Abnormal test structure acquisition method, verification method and related device

Abstract

Embodiments of the present application provide an abnormal test structure acquisition method, a verification method and related devices, wherein the abnormal test structure acquisition method comprises: acquiring a test data set; acquiring single-parameter abnormal data in which a driving current does not satisfy a corresponding driving current threshold or a leakage current does not satisfy a corresponding leakage current threshold, to obtain a single-parameter abnormal test structure set and a single-parameter normal test structure set; constructing a two-parameter feature according to the driving current and the leakage current of each single-parameter normal test structure in the single-parameter normal test structure set, acquiring a two-parameter feature threshold; acquiring two-parameter abnormal data in which the two-parameter feature does not satisfy the two-parameter feature threshold, acquiring each two-parameter abnormal test structure corresponding to the two-parameter abnormal data, to obtain a two-parameter abnormal test structure set; and obtaining an abnormal test structure set according to the single-parameter abnormal test structure set and the two-parameter abnormal test structure set. The method provided in the embodiments of the present application can improve the accuracy of acquiring abnormal test structures.

Description

Technical Field

[0001] This application relates to the field of semiconductor technology, and in particular to a method for obtaining abnormal test structures, a verification method, and related apparatus. Background Technology

[0002] Due to factors such as manufacturing processes, defective chips are produced during chip manufacturing. To improve chip manufacturing yield and reduce manufacturing costs, chips are tested at different stages. The test results are used to understand the impact of the manufacturing process on chip yield, allowing for adjustments to the process in subsequent batches to improve the yield of chips produced later.

[0003] After wafer fabrication is completed, different batches of wafers undergo WAT (Wafer Acceptance Test) to measure the electrical parameters of specific test structures, thereby obtaining WAT test data. By filtering the test data of parameters in the WAT test data that are correlated with the performance loss items of CP (Chip Performance Test) yield, abnormal test data of WAT test can be obtained. Then, the specific abnormal test structure corresponding to the abnormal test data (referred to as abnormal test structure in this article for convenience) can be determined, and process adjustments can be guided.

[0004] However, existing methods for obtaining abnormal test structures have low accuracy.

[0005] Therefore, improving the accuracy of obtaining abnormal test structures has become an urgent technical problem to be solved. Summary of the Invention

[0006] The technical problem addressed by the embodiments of this application is how to improve the accuracy of obtaining abnormal test structures.

[0007] To address the aforementioned problems, in a first aspect, embodiments of this application provide a method for obtaining anomaly test structures, comprising:

[0008] Obtain a test dataset, which includes the drive current and leakage current of WAT wafer acceptability tests for each specific test structure in a specific test structure set;

[0009] Obtain single-parameter abnormal data where the drive current does not meet the corresponding drive current threshold or the leakage current does not meet the corresponding leakage current threshold, and obtain the single-parameter abnormal test structure set and single-parameter normal test structure set corresponding to the single-parameter abnormal data.

[0010] Based on the driving current and leakage current of each single-parameter normal test structure in the single-parameter normal test structure set, construct a two-parameter feature and obtain the two-parameter feature threshold.

[0011] Obtain biparametric abnormal data where the biparametric features do not meet the biparametric feature threshold, and obtain each biparametric abnormal test structure corresponding to the biparametric abnormal data to obtain a biparametric abnormal test structure set;

[0012] An anomaly test structure set is obtained based on the single-parameter anomaly test structure set and the two-parameter anomaly test structure set.

[0013] Secondly, embodiments of this application provide a verification method for an anomaly test structure acquisition method, comprising:

[0014] By using the abnormal test structure acquisition method described in any of the preceding items, obtain the two-parameter verification abnormal test structure set and the verification abnormal test structure set corresponding to the verification dataset in the specific test structure set;

[0015] Obtain the normal test structure set based on the specific test structure set and the abnormal test structure set;

[0016] Obtain the proportions of device speed-related yield loss terms and defect-related yield loss terms for wafer testing corresponding to the dual-parameter verification abnormal test structure set and the verification normal test structure set, respectively.

[0017] When it is determined that the proportion of device speed-related yield loss items in the dual-parameter verification abnormal test structure set is greater than the proportion of device speed-related yield loss items in the verification normal test structure set, and the proportion of defect-related yield loss items in the dual-parameter verification abnormal test structure set is greater than the proportion of defect-related yield loss items in the verification normal test structure set, the abnormal test structure acquisition method is determined to be effective.

[0018] Thirdly, embodiments of this application also provide an anomaly test structure acquisition device, comprising:

[0019] A test dataset acquisition unit is adapted to acquire a test dataset, which includes the drive current and leakage current of WAT wafer acceptability tests for each specific test structure in a specific test structure set.

[0020] The single-parameter normal test structure set acquisition unit is adapted to acquire single-parameter abnormal data where the driving current does not meet the corresponding driving current threshold or the leakage current does not meet the corresponding leakage current threshold, and obtain the single-parameter abnormal test structure set and single-parameter normal test structure set corresponding to the single-parameter abnormal data.

[0021] The dual-parameter feature threshold acquisition unit is adapted to construct dual-parameter features and acquire dual-parameter feature thresholds based on the drive current and leakage current of each single-parameter normal test structure in the single-parameter normal test structure set.

[0022] The two-parameter anomaly test structure set acquisition unit is adapted to acquire two-parameter anomaly data where the two-parameter features do not meet the two-parameter feature thresholds, and acquire each two-parameter anomaly test structure corresponding to the two-parameter anomaly data to obtain a two-parameter anomaly test structure set.

[0023] The abnormal test structure set acquisition unit is adapted to obtain an abnormal test structure set based on the single-parameter abnormal test structure set and the two-parameter abnormal test structure set.

[0024] Fourthly, embodiments of this application also provide a verification apparatus for an anomaly test structure acquisition method, comprising:

[0025] The structure set acquisition unit is adapted to acquire, by means of the abnormal test structure acquisition method described in any of the preceding items, a two-parameter verification abnormal test structure set and a verification abnormal test structure set in the verification specific test structure set corresponding to the verification dataset;

[0026] The verification normal test structure set acquisition unit is adapted to acquire the verification normal test structure set based on the verification specific test structure set and the verification abnormal test structure set;

[0027] The yield loss item ratio acquisition unit is adapted to acquire the device speed-related yield loss item ratio and the defect-related yield loss item ratio of the wafer test corresponding to the dual-parameter verification abnormal test structure set and the verification normal test structure set, respectively.

