring the automatic detection roulette wheel, the weight and importance of each defect factor in the overall defect detection can be intuitively displayed, thereby improving the efficiency and accuracy of defect detection.

[0036] Furthermore, the automatic detection wheel is turned multiple times according to a preset number of detection samples to obtain an automatic detection plan, including: turning the automatic detection wheel multiple times according to a preset number of detection samples to obtain an initial automatic detection plan; extracting K initial detection sample quantities of K defect features in the initial automatic detection plan, and calculating K random inspection coefficients based on the K initial detection sample quantities; calculating the random inspection similarity between the K wheel coefficients and the K random inspection coefficients, and when the random inspection similarity meets a preset similarity threshold, using the initial automatic detection plan as the automatic detection plan.

[0037] Specifically, the automatic detection wheel is randomly turned multiple times according to the preset number of detection samples. After turning the wheel multiple times, the sample size of the defect features included in each detection is recorded according to the defect features marked on the wheel and the corresponding divided areas, forming an initial automatic detection plan.

[0038] From the initial automatic detection scheme, K initial detection sample sizes of K defect features are extracted, and the sample size represents the detection frequency or quantity of each defect feature in the initial detection scheme. Based on the K initial detection sample sizes, the sampling coefficient of each defect feature is calculated, and the sampling coefficient is the ratio of the detection sample size to the total sample size, or other indicators reflecting the detection frequency. Then, the K roulette coefficients, that is, the weight ratio, are compared with the K sampling coefficients to calculate the sampling similarity between them. The sampling similarity can be obtained by correlation analysis or calculating the similarity metric between two coefficient vectors, such as cosine similarity. A similarity threshold is preset to determine whether the sampling similarity is high enough. If the sampling similarity meets or exceeds this threshold, it means that the sample size distribution in the initial automatic detection scheme is relatively consistent with the weight distribution on the roulette, that is, the initial automatic detection scheme is reasonable, and it is used as the final automatic detection scheme for the subsequent automatic detection process. In this way, it can be ensured that the automatic detection scheme matches the roulette coefficient, and the accuracy and effectiveness of automatic detection can be improved.

[0039] Furthermore, the automatic detection roulette wheel is turned multiple times according to a preset number of detection samples to obtain an automatic detection scheme, and the method also includes: when the sampling similarity does not meet the preset similarity threshold, the automatic detection roulette wheel is turned multiple times according to the preset number of detection samples to obtain an iterative automatic detection scheme; the iterative sampling similarity of K iterative sampling coefficients and K roulette wheel coefficients in the iterative automatic detection scheme is calculated, and if the iterative sampling similarity meets the preset similarity threshold, the iterative automatic detection scheme is used as the automatic detection scheme.

[0040] Specifically, when the sampling similarity does not meet the preset similarity threshold, it means that the sample size distribution in the initial automatic detection scheme is quite different from the weight distribution on the roulette wheel, and it is necessary to dial the automatic detection roulette wheel multiple times according to the preset number of detection samples. Based on the results of multiple dialing of the roulette wheel, the detection data is collected and sorted to form an iterative automatic detection scheme, which contains new detection sample sizes for K defect features. Extract K iterative detection sample sizes of K defect features in the iterative automatic detection scheme, and calculate the iterative sampling coefficient of each defect feature based on these iterative detection sample sizes. Compare the K roulette coefficients, that is, the weight ratio, with the K iterative sampling coefficients to calculate the iterative sampling similarity between them, which can also be obtained by correlation analysis or calculating the similarity measure between two coefficient vectors, such as cosine similarity and other methods to measure similarity. Check whether the iterative sampling similarity meets the preset similarity threshold. If it does, it means that the iterative automatic detection scheme matches the roulette coefficient, and the iterative automatic detection scheme is used as the automatic detection scheme. If the iterative sampling similarity still does not meet the preset similarity threshold, you can consider continuing the iteration, or adjusting the number of test samples, roulette coefficient and other parameters, and then repeat the above steps until an automatic detection solution that meets the requirements is found. Through this iterative method, the automatic detection solution can be gradually optimized to ensure that it matches the roulette coefficient, thereby improving the accuracy and reliability of automatic detection of semiconductor wafer defects.

[0041] The above description is only a preferred embodiment of the present application and is not intended to limit the present application. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present application should be included in the protection scope of the present application.

[0042] This specification and the drawings are merely exemplary illustrations of the present application and are deemed to cover any and all modifications, variations, combinations or equivalents within the scope of the present application. Obviously, a person skilled in the art may make various modifications and variations to the present application without departing from the scope of the present application. Thus, if these modifications and variations of the present application fall within the scope of the present application and its equivalents, the present application intends to include these modifications and variations.

