Data processing method and device, chip and computer readable storage medium
By normalizing the wafer test data in the semiconductor industry and calculating the spatial vector similarity, the problems of low data analysis efficiency and high computing power consumption in the existing technology are solved, and efficient and interpretable abnormal data identification and Parameter positioning.
Patent Information
- Application Number
- CN202411360549.1
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-09-26
- Publication Date
- 2025-01-14
AI Technical Summary
The prior art has low efficiency in the semiconductor industry and high computing power consumption. Deep learning methods have poor generalization and lack interpretability, making it difficult to effectively identify abnormalities in diversified test data.
By obtaining the test data set for normalization, the space vector to be tested is constructed, and the vector similarity is calculated to determine the abnormal data. An algorithm based on logic rules is used to avoid model training, save computing resources, and improve data processing efficiency.
It realizes the output of second-level results, reduces computing power consumption, improves data processing efficiency, has high generalization characteristics and interpretability, and can accurately locate parameters that need to be adjusted.
Smart Images
Figure CN119312247A_ABST
Abstract
Description
Technical Field
[0001] The present application relates to the field of data processing technology, and in particular to a data processing method, device, chip and computer-readable storage medium. Background Art
[0002] Modern chips can only be obtained through a long semiconductor industry manufacturing process. For example, it is necessary to go through the process of silicon ingot-purification-cutting-grinding-polishing-lithography and etching-ion implantation and lamination-cutting separation to obtain wafer-testing-packaging-chip. Among them, after obtaining the wafer, the basic unit Die (particle) on it must be subjected to various types of tests (for example: current and voltage test, resistance test, etc.) to determine whether it is qualified. The analysis of test results is a very important part in the semiconductor industry. This importance is reflected in two aspects. On the one hand, Die is the basic unit of the chip. If a Die fails the test, it cannot be used to manufacture chips and can only be discarded; on the other hand, based on the analysis of the test results, the possible causes of the problems in the semiconductor process can be inferred. If the inferred cause is correct, it will help improve the process, thereby improving the yield of the semiconductor process and reducing the probability of defects in the wafer as much as possible.
[0003] Based on the above, it can be understood that data analysis of wafer test results is extremely important to the semiconductor industry, but data analysis of wafer test results is very difficult. On the one hand, there are many types of tests, and the types of data obtained are different; on the other hand, the amount of data obtained from the test is large, and it is more computationally intensive and time-consuming to perform data analysis.
[0004] A known prior art is to use a deep learning method to detect wafer test patterns, but the deep learning method is not suitable for the semiconductor business scenario. Since the semiconductor manufacturing process is very long and there are many types of tests, the "patterns / modes" generated are inexhaustible. The model obtained by the deep learning method can only "recognize" the "patterns / modes" that have appeared in the training data, and cannot recognize the "patterns / modes" that are not in the training data, so the generalization of the deep learning solution is poor; on the other hand, deep learning lacks interpretability as a whole, and its "recognition" process is a black-boxed probability distribution result output. When the "pattern / mode" is recognized incorrectly, it lacks interpretability; finally, the image recognition algorithm using deep learning is generally executed using a convolutional neural network, and the operation of the convolutional neural network requires a lot of computing power, so the overall energy consumption and speed are not good enough. Summary of the invention
[0005] Based on this, it is necessary to provide a data processing method, device, chip, computer-readable storage medium and computer program product that can save computing resources and improve data processing efficiency in response to the above technical problems.
[0006] In a first aspect, the present application provides a data processing method, comprising:
[0007] Acquire at least one test data set, and perform normalization processing on the at least one test data set to obtain a normalized parameter set;
[0008] According to the normalized parameter set, constructing a space vector to be tested of each test data set;
[0009] According to the to-be-tested spatial vector of each of the test data sets and the spatial vector of the target data set, obtaining the vector similarity between each of the test data sets and the target data set;
[0010] Abnormal data is determined from the at least one test data set according to the vector similarity.
[0011] In one embodiment, the acquiring at least one test data set and normalizing the at least one test data set to obtain a normalized parameter set includes:
[0012] A test data set of at least one electrical parameter of a plurality of wafers to be inspected is obtained, and the test data set of at least one electrical parameter is normalized to obtain a normalized parameter set.
[0013] In one embodiment, the step of normalizing the test data set of the at least one electrical parameter to obtain a normalized parameter set includes:
[0014] Determining data characteristics of each electrical parameter based on the test type of the electrical parameter;
[0015] Based on the data features, the test data set is normalized to obtain a normalized parameter set.
[0016] In one embodiment, the method of normalizing the test data set based on the data features to obtain a normalized parameter set further includes:
[0017] Based on the data features, all test values of the test data sets with different electrical parameters are normalized to the same data range.
[0018] In one embodiment, the normalizing the test data set based on the data features to obtain a normalized parameter set includes:
[0019] In a case where the data feature is that the value range of the test data set is greater than a preset value range, sorting the test values in the test data set, and determining a percentile value according to the sorting value of each test value in the test data set;
[0020] The corresponding normalization parameter is determined according to the percentile value corresponding to each test value to obtain a normalization parameter set.