[0028] The validity determination unit is adapted to determine that the abnormal test structure acquisition method is valid when it is determined that the proportion of device speed-related yield loss items in the dual-parameter verification abnormal test structure set is greater than the proportion of device speed-related yield loss items in the verification normal test structure set, and the proportion of defect-related yield loss items in the dual-parameter verification abnormal test structure set is greater than the proportion of defect-related yield loss items in the verification normal test structure set.

[0029] Fifthly, embodiments of this application also provide an electronic device, including at least one memory and at least one processor; the memory stores a program, and the processor calls the program to execute the abnormal test structure acquisition method as described in any of the foregoing claims or the verification method of the abnormal test structure acquisition method as described in the foregoing claims.

[0030] Sixthly, embodiments of this application also provide a storage medium storing one or more computer-executable instructions, which, when executed, implement the abnormal test structure acquisition method or the verification method of the abnormal test structure acquisition method as described in any of the preceding claims.

[0031] Compared with the prior art, the technical solution of this application has the following advantages:

[0032] The abnormal test structure acquisition method provided in this application embodiment firstly acquires a test dataset, which includes the drive current and leakage current of each specific test structure in a specific test structure set for WAT wafer acceptability testing. Then, it acquires single-parameter abnormal data where the drive current does not meet the corresponding drive current threshold or the leakage current does not meet the corresponding leakage current threshold, thereby obtaining a single-parameter abnormal test structure set and a single-parameter normal test structure set corresponding to the single-parameter abnormal data. Based on the drive current and leakage current of each single-parameter normal test structure in the single-parameter normal test structure set, it constructs a two-parameter feature, acquires a two-parameter feature threshold, and then acquires two-parameter abnormal data where the two-parameter feature does not meet the two-parameter feature threshold. It also acquires each two-parameter abnormal test structure corresponding to the two-parameter abnormal data, thereby obtaining a two-parameter abnormal test structure set. Finally, it obtains an abnormal test structure set based on the single-parameter abnormal test structure set and the two-parameter abnormal test structure set.

[0033] As can be seen, the technical solution provided in this application can obtain objective dual-parameter feature thresholds by obtaining dual-parameter feature thresholds based on the driving current and leakage current of each single-parameter normal test structure. This allows for objective comparison and acquisition when obtaining dual-parameter abnormal test structures through dual-feature parameter thresholds, reducing false positives and false negatives, improving the accuracy of acquisition, and consequently reducing the costs caused by subsequent false negatives of specific test structure anomalies. Attached Figure Description

[0034] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only embodiments of this application. For those skilled in the art, other drawings can be obtained based on the provided drawings without creative effort.

[0035] Figure 1 A flowchart illustrating the process of obtaining an anomaly test structure is shown.

[0036] Figure 2 This is a schematic diagram illustrating the correlation between the drive current of low threshold voltage devices and the product performance frequency parameter in the CP yield performance loss term.

[0037] Figure 3 This is a schematic diagram illustrating the correlation between leakage current of low threshold voltage devices and static leakage current parameters in the CP yield performance loss term.

[0038] Figure 4A schematic diagram for obtaining abnormal data of device drive current parameters using SPC control chart and set control range;

[0039] Figure 5 A schematic diagram for obtaining abnormal data of device leakage current parameters using SPC control chart and set control range;

[0040] Figure 6 This is a schematic diagram showing the correlation between the drive current parameters and leakage current parameters of each specific test structure after processing.

[0041] Figure 7 This is a flowchart illustrating the method for obtaining abnormal test structures provided in an embodiment of this application;

[0042] Figure 8 This is a schematic diagram of the test dataset for the abnormal test structure acquisition method provided in the embodiments of this application;

[0043] Figure 9 The product specification data limits for the SPC control charts provided in the embodiments of this application;

[0044] Figure 10 This is a schematic diagram showing the distribution range of the driving current values.

[0045] Figure 11 This is a schematic diagram showing the distribution range of leakage current values.

[0046] Figure 12 This is a schematic diagram illustrating the use of a two-parameter threshold to obtain two-parameter outlier data.

[0047] Figure 13 This is a flowchart illustrating the verification method of the abnormal test structure acquisition method provided in the embodiments of this application;

[0048] Figure 14 This shows the yield loss corresponding to the aforementioned product A in the CP test;

[0049] Figure 15 An optional block diagram of the anomaly test structure acquisition device provided in an embodiment of this application is shown;

[0050] Figure 16 An optional block diagram of the verification apparatus for the abnormal test structure acquisition method provided in the embodiments of this application is shown. Detailed Implementation

[0051] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of this application.

[0052] In a WAT ​​anomaly testing structure, in order to obtain WAT test anomaly data, the SPC (Statistical Process Control) principle is followed to process the WAT test data. SPC refers to the analysis, evaluation and process control of the production process using mathematical statistics methods, so as to maintain the process in a controlled state that is only affected by random factors, in order to achieve the purpose of quality control.

[0053] For details, please refer to Figure 1 , Figure 1 A flowchart illustrating the process of obtaining an anomaly test structure is shown.

[0054] like Figure 1 As shown, the process may include the following steps:

[0055] S00, obtain WAT test data.

[0056] WAT testing is performed on the wafer fabrication process to measure the electrical parameters of specific test structures and obtain WAT test data for specific test structures of each batch of wafer fabrication.

[0057] The WAT test data involves many parameters. Research has found that the device drive current and leakage current parameters of different threshold voltage devices (including conventional threshold voltage devices, low threshold voltage devices, ultra-low threshold voltage devices, etc.) are often correlated with the performance loss items of wafer testing (CP) yield (such as product performance frequency parameter test items and static leakage current test items). Therefore, by tracing the abnormal data of drive current and leakage current parameters in the WAT parameters, the yield loss problem of wafer testing can be found. Furthermore, by tracing the abnormal data of drive current and leakage current parameters in the WAT parameters, the specific test structure with the problem can be screened out, avoiding subsequent processing and reducing processing costs.

[0058] It should be noted that wafer testing (CP) is a quality inspection step following wafer testing (WAT) in order to detect various manufacturing defects, select defective products, thereby improving the yield rate of chips leaving the factory and reducing the cost of subsequent packaging.

[0059] Specifically, the correlation between the device drive current parameters and leakage current parameters of devices with different threshold voltages and the performance loss terms of wafer testing (CP) yield can be found in [reference needed]. Figure 2 and Figure 3 , Figure 2 This diagram illustrates the correlation between the drive current of low threshold voltage devices and the product performance frequency parameter in the CP yield performance loss term. Figure 3 This is a schematic diagram illustrating the correlation between leakage current of low threshold voltage devices and the static leakage current parameter in the CP yield performance loss term.