Claims

1. A semiconductor wafer defect automatic detection method based on deep learning, characterized in that: The method comprises: Interacting with a quality monitoring unit of a target wafer production line to extract a plurality of wafer defect detection data within a preset detection window, wherein each wafer defect detection data includes a production time mark; Performing defect feature recognition on the plurality of wafer defect detection data to obtain a plurality of defect feature sets; Clustering the multiple defect feature sets using the defect feature type as an index to obtain K defect feature clusters, each defect feature cluster including multiple defect feature values; Traversing the K defect feature clusters to perform defect feature value edge recognition and generate K defect feature edge intervals; Comparing the K defect feature edge intervals with the target wafer quality feature tolerance interval to determine K defect factors; Based on the K defect factors, an automatic detection wheel is configured, and the automatic detection wheel is dialed multiple times according to a preset number of detection samples to obtain an automatic detection scheme, wherein the automatic detection scheme includes K detection sample quantities of K defect features; Based on the automatic detection scheme, automatic detection of product defects of the target wafer production line is performed to obtain defect detection results.

2. The method for automatic semiconductor wafer defect detection based on deep learning as claimed in claim 1, characterized in that: Traversing the K defect feature clusters to perform defect feature value edge recognition, generating K defect feature edge intervals, including: Extracting a first defect feature cluster from the K defect feature clusters, wherein the first defect feature cluster includes a plurality of first defect feature values; Randomly extracting M first defect characteristic values ​​from the plurality of first defect characteristic values ​​as M transition particles, wherein the M transition particles have M transition densities; The M transition particles move in the first defect feature cluster according to a preset step length to obtain M stage transition particles, wherein the M stage transition particles have M stage transition densities; Iteratively updating the M transition particles according to the M stage transition particles to obtain M updated transition particles; Determine a first defect feature edge interval of the first defect feature cluster based on the M updated transition particles; Perform defect feature value edge recognition on the remaining K-1 defect feature clusters to generate K-1 defect feature edge intervals; The K defect feature edge intervals are generated according to the first defect feature edge interval and the K-1 defect feature edge intervals.

3. The method for automatic semiconductor wafer defect detection based on deep learning as claimed in claim 2, characterized in that: Iteratively updating the M transition particles according to the M stage transition particles to obtain M updated transition particles, including: Comparing the M transition densities with the M stage transition densities, when the M stage transition densities are greater than the M transition densities, updating the M transition particles using the M stage transition particles to obtain M stage updated transition particles; And taking the M stage updated transition particles as the starting point, moving in the first defect feature cluster according to the M first transition directions according to the preset step length, iteratively updating according to the movement result until the preset number of iterations is met, and obtaining the M updated transition particles.

4. The method for automatic semiconductor wafer defect detection based on deep learning as claimed in claim 3, characterized in that: include: When the M stage transition densities are less than or equal to the M transition densities, adding the M first transition directions into a taboo table, wherein the taboo table has a taboo iteration number; Taking the M transition particles as the starting point again, move in the first defect feature cluster in a direction other than the transition direction stored in the taboo table according to a preset step size, and perform iterative update according to the movement result until the preset iteration parameters are met to obtain the M updated transition particles.

5. The method for automatic semiconductor wafer defect detection based on deep learning as claimed in claim 3, characterized in that: Determining a first defect feature edge interval of the first defect feature cluster based on the M updated transition particles includes: Taking the updated transition particle corresponding to the maximum transition density among the M updated transition particles as the target transition particle; Taking the target transition particle as the midpoint of the interval and the preset step length as the interval radius, constructing a first partition interval; Taking the maximum value of a plurality of first defect feature values ​​in the first defect feature cluster and located in the first divided interval as the first right endpoint; Taking the minimum value of a plurality of first defect feature values ​​in the first defect feature cluster located in the first divided interval as the first left endpoint; The first defect feature edge interval is constructed according to the first left endpoint and the first right endpoint.

6. The method for automatic semiconductor wafer defect detection based on deep learning as claimed in claim 1, characterized in that: Configuring the automatic detection roulette based on the K defect factors includes: Collect the preset roulette area; Calculate the ratios of the K defect factors to the sum of the K defect factors respectively, and use the calculation results as K roulette coefficients; The K wheel coefficients are used as weight ratios, the preset wheel area is divided according to the weight ratios, and K defect features are marked in corresponding divided areas of the wheel to generate the automatic detection wheel.

7. The method for automatic semiconductor wafer defect detection based on deep learning as claimed in claim 6, characterized in that: Turn the automatic detection wheel multiple times according to the preset number of test samples to obtain the automatic detection plan, including: Turn the automatic detection wheel multiple times according to the preset number of detection samples to obtain an initial automatic detection plan; Extracting K initial detection sample sizes of K defect features in the initial automatic detection scheme, and calculating K sampling coefficients based on the K initial detection sample sizes; The sampling similarity between the K roulette coefficients and the K sampling coefficients is calculated, and when the sampling similarity meets a preset similarity threshold, the initial automatic detection scheme is used as the automatic detection scheme.

8. The method for automatic semiconductor wafer defect detection based on deep learning as claimed in claim 7, characterized in that: include: When the sampling similarity does not meet the preset similarity threshold, the automatic detection wheel is turned multiple times according to the preset number of detection samples to obtain an iterative automatic detection solution; The iterative sampling similarity of the K iterative sampling coefficients and the K roulette wheel coefficients in the iterative automatic detection scheme is calculated. If the iterative sampling similarity meets a preset similarity threshold, the iterative automatic detection scheme is used as the automatic detection scheme.

Images