[0021] In one embodiment, the normalizing the test data set based on the data features to obtain a normalized parameter set includes:
[0022] When the data feature is that the number of identical test values in the test data set is greater than a preset number, sorting the test values in the test data set to obtain a sorting value of each test value in the test data set;
[0023] A normalized parameter for each test value is determined based on the maximum value among the ranked values to obtain a normalized parameter set.
[0024] In one embodiment, constructing the space vector to be tested of each test data set according to the normalized parameter set includes:
[0025] Partitioning the wafer to be inspected corresponding to the test data set to obtain a plurality of wafer partitions;
[0026] Determine the characteristic parameters of each wafer partition according to the position intervals corresponding to the plurality of wafer partitions and the target positions corresponding to the test values in the normalized parameter set; the target position is the position of each test value on the wafer to be tested;
[0027] According to the characteristic parameters of each wafer partition, a corresponding space vector to be tested of each test data set is constructed.
[0028] In one embodiment, constructing a corresponding space vector to be tested of each test data set according to the characteristic parameters of each wafer partition includes:
[0029] Performing statistical processing on all normalized test values of each wafer partition to obtain characteristic parameters of each wafer partition;
[0030] According to each of the characteristic parameters, a corresponding space vector to be tested of each of the wafers to be tested is constructed.
[0031] In one embodiment, the partitioning process of the wafer to be inspected corresponding to the test data set to obtain a plurality of wafer partitions includes:
[0032] Determining a partitioning mode according to a preset spatial vector of the target wafer;
[0033] The wafer to be inspected is partitioned according to the partitioning mode to obtain a plurality of wafer partitions.
[0034] In one embodiment, obtaining the vector similarity between each test data set and the target data set according to the to-be-tested spatial vector of each test data set and the preset spatial vector of the target data set includes:
[0035] Performing similarity calculation based on the to-be-tested spatial vector of each of the test data sets and the preset spatial vector of the target data set to obtain the vector space distance between each of the test data sets and the target data set;
[0036] The vector similarity between each of the test data sets and the target data set is determined according to the vector space distance.
[0037] In one embodiment, the method further comprises:
[0038] Determine the corresponding abnormal wafer according to the abnormal data and obtain the corresponding process information;
[0039] The wafer process problem is determined according to the process information and the test data set of the abnormal wafer.
[0040] In a second aspect, the present application further provides a data processing device, comprising:
[0041] A data processing module, used to obtain at least one test data set, and perform normalization processing on the at least one test data set to obtain a normalized parameter set;
[0042] A pooling module, used for constructing a space vector to be tested of each test data set according to the normalization parameter set;
[0043] A similarity determination module, used to obtain the vector similarity between each of the test data sets and the target data set according to the to-be-tested spatial vector of each of the test data sets and the preset spatial vector of the target data set;
[0044] A determination module is used to determine abnormal data from the at least one test data set according to the vector similarity.
[0045] In a third aspect, the present application further provides a chip, including a memory and a processor, wherein the memory stores a computer program, and when the processor executes the computer program, the following steps are implemented:
[0046] Acquire at least one test data set, and perform normalization processing on the at least one test data set to obtain a normalized parameter set;
[0047] According to the normalized parameter set, constructing a space vector to be tested of each test data set;
[0048] According to the to-be-tested spatial vector of each of the test data sets and the spatial vector of the target data set, obtaining the vector similarity between each of the test data sets and the target data set;
[0049] Abnormal data is determined from the at least one test data set according to the vector similarity.
[0050] In a fourth aspect, the present application further provides a computer-readable storage medium having a computer program stored thereon, wherein when the computer program is executed by a processor, the following steps are implemented:
[0051] Acquire at least one test data set, and perform normalization processing on the at least one test data set to obtain a normalized parameter set;
[0052] According to the normalized parameter set, constructing a space vector to be tested of each test data set;
[0053] According to the to-be-tested spatial vector of each of the test data sets and the spatial vector of the target data set, obtaining the vector similarity between each of the test data sets and the target data set;
[0054] Abnormal data is determined from the at least one test data set according to the vector similarity.
[0055] In a fifth aspect, the present application further provides a computer program product, including a computer program, which implements the following steps when executed by a processor:
[0056] Acquire at least one test data set, and perform normalization processing on the at least one test data set to obtain a normalized parameter set;
[0057] According to the normalized parameter set, constructing a space vector to be tested of each test data set;
[0058] According to the to-be-tested spatial vector of each of the test data sets and the spatial vector of the target data set, obtaining the vector similarity between each of the test data sets and the target data set;
[0059] Abnormal data is determined from the at least one test data set according to the vector similarity.