[0060] like Figure 2 As shown, the product performance frequency parameter in the CP yield performance loss term of the low threshold voltage device drive current satisfies the univariate linear regression condition y=k0x+k1, as follows. Figure 3 As shown, the static leakage current parameter in the CP yield performance loss term of low threshold voltage devices also satisfies the univariate linear regression condition y=k0x+k1.

[0061] Therefore, it is necessary to obtain test data of the drive current parameters and leakage current parameters of different threshold voltage devices in the WAT parameters, in order to prepare for the subsequent acquisition of abnormal WAT test data and to find the specific test structure of the abnormality.

[0062] S01, obtain the SPC abnormal dataset and SPC normal dataset of WAT test data.

[0063] For the aforementioned device drive current parameters and device leakage current parameters, control ranges are set respectively. Test data that exceed their respective control ranges are SPC abnormal data. All SPC abnormal data constitute an SPC abnormal dataset. Test data that fall within their respective control ranges are SPC normal data. All SPC normal data constitute an SPC normal dataset.

[0064] The control range can be (median of test data - 3σ, median of test data + 3σ), where σ is the standard deviation, a statistic that measures the degree of data dispersion. The physical meaning of standard deviation is to measure the degree of deviation of a set of data from its mean; the larger the standard deviation, the greater the data dispersion, and vice versa. By continuously collecting data and calculating its mean and standard deviation, the stability and controllability of the production process can be evaluated.

[0065] By utilizing SPC control charts and defined control ranges, abnormal data can be acquired. For details, please refer to [reference needed]. Figure 4 and Figure 5 , Figure 4 This is a schematic diagram illustrating the acquisition of abnormal data of device drive current parameters using SPC control charts and set control ranges. Figure 5 This is a schematic diagram illustrating the acquisition of abnormal data of device leakage current parameters using SPC control charts and the set control range.

[0066] It is easy to understand that, for each specific test structure, during the WAT test, both the test data of the drive current parameter and the test data of the leakage current parameter are acquired. If either of them exceeds its respective control range, then the corresponding specific test structure is an abnormal structure. Only the test data of the specific test structure where both the drive current parameter and the leakage current parameter are within the above control range can be called SPC normal data, and the data set composed of them is the normal dataset.

[0067] Of course, the specific test structure corresponding to the SPC abnormal dataset is the SPC abnormal specific test structure, and the specific test structure corresponding to the SPC normal dataset is the SPC normal specific test structure.

[0068] Step S02: Process the normal SPC dataset to obtain the supplementary abnormal dataset.

[0069] The drive current parameters and leakage current parameters corresponding to each specific test structure in the normal SPC dataset are processed to obtain a supplementary abnormal dataset.

[0070] For details, please refer to Figure 6 , Figure 6 This is a schematic diagram showing the correlation between the drive current parameters and leakage current parameters of each specific test structure after processing.

[0071] As shown in the figure, the driving current parameters of each specific test structure in the SPC normal dataset are used as X-axis data and the leakage current parameters are used as Y-axis data. Outlier data are filtered out to obtain a supplementary abnormal dataset, thereby obtaining a supplementary abnormal specific test structure.

[0072] When supplementing abnormal datasets and obtaining specific test structures based on the above methods, it usually relies on human-calibrated experience, which can easily lead to misjudgment or omission, resulting in the omission of the identified problematic specific test structures.

[0073] Therefore, in order to solve the problem of low accuracy of the above-mentioned methods for obtaining specific test structures of anomalies, this application provides a method for obtaining anomaly test structures, which can improve the accuracy of obtaining specific test structures of anomalies by constructing a two-parameter feature threshold that is not affected by human subjectivity.

[0074] For details, please refer to Figure 7 , Figure 7 This is a flowchart illustrating the method for obtaining abnormal test structures provided in the embodiments of this application.

[0075] As shown in the figure, the abnormal test structure acquisition method provided in this application embodiment may include the following steps:

[0076] Step S10: Obtain the test dataset, which includes the drive current and leakage current of the WAT wafer acceptability test for each specific test structure in the specific test structure set.

[0077] As described in step S00, after the tape-out is completed, a WAT ​​test is performed on a specific test structure to obtain a test dataset. It is easy to understand that there is a lot of test data obtained after the WAT test. How to select the data for the abnormal test structure acquisition method provided in this application embodiment needs to be studied and determined. For this purpose, the WAT test data and wafer test data corresponding to the chips that have completed the WAT test and CP yield test can be used to determine the data.

[0078] Analysis of the historical data used for verification shows that the device drive current of different threshold voltage devices (including conventional threshold voltage, low threshold voltage, ultra-low threshold voltage, etc.) is strongly correlated with the product performance frequency parameter test item in the yield test, and the device leakage current of different threshold voltage devices is strongly correlated with the product static leakage current test item in the yield test.

[0079] Therefore, in the abnormal test structure acquisition method provided in this application embodiment, the acquired test dataset includes the drive current and leakage current of the WAT wafer acceptability test of each specific test structure in the specific test structure set.

[0080] For specific data presentation formats, please refer to [link / reference]. Figure 8 , Figure 8 This is a schematic diagram of the test dataset for the abnormal test structure acquisition method provided in the embodiments of this application.

[0081] As shown in the figure, each row represents information about a specific test structure, including the batch (lot), wafer number, the specific location of the test structure on the wafer, and the drive current and leakage current of devices with different threshold voltages.

[0082] Since abnormal data may be generated during WAT testing due to the testing method, to avoid interference from such data, in one specific implementation, the step of obtaining the test dataset may include:

[0083] Obtain the raw dataset for WAT wafer acceptability testing for each specific test structure;

[0084] Clean the invalid data in the original dataset.

[0085] First, obtain the original dataset for testing, and then clean the original dataset to remove obviously invalid data.

[0086] Specifically, the failure data can be data with values ​​of (+999, -999, 0).

[0087] The cleaning method can be to change the above data in the table to "unknown".

[0088] Of course, in other embodiments, the row containing the specific test structure corresponding to the above data can be deleted.

[0089] Step S11: Obtain single-parameter abnormal data where the driving current does not meet the corresponding driving current threshold or the leakage current does not meet the corresponding leakage current threshold, and obtain the single-parameter abnormal test structure set and the single-parameter normal test structure set corresponding to the single-parameter abnormal data.

[0090] After obtaining the test dataset, first obtain the single-parameter exception data.

[0091] In one specific implementation, the above-mentioned single-parameter abnormal data can be obtained using an SPC control chart. For this purpose, it is also necessary to obtain the drive current threshold and the leakage current threshold. Before obtaining the single-parameter test data where the drive current does not meet the corresponding drive current threshold or the leakage current does not meet the corresponding leakage current threshold, the product specification data limit (i.e., the control range) corresponding to the threshold voltage of each of the specific test structures is first obtained to obtain each of the drive current thresholds and each of the leakage current thresholds.

[0092] Please refer to the details. Figure 9 , Figure 9 The product specification data limits are provided for the SPC control charts in the embodiments of this application.

[0093] As shown in the figure, for different devices with different threshold voltages, minimum, maximum, and target values ​​for drive current, as well as minimum, maximum, and target values ​​for leakage current, are provided. The maximum and minimum values ​​for drive current form the drive current threshold, and the maximum and minimum values ​​for leakage current form the leakage current threshold.

[0094] It should be noted that the minimum, maximum, and target values ​​of the drive current for different devices corresponding to different threshold voltages, as well as the minimum, maximum, and target values ​​of the leakage current, can be adjusted and determined based on experience.

[0095] Of course, in another specific implementation, the product specification data limits corresponding to the threshold voltages of each of the specific test structures can also be called in advance to obtain each of the drive current thresholds and each of the leakage current thresholds.

[0096] Based on the test data of each specific test structure, the corresponding drive current threshold, and the corresponding leakage current threshold, test data exceeding the corresponding threshold range can be obtained, i.e., single-parameter abnormal data.

[0097] In one specific implementation, single-parameter abnormal data can be obtained through images. Of course, in other implementations, single-parameter abnormal data can also be obtained through other methods.

[0098] After obtaining the single-parameter anomaly data, the corresponding single-parameter anomaly test structure can be obtained based on the single-parameter anomaly data. It is easy to understand that single-parameter anomaly test data can be obtained based on both abnormal drive current and abnormal leakage current data. Then, regardless of the type of single-parameter anomaly test data, the corresponding single-parameter anomaly test structure can be obtained. Therefore, the obtained single-parameter anomaly test structures include:

[0099] Abnormal test data for a single parameter where only the drive current does not meet the corresponding drive current threshold;

[0100] Single-parameter abnormal test data where only the leakage current fails to meet the corresponding leakage current threshold; and

[0101] Abnormal test data for a single parameter where the drive current does not meet the corresponding drive current threshold and the leakage current does not meet the corresponding leakage current threshold.

[0102] The obtained single-parameter anomaly test structures form a single-parameter anomaly test structure set.

[0103] Based on the specific test structure set and the single-parameter abnormal test structure set corresponding to the test dataset, the single-parameter normal test structure set can be obtained.

[0104] Of course, there may still be abnormal test structures in the set of normal single-parameter test structures. In order to more comprehensively identify abnormal test structures, further steps are required:

[0105] Step S12: Based on the driving current and leakage current of each single-parameter normal test structure in the single-parameter normal test structure set, construct a two-parameter feature and obtain the two-parameter feature threshold.

[0106] In one specific implementation, a dual-parameter feature is established directly based on the drive current and leakage current from the WAT test data, with the drive current used as the X-axis data and the leakage current used as the Y-axis data.

[0107] However, the distribution ranges of drive current and leakage current differ significantly; please refer to [reference needed]. Figure 10 and Figure 11 , Figure 10 This is a schematic diagram showing the range of driving current values. Figure 11The diagram illustrates the distribution range of leakage current values. It shows that the driving current value distribution range is much greater than 1, while the leakage current value distribution range is relatively small, hovering around 1. Therefore, to avoid the driving current with a large value dominating the gradient update direction and causing continuous oscillations on the regression dataset plane, feature scaling and normalization can be performed before constructing the two-parameter features. Specifically, in one implementation, before constructing the two-parameter features based on the driving current and leakage current corresponding to the same specific test structure in the single-parameter normal dataset in step S12, and before obtaining the two-parameter feature threshold, the method may further include: performing feature scaling and normalization on the driving current in the single-parameter normal dataset.

[0108] Specifically, the drive current in the single-parameter normal dataset can be feature-scaled and normalized using the following formula:

[0109] x normalized = (x - min(x)) / (max(x) - min(x))

[0110] Where, x normalized --Drive current after feature scaling and normalization, and x normalized ∈[0,1];

[0111] x -- Drive current before feature scaling and normalization;

[0112] min(x) -- the minimum value of the drive current in the single-parameter normal dataset;

[0113] max(x) -- The maximum value of the drive current in the single-parameter normal dataset.

[0114] A two-parameter characteristic is established by using the leakage current and normalized drive current of each group corresponding to the same abnormal test structure.

[0115] After obtaining the two-parameter features, the two-parameter feature threshold is obtained based on the two-parameter features.

[0116] It is easy to understand that, since it is a two-parameter feature, the two-parameter feature threshold is the range of a region on a two-dimensional plane, which can be a square region, a circular region, or an irregular shape region.

[0117] In one specific implementation, the two-parameter feature threshold may include the MCD (Minimum Covariance Determinant) Mahalanobis distance threshold.

[0118] Specifically, the minimum covariance determinant Mahalanobis distance threshold of the MCD can be obtained through the following steps:

[0119] First, the covariance matrix is ​​obtained based on the drive current and leakage current corresponding to the same specific test structure in the single-parameter normal dataset;

[0120] Then, the minimum covariance determinant matrix is ​​obtained by using the MCD minimum covariance determinant algorithm to obtain the matrix with the smallest covariance determinant among the matrices formed by the subsets of the single-parameter normal dataset. The minimum covariance determinant matrix is ​​obtained by using the data elements formed by the driving current and the leakage current corresponding to the same specific test structure and the minimum covariance determinant matrix to obtain the Mahalanobis distance anomaly score threshold, thus obtaining the MCD Mahalanobis distance threshold.

[0121] To obtain the covariance matrix, we first establish the driving current matrix X = (X1, ..., X2). n ) t With the leakage current matrix Y = (Y1,...Y n ) t Where X1,...X n X is the normalized drive current. n Y n Corresponding to the same specific test structure, where (X) n Y n A single parameter can be understood as a data element, and a normal dataset with one parameter contains n data elements.