[0060] The above-mentioned wafer data processing method, device, chip, computer-readable storage medium and computer program product obtain a test data set of at least one electrical parameter, normalize the test data set of at least one electrical parameter, and obtain a normalized parameter set, so as to avoid the complex calculation caused by different types and numerical ranges of electrical parameters; at the same time, according to the normalized parameter set, construct the space vector to be tested of each test data set; according to the space vector to be tested of each test data set and the space vector of the target data set, obtain the vector similarity between each test data set and the target data set; determine abnormal data from at least one test data set according to the vector similarity, that is, By converting the test data into spatial vectors and calculating the distance between the spatial vectors, the vector similarity is obtained, and the abnormal data is determined based on the vector similarity. This method does not require model training and is implemented based on a logical rule-based algorithm. It has a small amount of code and consumes very little computing power. It is easy to deploy and implement, as well as to maintain and upgrade, saving computing resources and improving data processing efficiency. Furthermore, in actual production scenarios, results can basically be obtained in seconds. In addition, the rule-based algorithm makes the calculation process explainable. In scenarios where problems occur or updates are required, the relevant parameters that need to be adjusted can be accurately located, and it has a high degree of generalization. BRIEF DESCRIPTION OF THE DRAWINGS
[0061] In order to more clearly illustrate the technical solutions in the embodiments of the present application or related technologies, the drawings required for use in the embodiments of the present application or related technical descriptions will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present application.
[0062] Figure 1 A schematic diagram of an application environment of a data processing method in an embodiment;
[0063] Figure 2 is a flow chart of a data processing method in one embodiment;
[0064] Figure 3 A defect display diagram of a wafer test in one embodiment;
[0065] Figure 4 Schematic diagram of a data normalization method in one embodiment;
[0066] Figure 5 is a schematic flow chart of a method for determining a space vector to be measured in one embodiment;
[0067] Figure 6 is a flow chart of a method for processing characteristic parameters in one embodiment;
[0068] Figure 7 A schematic diagram of a wafer defect on a wafer in one embodiment;
[0069] Figure 8 is a schematic diagram of a wafer defect on a wafer in another embodiment;
[0070] Fig. 9 A schematic diagram of a flow chart of a method for determining vector similarity in one embodiment;
[0071] Fig.10 is a flow chart of a data processing method in another embodiment;
[0072] Fig.11 is a structural block diagram of a wafer data processing device in one embodiment;
[0073] Fig.12 FIG. 4 is a diagram showing the internal structure of a computer device in one embodiment. DETAILED DESCRIPTION
[0074] In order to make the purpose, technical solution and advantages of the present application more clearly understood, the present application is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application and are not used to limit the present application.
[0075] In the description of the embodiments of the present application, “plurality” means more than two, unless otherwise clearly and specifically defined.
[0076] Reference to "embodiments" herein means that a particular feature, structure, or characteristic described in conjunction with the embodiments may be included in at least one embodiment of the present application. The appearance of the phrase in various locations in the specification does not necessarily refer to the same embodiment, nor is it an independent or alternative embodiment that is mutually exclusive with other embodiments. It is explicitly and implicitly understood by those skilled in the art that the embodiments described herein may be combined with other embodiments.
[0077] As the core component of modern electronic devices, chips have a wide range of applications. Chips are not only used in computers and servers, mobile devices, embedded systems and electronic products, but also in navigation systems, radars, communication equipment, etc.
[0078] Chips need to go through a long semiconductor industry manufacturing process to be obtained. For example, it is necessary to go through the process of silicon ingot-purification-cutting-grinding-polishing-lithography and etching-ion implantation and coating-cutting separation to obtain wafers-testing-packaging-chips. Among them, after obtaining the wafer, the basic unit Die (particle) on it must be subjected to various types of tests, such as current and voltage tests, resistance tests, etc., to determine whether it is qualified. The analysis of test results is a very important part in the semiconductor industry. This importance is reflected in two aspects. On the one hand, Die is the basic unit of the chip. If a Die test fails, it will not be used to manufacture chips, but can only be discarded; on the other hand, based on the analysis of the test results, the possible causes of the problems in the semiconductor process can be inferred. If the inferred cause is correct, it will help improve the process, thereby improving the yield of the semiconductor process and reducing the probability of defects in the wafer as much as possible. At present, with the improvement of semiconductor lithography accuracy, it is becoming increasingly difficult to improve product yield. In order to ensure the product yield, it is necessary to identify the faults in the wafer for analysis. Therefore, the wafer needs to be tested.
[0079] In the semiconductor manufacturing process, there are multiple production batches in the production plant. In each production batch, the number of wafers produced is also huge. Moreover, a production plant often has multiple production lines, so the number of wafers produced is very large. Traditionally, deep learning methods are used to detect wafer test patterns. The image recognition algorithm using deep learning is generally executed using convolutional neural networks. The operation of convolutional neural networks requires a lot of computing power, so there are problems such as high computing power resource consumption and low data processing efficiency.