[0122] Then, using the covariance formula:

[0123]

[0124] Calculate the covariance matrix

[0125]

[0126] Thus, the covariance matrix z is obtained.

[0127] To obtain the minimum covariance determinant Mahalanobis distance threshold for MCD, we first obtain the minimum covariance matrix:

[0128] The minimum covariance determinant Mahalanobis distance algorithm (MCD) is used to exhaustively select h data elements from n data elements in a single-parameter normal dataset to form a subset, where h >= (n+p+1) / 2, p = 2 elements. The determinant of the covariance matrix corresponding to each subset is obtained, and the covariance matrix with the smallest determinant is obtained.

[0129] Obtain the minimum covariance determinant matrix Then, obtain the minimum covariance determinant matrix. The mean estimator of the corresponding h data elements Covariance estimator Σ i :

[0130]

[0131]

[0132] Where, x i-- Minimum covariance determinant matrix Each data element in;

[0133] K MCD (h,n,p) -- MCD proportionality constant, used to ensure the consistency and unbiasedness of the covariance estimator.

[0134] Then calculate the distance between each of the n data elements in the single-parameter normal dataset and the center of the subset:

[0135]

[0136] Therefore, the minimum covariance determinant matrix formed for a fixed subset of data. You can get n distances.

[0137] Next, sort the n distances in ascending order, select the h data elements with the highest distances to form a new sample set, and iterate the calculation until ∑n=∑n-1, then record d(n)=d(n). last The threshold for Mahalanobis distance anomaly scores.

[0138] It is easy to understand that as h changes, multiple Mahalanobis distance anomaly score thresholds can be obtained, and the boundary of the region enclosed by each Mahalanobis distance anomaly score threshold is the minimum covariance determinant Mahalanobis distance threshold of the MCD.

[0139] Step S13: Obtain dual-parameter abnormal data where the dual-parameter features do not meet the dual-parameter feature threshold, and obtain each dual-parameter abnormal test structure corresponding to the dual-parameter abnormal data to obtain a set of dual-parameter abnormal test structures.

[0140] By obtaining the two-parameter features and the two-parameter feature threshold, and acquiring the two-parameter abnormal data that does not meet the two-parameter feature threshold, the corresponding two-parameter abnormal test structure and the two-parameter abnormal test structure set can be obtained.

[0141] Specifically, you can refer to Figure 12 , Figure 12 This is a schematic diagram illustrating the use of a two-parameter threshold to obtain two-parameter abnormal data.

[0142] Step S14: Obtain the anomaly test structure set based on the single-parameter anomaly test structure set and the two-parameter anomaly test structure set.

[0143] The single-parameter anomaly test structure set and the two-parameter anomaly test structure set are merged to obtain the anomaly test structure set.

[0144] As can be seen, the technical solution provided in this application can obtain objective dual-parameter feature thresholds by obtaining dual-parameter feature thresholds based on the driving current and leakage current of each single-parameter normal test structure. This allows for objective comparison and acquisition when obtaining dual-parameter abnormal test structures through dual-feature parameter thresholds, reducing false positives and false negatives, improving the accuracy of acquisition, and consequently reducing the costs caused by subsequent false negatives of specific test structure anomalies.

[0145] The following uses product A as an example to illustrate the method for obtaining abnormal test structures provided in this application:

[0146] 1. Use the low threshold voltage device of product A as the anchor device for analysis, and obtain the corresponding WAT test data, including drive current, leakage current, batch (lot), wafer (wafer) number, and specific location information of each specific test structure on the wafer. Taking product A as an example, generate a data list with 51,503 rows and 19 columns.

[0147] 2. Retrieve invalid data from the data list and change the data with the value (+999, -999, 0) to "unknown".

[0148] 3. Import product specification data limits from the SPC control chart;

[0149] 4. Based on the drive current and drive current threshold, 400 single-parameter abnormal data points were selected; based on the leakage current and leakage current threshold, 42 single-parameter abnormal data points were selected; the union of the two was obtained to get 403 single-parameter abnormal test structures and 51,100 single-parameter normal test structures.

[0150] 5. Establish dual-parameter features using 51,100 single-parameter normal test structures, and obtain the minimum covariance determinant Mahalanobis distance threshold of MCD.

[0151] 6. Based on the two-parameter features and the Mahalanobis distance threshold of the minimum covariance determinant of the MCD, 458 two-parameter anomaly test structures were obtained.

[0152] 7. Based on 403 single-parameter anomaly test structures and 458 two-parameter anomaly test structures, 861 anomaly test structures were obtained.

[0153] In addition, to verify the effectiveness of the abnormal test structure acquisition method provided in this application embodiment, this application embodiment also provides a verification method for the abnormal test structure acquisition method. It is easy to understand that the verification method for the abnormal test structure acquisition method is used before the abnormal test structure is specifically used to acquire the abnormal test structure, so as to demonstrate through data that the abnormal test structure acquisition method provided in this application embodiment can improve the accuracy of the acquired abnormal test structure.

[0154] Please refer to the details. Figure 13 , Figure 13 This is a flowchart illustrating the verification method of the abnormal test structure acquisition method provided in the embodiments of this application.

[0155] As shown in the figure, the verification method provided in this application embodiment includes:

[0156] Step S20: Obtain the two-parameter verification anomaly test structure set and the verification anomaly test structure set corresponding to the verification dataset using the anomaly test structure acquisition method.

[0157] For specific acquisition methods, please refer to the aforementioned implementation description, which will not be repeated here.

[0158] It is easy to understand that in the verification method, the data used is the verification dataset. The corresponding verification-specific test structure has been processed and CP test has been performed to obtain the wafer yield loss dataset for subsequent verification.

[0159] In addition, for the validation dataset, the obtained results are a single-parameter validation anomaly test structure set, a two-parameter validation anomaly test structure set, and a validation anomaly test structure set.

[0160] Step S21: Obtain the normal test structure set based on the specific test structure set and the abnormal test structure set.

[0161] By removing each abnormal test structure from a specific set of test structures, you can obtain a set of normal test structures.

[0162] Step S22: Obtain the proportion of device speed-related yield loss items and the proportion of defect-related yield loss items for wafer testing corresponding to the dual-parameter verification abnormal test structure set and the verification normal test structure set, respectively.

[0163] Since the verification dataset used for verification has been processed and CP testing has been performed on the corresponding verification-specific test structures, the proportion of device speed-related yield loss and defect-related yield loss of the verification-specific test structures in each structure set during CP testing can be obtained based on the dual-parameter verification abnormal test structure set and the verification normal test structure set.