[0080] In order to solve this technical problem, the electrical parameters of the wafer to be inspected are converted into spatial vectors, and the distance between the spatial vectors to be tested is calculated. The vector similarity is determined based on the distance between the preset spatial vector of the target wafer and the spatial vector to be tested, and then the target wafer is determined from multiple wafers to be inspected. Based on the spatial vector, the target wafer can be acquired efficiently, quickly and accurately, saving computing resources and improving data processing efficiency.
[0081] The data processing method provided in the embodiment of the present application can be applied to Figure 1In the application environment shown. Among them, the terminal 102 communicates with the cloud 104 through the network. The data storage system can store the data that the server 104 needs to process. The data storage system can be integrated on the cloud 104. The cloud 104 obtains at least one test data set from the terminal. Normalize at least one test data set to obtain a normalized parameter set; according to the normalized parameter set, construct the test space vector of each test data set; according to the test space vector of each test data set and the space vector of the target data set, obtain the vector similarity between each test data set and the target data set; determine abnormal data from at least one test data set according to the vector similarity. Among them, the terminal 102 can be but is not limited to various personal computers, laptops, smart phones, tablets, Internet of Things devices and portable wearable devices. The Internet of Things devices can be smart speakers, smart TVs, smart air conditioners, smart car devices, projection devices, etc. The cloud 104 can be a cloud server that provides cloud computing services. In an exemplary embodiment, such as Figure 2 As shown, a data processing method is provided. In this embodiment, the method is applied to Figure 1 In this embodiment, the method includes steps 102 to 110, wherein:
[0082] Step 202: Obtain at least one test data set, and perform normalization processing on the at least one test data set to obtain a normalized parameter set.
[0083] The specific data of the test data set can be determined according to the corresponding application scenario. The test data set can be understood as the test data obtained during the wafer testing process, and can also be audio data or medical image data, etc.
[0084] Taking the test data set as an example, the test data obtained during the wafer test process can be understood as the test data obtained, and the wafer test can be but not limited to at least one of resistance test, voltage test and current test. The electrical parameter can be but not limited to at least one of voltage, current and resistance. For example, in the case where the wafer test is a current test, the electrical parameter obtained can be the leakage current of the wafer. In the case where the wafer test is a voltage test, the electrical parameter obtained can be the voltage of the wafer, and the voltage can be the operating voltage. The test method of the wafer can be implemented by existing methods, which will not be repeated here. The test data set can be the test values of different electrical parameters. A test data set of at least one electrical parameter of multiple wafers to be tested is obtained. It can be understood that the electrical performance parameters corresponding to the test data sets of the wafers to be tested can be the same or different. For example, the electrical performance parameters corresponding to the test data sets of three wafers to be tested are all leakage current. For another example, the electrical performance parameters corresponding to the test data sets of three wafers to be tested are voltage, current and resistance, respectively. This scheme can realize data comparison between different wafers or different electrical parameters, so the normalization processing can realize the similarity judgment of various test data sets within the same numerical range.
[0085] It is understandable that there are many production processes in the wafer production process. Wafers need to be tested in different production processes. Wafer tests in different processes can be the same or different. The electrical parameters obtained by different wafer tests are different. A wafer includes multiple wafer basic units, such as bare die. During the test of the wafer, the electrical parameters of each wafer basic unit will be obtained. Therefore, the corresponding test data set can be obtained according to the electrical parameters of each wafer basic unit on each wafer.
[0086] The test data set of the wafer to be inspected can be obtained by taking the production line as a dimension, and obtaining the test data set of the wafer to be inspected of each production line respectively, or by taking the production batch as a dimension, and obtaining the test data set of each batch of wafers to be inspected respectively.
[0087] It is understandable that, since corresponding test data will be obtained in the wafer test, and the types and numerical ranges of electrical parameters of different production processes will be different, in order to ensure the reliability of data processing results, the acquired test data needs to be normalized.
[0088] Normalization processing is to normalize the data to the same preset numerical range. The preset numerical range can be determined according to actual needs. For example, the preset numerical range can be the interval [0, 1]. The specific method of normalization can be any one of minimum-maximum normalization, maximum value normalization, and Z-score normalization. Among them, minimum-maximum normalization can be to determine the maximum and minimum values in the test data set, map the maximum value to 1, and map the minimum value to 0, and linearly map other data in the test data set to [0, 1]. The specific mapping method can be to subtract the minimum value -min from the other data value in the test data set to obtain the difference, and divide the difference by the difference between the maximum value max and the minimum value min, which can be specifically expressed as: value_normalized = (value-min) / (max-min).
[0089] Furthermore, when normalizing different types of electrical parameters, the influences and effects of different types of electrical parameters are different, and the electrical parameters need to be sorted, and the sorting methods include ascending and descending. For electrical parameters sorted in ascending order, the larger value will be mapped to 1 and the smaller value will be mapped to 0 after normalization; for parameters sorted in descending order, the larger value will be mapped to 0 and the smaller value will be mapped to 1 after normalization. For example, electrical parameter 1 is sorted in ascending order, and electrical parameter 2 is sorted in descending order.