[0164] It should be noted that defect-related yield loss items can specifically include scanmix, scanchain, and Jtag.

[0165] JTAG testing is a commonly used chip testing method that tests the chip's electrical performance and functional correctness by inputting test modes into the chip's JTAG interface.

[0166] Scan testing, including scanmix and scanchain, is a chip testing method based on scanchain. It tests the electrical performance and logical correctness of a chip by inputting test modes into the scanchain.

[0167] In one specific implementation, the set of dual-parameter verification anomaly test structures is the set of MCD minimum covariance determinants, and the set of verification normal test structures is the set of verification-specific test structures remaining after removing the SPC set and the MCD minimum covariance determinant set from all verification-specific test structures.

[0168] Please refer to Figure 14 , Figure 14 The yield loss corresponding to the aforementioned product A in the CP test is shown.

[0169] As shown in the figure, for product A, yield losses were obtained for different Hbinning items, including the SPC set, the MCD minimum covariance determinant set, the SPC+MCD minimum covariance determinant set, and the normal value set (i.e., the verification normal test structure set).

[0170] Step S23: Determine whether the proportion of device speed-related yield loss items in the dual-parameter verification abnormal test structure set is greater than the proportion of device speed-related yield loss items in the verification normal test structure set, and whether the proportion of defect-related yield loss items in the dual-parameter verification abnormal test structure set is greater than the proportion of defect-related yield loss items in the verification normal test structure set. If yes, proceed to step S24; otherwise, proceed to step S25.

[0171] Step S24: Determine that the abnormal test structure acquisition method is effective.

[0172] Step S25: Determine that the abnormal test structure acquisition method is invalid.

[0173] In this way, by using the aforementioned method, it can be determined whether the abnormal test structure acquisition method provided in the embodiments of this application can truly improve the accuracy of abnormal test structure acquisition, and ensure that the abnormal test structure acquisition method is truly effective. This will enable the use of the abnormal test structure acquisition method provided in the embodiments of this application in subsequent applications to find more abnormal specific test structures, reduce false positives and false negatives, and reduce the cost caused by the false negatives of specific test structure anomalies.

[0174] To address the aforementioned issues, this application also provides an anomaly test structure acquisition device. This device can be considered as a functional module required to implement the anomaly test structure acquisition method provided in this application. The device described below can be referred to in correspondence with the method described above.

[0175] As an optional implementation, Figure 15 An optional block diagram of the anomaly test structure acquisition device provided in an embodiment of this application is shown. For example... Figure 15 As shown, the device may include:

[0176] Test dataset acquisition unit 100 is adapted to acquire test dataset, which includes drive current and leakage current of WAT wafer acceptability test for each specific test structure in a specific test structure set.

[0177] The single-parameter normal test structure set acquisition unit 110 is adapted to acquire single-parameter abnormal data where the driving current does not meet the corresponding driving current threshold or the leakage current does not meet the corresponding leakage current threshold, and obtain the single-parameter abnormal test structure set and single-parameter normal test structure set corresponding to the single-parameter abnormal data.

[0178] The dual-parameter feature threshold acquisition unit 120 is adapted to construct dual-parameter features and acquire dual-parameter feature thresholds based on the driving current and leakage current of each single-parameter normal test structure in the single-parameter normal test structure set.

[0179] The two-parameter anomaly test structure set acquisition unit 130 is adapted to acquire two-parameter anomaly data where the two-parameter features do not meet the two-parameter feature thresholds, and acquire each two-parameter anomaly test structure corresponding to the two-parameter anomaly data to obtain a two-parameter anomaly test structure set.

[0180] The abnormal test structure set acquisition unit 140 is adapted to obtain an abnormal test structure set based on the single-parameter abnormal test structure set and the two-parameter abnormal test structure set.

[0181] In some embodiments, the two-parameter feature threshold includes the MCD minimum covariance determinant Mahalanobis distance threshold.

[0182] In some embodiments, the two-parameter anomaly test structure set acquisition unit 130 is adapted to acquire the minimum covariance determinant Mahalanobis distance threshold of the MCD, including:

[0183] The covariance matrix is ​​obtained based on the drive current and leakage current corresponding to the same specific test structure in the single-parameter normal dataset;

[0184] The minimum covariance determinant matrix is ​​obtained by using the MCD minimum covariance determinant algorithm to obtain the matrix with the smallest covariance determinant among the matrices formed by the subsets of the single-parameter normal dataset. The minimum covariance determinant matrix is ​​then obtained by using the data elements formed by the driving current and the leakage current corresponding to the same specific test structure and the minimum covariance determinant matrix to obtain the Mahalanobis distance anomaly score threshold, thus obtaining the MCD minimum covariance determinant Mahalanobis distance threshold.

[0185] Specifically, the test dataset acquisition unit 100 is suitable for acquiring test datasets including:

[0186] Obtain the raw dataset for WAT wafer acceptability testing for each specific test structure;

[0187] Clean the invalid data in the original dataset.

[0188] In some embodiments, the anomaly test structure acquisition device further includes:

[0189] The threshold acquisition unit is adapted to acquire the product specification data boundary corresponding to the threshold voltage of each of the specific test structures, and to obtain each of the drive current thresholds and each of the leakage current thresholds.

[0190] In some embodiments, the anomaly test structure acquisition device further includes:

[0191] The feature scaling and normalization unit is adapted to perform feature scaling and normalization on the drive current in the single-parameter normal dataset.

[0192] Specifically, the feature scaling and normalization unit is adapted to perform feature scaling and normalization on the drive current in the single-parameter normal dataset, including:

[0193] The driving current in the single-parameter normal dataset is feature-scaled and normalized using the following formula:

[0194] x normalized = (x - min(x)) / (max(x) - min(x))

[0195] Where, x normalized --Drive current after feature scaling and normalization, and x normalized ∈[0,1];

[0196] x -- Drive current before feature scaling and normalization;

[0197] min(x) -- the minimum value of the drive current in the single-parameter normal dataset;

[0198] max(x) -- The maximum value of the drive current in the single-parameter normal dataset.

[0199] As can be seen, the technical solution provided in this application can obtain objective dual-parameter feature thresholds by obtaining dual-parameter feature thresholds based on the driving current and leakage current of each single-parameter normal test structure. This allows for objective comparison and acquisition when obtaining dual-parameter abnormal test structures through dual-feature parameter thresholds, reducing false positives and false negatives, improving the accuracy of acquisition, and consequently reducing the costs caused by subsequent false negatives of specific test structure anomalies.

[0200] To address the aforementioned issues, embodiments of this application also provide a verification apparatus for an anomaly test structure acquisition method. This apparatus can be considered as a functional module required to implement the verification method for the anomaly test structure acquisition method provided in this application. The apparatus described below can be referred to in correspondence with the method described above.

[0201] As an optional implementation, Figure 16 An optional block diagram of the verification apparatus for the abnormal test structure acquisition method provided in an embodiment of this application is shown. For example... Figure 16 As shown, the device may include:

[0202] The structure set acquisition unit 200 is adapted to acquire, through any of the aforementioned abnormal test structure acquisition methods, the two-parameter verification abnormal test structure set and the verification abnormal test structure set in the verification specific test structure set corresponding to the verification dataset;

[0203] The verification normal test structure set acquisition unit 210 is adapted to acquire the verification normal test structure set based on the verification specific test structure set and the verification abnormal test structure set.

[0204] The yield loss item ratio acquisition unit 220 is adapted to acquire the device speed-related yield loss item ratio and defect-related yield loss item ratio of the wafer test corresponding to the dual-parameter verification abnormal test structure set and the verification normal test structure set, respectively.

[0205] The validity determination unit 230 is adapted to determine that the abnormal test structure acquisition method is valid when it is determined that the proportion of device speed-related yield loss items in the dual-parameter verification abnormal test structure set is greater than the proportion of device speed-related yield loss items in the verification normal test structure set, and the proportion of defect-related yield loss items in the dual-parameter verification abnormal test structure set is greater than the proportion of defect-related yield loss items in the verification normal test structure set.

[0206] This application also provides an electronic device, including at least one memory and at least one processor; the memory stores a program, and the processor calls the program to execute the abnormal test structure acquisition method as described in any of the foregoing embodiments or the verification method of the abnormal test structure acquisition method as described in any of the foregoing embodiments.

[0207] This application also provides a storage medium storing one or more computer-executable instructions. When the one or more computer-executable instructions are executed, they implement the abnormal test structure acquisition method or the verification method of the abnormal test structure acquisition method as described in any of the foregoing embodiments.

[0208] While the embodiments disclosed above are described in this application, this application is not limited thereto. Any person skilled in the art can make various modifications and alterations without departing from the spirit and scope of this application; therefore, the scope of protection of this application should be determined by the scope defined in the claims.

Claims

1. A method for obtaining anomaly test structures, characterized in that, include: Obtain a test dataset, which includes the drive current and leakage current of WAT wafer acceptability tests for each specific test structure in a specific test structure set; wherein, there is a correlation between the device drive current parameters and leakage current parameters of different threshold voltage devices and the performance loss term of wafer test yield. Obtain the product specification data limits corresponding to the threshold voltage of each of the specific test structures, and obtain each of the drive current thresholds and each of the leakage current thresholds; Obtain single-parameter abnormal data where the drive current does not meet the corresponding drive current threshold or the leakage current does not meet the corresponding leakage current threshold, and obtain the single-parameter abnormal test structure set and single-parameter normal test structure set corresponding to the single-parameter abnormal data. The driving current in the single-parameter normal dataset is subjected to feature scaling and normalization. Based on the driving current and leakage current of each single-parameter normal test structure in the single-parameter normal test structure set, a two-parameter feature is constructed, and a two-parameter feature threshold is obtained, wherein the two-parameter feature threshold includes the MCD minimum covariance determinant Mahalanobis distance threshold. Obtain biparametric abnormal data where the biparametric features do not meet the biparametric feature threshold, and obtain each biparametric abnormal test structure corresponding to the biparametric abnormal data to obtain a biparametric abnormal test structure set; Based on the single-parameter anomaly test structure set and the two-parameter anomaly test structure set, an anomaly test structure set is obtained; The MCD minimum covariance determinant Mahalanobis distance threshold is obtained through the following steps: Based on the driving current and leakage current corresponding to the same specific test structure in the single-parameter normal dataset, a covariance matrix is ​​obtained; using the MCD minimum covariance determinant algorithm, the matrix with the smallest covariance determinant among the matrices formed by the subsets of the single-parameter normal dataset is obtained, thus obtaining the minimum covariance determinant matrix; using the data elements formed by the driving current and leakage current corresponding to the same specific test structure and the minimum covariance determinant matrix, a Mahalanobis distance anomaly score threshold is obtained, thus obtaining the MCD minimum covariance determinant Mahalanobis distance threshold. The step of obtaining the Mahalanobis distance anomaly score threshold by using data elements formed by the driving current and the leakage current corresponding to the same specific test structure and the minimum covariance determinant matrix, and obtaining the MCD minimum covariance determinant Mahalanobis distance threshold, includes: Obtain the minimum covariance determinant matrix The mean estimator of the corresponding h data elements Estimator of covariance ∑ i Where h>=(n+p+1) / 2, p=2 elements, and n is the number of data elements in the single-parameter normal dataset; Calculate the distance between each of the n data elements in a single-parameter normal dataset and the center of the subset corresponding to the minimum covariance determinant matrix; Sort the n distances in ascending order, select the h data elements with the highest distances to form a new sample set, and iterate through the calculation until ∑n=∑n-1, then record d(n)=d(n). last The threshold for Mahalanobis distance anomaly scores; Based on the change of h, multiple Mahalanobis distance anomaly score thresholds are obtained, and the boundary of the region enclosed by each Mahalanobis distance anomaly score threshold is used as the minimum covariance determinant Mahalanobis distance threshold of MCD. The method includes a verification method for the abnormal test structure acquisition method before acquiring the abnormal test structure. This verification method includes: acquiring a two-parameter verification abnormal test structure set and a verification abnormal test structure set corresponding to a specific verification test structure set using the abnormal test structure acquisition method; acquiring a verification normal test structure set based on the specific verification test structure set and the verification abnormal test structure set; acquiring the proportions of device speed-related yield loss items and defect-related yield loss items for wafer testing corresponding to the two-parameter verification abnormal test structure set and the verification normal test structure set, respectively; and determining that the abnormal test structure acquisition method is effective when it is determined that the proportion of device speed-related yield loss items in the two-parameter verification abnormal test structure set is greater than that in the verification normal test structure set, and the proportion of defect-related yield loss items in the two-parameter verification abnormal test structure set is greater than that in the verification normal test structure set.