[0090] Exemplarily, at least one test data set is obtained, and for each test data set, the maximum value and the minimum value are determined from the test data set, and the maximum value is mapped to 1 and the minimum value is mapped to 0 respectively, and the difference is obtained by subtracting the minimum value -min from the other data values in the test data set, and the difference is divided by the difference between the maximum value max and the minimum value min to obtain a normalized parameter set.
[0091] Step 204: construct a corresponding space vector to be tested for each test data set according to the normalized parameter set.
[0092] The method of constructing the space vector to be tested can be determined by converting each normalized parameter in the normalized parameter set of each test data set into an array element to obtain a corresponding array. The method of constructing the space vector to be tested can also be determined by partitioning the test data set to obtain the normalized parameters of each partition, and constructing the corresponding space vector to be tested according to the normalized parameters of each partition.
[0093] Step 206 , obtaining the vector similarity between each test data set and the target data set according to the to-be-tested spatial vector of each test data set and the spatial vector of the target data set.
[0094] The method for determining the spatial vector of the target data set may also be implemented by the method for determining the spatial vector to be tested, and the method for determining the spatial vector of the target data set is not limited. Calculating the vector similarity may be calculating the vector space distance based on the spatial vector to be tested and the spatial vector of the target data set, and using the obtained vector space distance to represent the similarity between the test data set and the target data set.
[0095] Furthermore, taking the test data set as a wafer test set as an example, since the defect mode on the wafer (or the failure mode of the wafer) may be consistent between similar wafers, the wafer similarity between two wafers can represent the defect similarity (or failure similarity) between the two wafers, and the wafer defects can be displayed in the form of a pattern. Figure 3 As shown in FIG. 1 , a defect display diagram of a wafer test in an exemplary embodiment is shown. The defect at the lower right of the wafer forms a shape similar to a "hat". This defect pattern may mean that there are problems to be solved in a process or several processes upstream of the wafer. If multiple test results "similar" to the "hat" pattern below can be found in other wafers of the same batch or produced by the same manufacturer, the reliability of the above inference (i.e., there are problems to be solved in the process) can be further improved.
[0096] Step 208: determine abnormal data from at least one test data set according to the vector similarity.
[0097] Among them, determining abnormal data based on vector similarity can be to determine the vector similarity within a preset similarity range as the target vector similarity, and determine the test data corresponding to the target vector similarity as abnormal data; or it can be to sort the multiple vector similarities obtained, obtain test data corresponding to a preset number of vector similarities arranged in order, and determine the test data as abnormal data.
[0098] The above data processing method obtains a test data set of at least one electrical parameter, normalizes the test data set of at least one electrical parameter, and obtains a normalized parameter set, thereby avoiding the situation where the calculation is complicated due to different types and different numerical ranges of electrical parameters; at the same time, according to the normalized parameter set, a space vector to be tested of each test data set is constructed; according to the space vector to be tested of each test data set and the space vector of the target data set, the vector similarity between each test data set and the target data set is obtained; abnormal data is determined from at least one test data set according to the vector similarity, that is, by converting the test data into a spatial vector, and by calculating the distance between the spatial vectors, the vector similarity is obtained, and the abnormal data is determined according to the vector similarity. This method does not require model training and is implemented in an algorithm based on logical rules. The code volume is small, the consumption of computing power is extremely small, and it is easy to deploy and implement, and it is also easy to maintain and upgrade, saving computing power resources and improving data processing efficiency. Further, in actual production scenarios, the results can basically be obtained in seconds. In addition, the rule-based algorithm makes the calculation process explainable. In scenarios where problems occur or updates are required, the relevant parameters that need to be adjusted can be accurately located, and it has a high degree of generalization.
[0099] Optionally, in an exemplary embodiment, a test data set of at least one electrical parameter of a plurality of wafers to be inspected is obtained, and the test data set of at least one electrical parameter is normalized to obtain a normalized parameter set.
[0100] The test data set of the electrical parameters includes the data as described above, which is not limited here.
[0101] Exemplarily, a test data set of each of the multiple wafers to be tested from different batches of the same production line is obtained, and for each wafer to be tested in the multiple wafers to be tested, the maximum value and the minimum value are determined from the test data set of each wafer to be tested, and the maximum value is mapped to 1, and the minimum value is mapped to 0, respectively. The difference is obtained by subtracting the minimum value -min from the other data value in the test data set in turn, and the difference is divided by the difference between the maximum value max and the minimum value min to obtain the normalized parameter set of each wafer to be tested. For example, the electrical parameter of wafer 1 to be tested is leakage current, and the electrical parameter of wafer 2 to be tested is voltage. When the electrical parameters of wafers 1 to be tested and wafers 2 to be tested are different, different types of electrical parameters can be normalized to the same numerical range through normalization processing. For different types of electrical parameters, there is no need to perform conversion between the electrical parameters, and different types of electrical parameters only need to be normalized to the same numerical range, which is convenient for calculating similarity.
[0102] Furthermore, after determining the target wafer, the electrical parameters of each wafer basic unit of the target wafer are obtained; based on a preset conversion relationship between the electrical parameters and the pixel values, the electrical parameters of each wafer basic unit are converted to obtain their respective corresponding pixel values; based on the relative positions of each wafer basic unit on the target wafer and the pixel values corresponding to each wafer basic unit, the target wafer is rendered to the display interface according to a preset ratio to obtain a wafer pattern of the target wafer, and based on the wafer pattern, image processing is performed to determine the fault information that caused the wafer pattern.
[0103] Since there are multiple production processes for wafers, the electrical parameters of the wafers determined by each production process may be different. In order to avoid the influence of abnormal values and improve calculation reliability, a data normalization method is provided. In an exemplary embodiment, Figure 4 As shown, step 204 includes steps 402 to 404. Among them:
[0104] Step 402: Determine data characteristics of each electrical parameter based on the test type of the electrical parameter.
[0105] The data feature includes at least one of a value range, the number of identical values in a data set, and a data type. It is understandable that there is a corresponding relationship between the test type and the data feature.
[0106] Step 404: normalize the test data set based on the data features to obtain a normalized parameter set.
[0107] The normalization process may be to normalize all test values of the test data sets of different electrical parameters to the same data range.
[0108] Exemplarily, based on data features, all test values of test data sets with different electrical parameters are normalized to the same data range.
[0109] It is understandable that different data features correspond to different normalization processing methods. Based on the data features, the test data set is normalized to obtain the normalized parameter set in the following two ways:
[0110] Method 1: When the data feature is that the numerical range of the test data set is greater than the preset numerical range, the test values in the test data set are sorted, and the percentile value is determined according to the sort value of each test value in the test data set; the corresponding normalization parameter is determined according to the percentile value corresponding to each test value to obtain a normalization parameter set.
[0111] It is understandable that the numerical range of the test data set is greater than the preset numerical range, indicating that the acquired electrical parameters may be unevenly distributed. In order to ensure the accuracy of the data, percentile values are used to determine the range of data scaling instead of simply using the minimum and maximum values of all data.
[0112] The percentile value can be determined by the percentrank function, that is, given a value, the function returns the percentile rank of this value in the data set. percentrank(value) is a function that calculates the percentile rank.
[0113] Exemplarily, when the data feature is that the numerical range of the test data set is greater than the preset numerical range, the electrical parameters in the test data set of each wafer to be inspected are sorted, and their percentile ranking in the data set is calculated to perform normalization processing to obtain a standardized numerical value, which can be expressed as: value_normalized = percentrank(value), and then the normalized parameter set of each wafer to be inspected can be obtained.
[0114] Method 2: When the data feature is that the number of identical test values in the test data set is greater than a preset number, the test values in the test data set are sorted to obtain a sorted value of each test value in the test data set; a normalized parameter for each test value is determined based on the maximum value in the sorted value to obtain a normalized parameter set.
[0115] The ranking value divided by the number of electrical parameters in the test data set can be expressed as: value_normalized=rank(value) / max(rank(value)).
[0116] It is understandable that when there are a large number of identical values in the test data set, if percentile values are used for normalization, the maximum value may not be guaranteed to be 1. Therefore, the rank of each data value can be divided by the maximum value of all rankings to achieve normalization, ensuring that the maximum value is 1, while dealing with the problem of duplicate values that may exist in the data set.
[0117] In the above embodiment, the data features of the test data set of each wafer to be tested are determined; the corresponding normalization method is determined based on the data features; the test data set is normalized according to the corresponding normalization method to obtain a normalized parameter set, thereby ensuring the reliability of normalization.
[0118] Since a wafer includes multiple wafer basic units, the corresponding space vector to be measured is constructed based on the electrical parameters of all wafer basic units of the wafer. The vector dimension of the space vector to be measured is relatively large, and the amount of calculation is large. In order to reduce the complexity of data processing and ensure the reliability of similarity, a method for determining the space vector to be measured is provided. In an exemplary embodiment, Figure 5 As shown, it includes steps 502 to 506, wherein:
[0119] Step 502 , partitioning the wafer to be inspected corresponding to the test data set to obtain a plurality of wafer partitions.
[0120] Among them, the partitioning method can be any one of axial partitioning, radial partitioning and mask partitioning, or any combination of two or a combination of the three. Axial partitioning can be understood as partitioning along the axial direction of the wafer. Radial partitioning can be understood as partitioning along the radial direction based on the center of the wafer. Mask partitioning can be understood as partitioning control of the distribution of exposure light sources during the photolithography process.
[0121] Step 504 , determining characteristic parameters of each wafer partition according to position intervals corresponding to the plurality of wafer partitions and target positions corresponding to each test value in the normalized parameter set; the target position is the position of each test value on the wafer to be tested.