2. The method for obtaining abnormal test structures as described in claim 1, characterized in that, The steps for obtaining the test dataset include: Obtain the raw dataset for WAT wafer acceptability testing for each specific test structure; Clean the invalid data in the original dataset.

3. The method for obtaining abnormal test structures as described in claim 1, characterized in that, The step of performing feature scaling and normalization on the driving current in the single-parameter normal dataset includes: The driving current in the single-parameter normal dataset is feature-scaled and normalized using the following formula: x normalized =(x-min(x)) / (max(x)-min(x)) Where, x normalized --Drive current after feature scaling and normalization, and x normalized ∈[0,1]; x -- Drive current before feature scaling and normalization; min(x) -- the minimum value of the drive current in the single-parameter normal dataset; max(x) -- The maximum value of the drive current in the single-parameter normal dataset.

4. An anomaly test structure acquisition device, characterized in that, include: The test dataset acquisition unit is adapted to acquire a test dataset, which includes the drive current and leakage current of WAT wafer acceptability tests for each specific test structure in a specific test structure set; wherein, there is a correlation between the device drive current parameters and leakage current parameters of different threshold voltage devices and the performance loss term of wafer test yield. The threshold acquisition unit is adapted to acquire the product specification data boundary corresponding to the threshold voltage of each of the specific test structures, and to obtain each of the drive current thresholds and each of the leakage current thresholds; The single-parameter normal test structure set acquisition unit is adapted to acquire single-parameter abnormal data where the driving current does not meet the corresponding driving current threshold or the leakage current does not meet the corresponding leakage current threshold, and obtain the single-parameter abnormal test structure set and single-parameter normal test structure set corresponding to the single-parameter abnormal data. The feature scaling and normalization unit is adapted to perform feature scaling and normalization on the drive current in the single-parameter normal dataset; A dual-parameter feature threshold acquisition unit is adapted to construct dual-parameter features and acquire dual-parameter feature thresholds based on the driving current and leakage current of each single-parameter normal test structure in the single-parameter normal test structure set, wherein the dual-parameter feature thresholds include the MCD minimum covariance determinant Mahalanobis distance threshold. The two-parameter anomaly test structure set acquisition unit is adapted to acquire two-parameter anomaly data where the two-parameter features do not meet the two-parameter feature thresholds, and acquire each two-parameter anomaly test structure corresponding to the two-parameter anomaly data to obtain a two-parameter anomaly test structure set. An anomaly test structure set acquisition unit is adapted to obtain an anomaly test structure set based on the single-parameter anomaly test structure set and the two-parameter anomaly test structure set; The dual-parameter abnormal test structure set acquisition unit is adapted to obtain the MCD minimum covariance determinant Mahalanobis distance threshold by: obtaining a covariance matrix based on the driving current and leakage current corresponding to the same specific test structure in the single-parameter normal dataset; using the MCD minimum covariance determinant algorithm to obtain the matrix with the smallest covariance determinant among the matrices formed by the subsets of the single-parameter normal dataset, thus obtaining the minimum covariance determinant matrix; and using the data elements formed by the driving current and leakage current corresponding to the same specific test structure and the minimum covariance determinant matrix to obtain the Mahalanobis distance abnormal score threshold, thus obtaining the MCD minimum covariance determinant Mahalanobis distance threshold. The step of obtaining the Mahalanobis distance anomaly score threshold by using data elements formed by the driving current and the leakage current corresponding to the same specific test structure and the minimum covariance determinant matrix, and obtaining the MCD minimum covariance determinant Mahalanobis distance threshold, includes: Obtain the minimum covariance determinant matrix The mean estimator of the corresponding h data elements Estimator of covariance ∑ i Where h>=(n+p+1) / 2, p=2 elements, and n is the number of data elements in the single-parameter normal dataset; Calculate the distance between each of the n data elements in a single-parameter normal dataset and the center of the subset corresponding to the minimum covariance determinant matrix; Sort the n distances in ascending order, select the h data elements with the highest distances to form a new sample set, and iterate through the calculation until ∑n=∑n-1, then record d(n)=d(n). last The threshold for Mahalanobis distance anomaly scores; Based on the change of h, multiple Mahalanobis distance anomaly score thresholds are obtained, and the boundary of the region enclosed by each Mahalanobis distance anomaly score threshold is used as the minimum covariance determinant Mahalanobis distance threshold of MCD. The method includes a verification method for the abnormal test structure acquisition method before acquiring the abnormal test structure. This verification method includes: acquiring a two-parameter verification abnormal test structure set and a verification abnormal test structure set corresponding to a specific verification test structure set using the abnormal test structure acquisition method; acquiring a verification normal test structure set based on the specific verification test structure set and the verification abnormal test structure set; acquiring the proportions of device speed-related yield loss items and defect-related yield loss items for wafer testing corresponding to the two-parameter verification abnormal test structure set and the verification normal test structure set, respectively; and determining that the abnormal test structure acquisition method is effective when it is determined that the proportion of device speed-related yield loss items in the two-parameter verification abnormal test structure set is greater than that in the verification normal test structure set, and the proportion of defect-related yield loss items in the two-parameter verification abnormal test structure set is greater than that in the verification normal test structure set.

5. The abnormal test structure acquisition device as described in claim 4, characterized in that, Test dataset acquisition unit, suitable for acquiring test datasets including: Obtain the raw dataset for WAT wafer acceptability testing for each specific test structure; Clean the invalid data in the original dataset.

6. The abnormal test structure acquisition device as described in claim 4, characterized in that, The feature scaling and normalization unit, adapted to perform feature scaling and normalization on the drive current in the single-parameter normal dataset, includes: The driving current in the single-parameter normal dataset is feature-scaled and normalized using the following formula: x normalized =(x-min(x)) / (max(x)-min(x)) Where, x normalized --Drive current after feature scaling and normalization, and x normalized ∈[0,1]; x -- Drive current before feature scaling and normalization; min(x) -- the minimum value of the drive current in the single-parameter normal dataset; max(x) -- The maximum value of the drive current in the single-parameter normal dataset.

7. An electronic device, characterized in that, It includes at least one memory and at least one processor; the memory stores a program, and the processor calls the program to execute the abnormal test structure acquisition method as described in any one of claims 1-3.

8. A storage medium, characterized in that, The storage medium stores one or more computer-executable instructions, which, when executed, implement the abnormal test structure acquisition method as described in any one of claims 1-3.

Images