[0122] Step 506 , constructing a corresponding space vector to be tested for each wafer to be tested according to the characteristic parameters of each wafer partition.
[0123] The characteristic parameter may be determined based on all normalized parameters on each wafer partition, and specifically may be determined by performing statistical processing on all normalized parameters on each wafer partition.
[0124] Exemplarily, according to the position intervals corresponding to multiple wafer partitions and the target positions corresponding to each test value in the normalized parameter set, based on the corresponding relationship between the position intervals and the target positions, and according to the normalized parameter set belonging to the position interval, the characteristic parameters of each wafer partition are determined.
[0125] It can be understood that the characteristic parameters of each wafer partition on each wafer are converted into an array, that is, the space vector to be measured is obtained, wherein the number of wafer partitions is equal to the length of the array, and each value in the array represents the statistical value (i.e., characteristic parameter) of the corresponding partition.
[0126] In an exemplary embodiment, Figure 6 As shown, a method for processing characteristic parameters is provided, including steps 602 to 604, wherein:
[0127] Step 602 , statistically process all normalized test values of each wafer partition to obtain characteristic parameters of each wafer partition.
[0128] The statistical processing may be various statistical algorithms such as the average value, median, maximum value, minimum value, and any combination thereof. It is understandable that the determination of the average value, median, maximum value, and minimum value can be achieved by existing methods, which will not be elaborated here.
[0129] Step 604 : constructing a corresponding space vector to be tested for each wafer to be tested according to each characteristic parameter.
[0130] The space vector to be measured is constructed by converting each characteristic parameter into an array, and the placement order of each element in the array may be the same.
[0131] Exemplarily, a space vector to be tested is constructed according to characteristic parameters of all wafer partitions of each wafer to be tested.
[0132] In the above embodiment, by performing statistical processing on the normalized parameters of each wafer partition, a corresponding space vector to be tested of each wafer to be tested is constructed, thereby ensuring the accuracy of similarity calculation.
[0133] In the above embodiment, each wafer is partitioned to obtain normalized parameters of each partition, and for each wafer to be tested, the test values of each partition of each wafer to be tested are statistically analyzed to obtain characteristic parameters, and the characteristic parameters of each partition of each wafer to be tested are converted into an array, and the similarity between two wafers to be tested is detected based on the array to reduce the amount of data processing.
[0134] It is understandable that there is a certain relationship between the similarity between the wafers to be tested and the partitioning mode of the wafers to be tested. In an exemplary embodiment, the wafers to be tested corresponding to the test data set are partitioned to obtain multiple wafer partitions, including:
[0135] A partitioning pattern is determined according to a preset spatial vector of the target wafer; and a wafer to be inspected is partitioned according to the partitioning pattern to obtain a plurality of wafer partitions.
[0136] Among them, the target image can be understood as the manifestation of wafer defects, for example Figure 7 and Figure 8As shown, they are schematic diagrams of wafer defects on a wafer in an example. It is understandable that in actual application scenarios, due to the diversity of wafer test data, the test results of wafers are expressed in different forms. For example, if the test data is expressed in binary, then the wafer pattern can be a black and white grayscale image, where the white part represents that the basic unit test of the wafer has passed, and the black part represents that the test has failed and there is a fault. If the test result is an analog quantity, the wafer pattern can also be displayed in the form of a color pattern.
[0137] The partitioning mode is determined according to the preset spatial vector of the target wafer, and each wafer to be inspected is partitioned to obtain multiple wafer partitions. The actual wafer area where the target pattern on the target wafer is located can be used as the starting point, and each wafer to be inspected can be divided into multiple wafer partitions according to the size of the actual wafer area. Through this partitioning process, potential problems on the wafer can be more accurately identified and located.
[0138] In an exemplary embodiment, Fig. 9 As shown, a method for determining vector similarity is provided, including steps 902 to 904, wherein:
[0139] Step 902 , performing similarity calculation based on the to-be-tested spatial vector of each test data set and the preset spatial vector of the target data set, to obtain the vector space distance between each test data set and the target data set.
[0140] Among them, the vector space distance represents the similarity of the wafers. The smaller the space distance, the higher the similarity of the two sets of data, that is, the more similar the patterns / modes of the two images are. Similarity calculation methods include Euler distance, Pearson correlation coefficient, cosine similarity, etc.
[0141] Exemplarily, the Euler distance is calculated based on the space vector to be tested of each test data set and the preset space vector of the target data set, and the vector space distance between each test data set and the target wafer is obtained based on the calculated Euler distance.
[0142] Further, in an exemplary embodiment, according to the space vector to be tested of each test data set and the preset space vector of the target wafer, the Euler distance is calculated to obtain the Euler distance, and then the Euler distance is weighted to obtain the vector space distance between each wafer to be tested and the target wafer. The weighted calculation can be implemented in an existing manner, and the weighted calculation fully considers the electrical parameters of the wafer, ensuring the reliability of the vector similarity.
[0143] Step 904 : determining the vector similarity between each test data set and the target data set according to the vector space distance.
[0144] In this embodiment, by calculating vector similarity and determining the vector similarity between each test data set and the target wafer based on the vector similarity, it is possible to screen out abnormal data. Furthermore, in the wafer data test, wafers to be tested with preset patterns can be screened out.
[0145] On the basis of determining the similarity, it is necessary to analyze the problems in the wafer production process, and then it is necessary to screen out the target wafers. In an exemplary embodiment, the target wafer is determined from multiple wafers to be detected based on multiple vector similarities, including: sorting the vector similarities to obtain sorted vector similarities; and determining a preset number of target wafers arranged in order from multiple wafers to be detected based on the sorted vector similarities. The preset number is determined based on actual needs, for example, it can be 10 or 20.
[0146] It is understandable that the wafer process problem needs to be analyzed based on the acquired similar wafer test data. In an exemplary embodiment, after determining the abnormal data, the following operations may also be performed:
[0147] Determine the corresponding abnormal wafer based on the abnormal data and obtain the corresponding process information; determine the wafer process problem based on the process information and the test data set of the abnormal wafer.
[0148] Among them, determining abnormal data from at least one test data set based on multiple vector similarities can be done by sorting the vector similarities in descending order according to the magnitude of the vector similarities, obtaining the sorted vector similarities, and obtaining the first N determined from the sorted vector similarities, where the number of N can be determined based on actual needs and is not limited here. The corresponding process information can be obtained based on the abnormal wafer, and the process information includes production information of each link of wafer production, such as process parameters and equipment parameters of production equipment.
[0149] By performing fault analysis on the wafer pattern of the target wafer according to the production information, the wafer production problem can be obtained, and the problems that may arise in the process can be solved based on the wafer production problem, thereby effectively improving the yield of semiconductor manufacturing.
[0150] In an exemplary embodiment, Fig.10 As shown, a wafer data processing method is provided, including steps 1002 to 1016, wherein:
[0151] Step 1002 , obtaining a test data set of at least one electrical parameter of a plurality of wafers to be tested.
[0152] Step 1004 , normalizing the test data set of at least one electrical parameter to obtain a normalized parameter set.
[0153] Step 1006 , partitioning the wafer to be inspected corresponding to the test data set to obtain a plurality of wafer partitions.
[0154] Step 1008, determining characteristic parameters of each wafer partition according to position intervals corresponding to the plurality of wafer partitions and target positions corresponding to each test value in the normalized parameter set; the target position is the position of each test value on the wafer to be tested.
[0155] Step 1010: construct a corresponding space vector to be tested for each wafer to be tested according to the characteristic parameters of each wafer partition.
[0156] Step 1012: Obtain the vector similarity between each test data set and the target data set according to the to-be-tested spatial vector of each test data set and the spatial vector of the target data set.
[0157] Step 1014 , obtaining corresponding production information according to the wafers to be inspected corresponding to the multiple vector similarities.
[0158] Step 1016, performing a fault analysis on the wafer pattern of the wafer to be inspected based on the production information to determine the wafer production problem.
[0159] It is understandable that the normalized parameter set, the space vector to be measured and the vector similarity can be implemented in the manner described in the above embodiments, which will not be described in detail here.
[0160] The above wafer data processing method converts wafer test data into space vectors, calculates the distance between space vectors, and determines the pattern similarity of different wafers under test conditions based on the distance between space vectors. Thus, wafers with similar test characteristics can be obtained efficiently, quickly and accurately, and problems in the production process can be further inferred based on these wafers. Therefore, based on this method, the yield rate in the semiconductor manufacturing field can be effectively improved.
[0161] It should be understood that, although the various steps in the flowcharts involved in the above-mentioned embodiments are displayed in sequence according to the indication of the arrows, these steps are not necessarily executed in sequence according to the order indicated by the arrows. Unless there is a clear explanation in this article, the execution of these steps does not have a strict order restriction, and these steps can be executed in other orders. Moreover, at least a part of the steps in the flowcharts involved in the above-mentioned embodiments can include multiple steps or multiple stages, and these steps or stages are not necessarily executed at the same time, but can be executed at different times, and the execution order of these steps or stages is not necessarily to be carried out in sequence, but can be executed in turn or alternately with other steps or at least a part of the steps or stages in other steps.
[0162] Based on the same inventive concept, the embodiment of the present application also provides a wafer data processing device for implementing the wafer data processing method involved above. The implementation scheme for solving the problem provided by the device is similar to the implementation scheme recorded in the above method, so the specific limitations in one or more wafer data processing device embodiments provided below can refer to the limitations of the wafer data processing method above, and will not be repeated here.
[0163] In an exemplary embodiment, Fig.11 As shown, a wafer data processing device is provided, including: a data processing module 1102, a pooling module 1104, a similarity determination module 1106 and a determination module 1108, wherein:
[0164] The data processing module 1102 is used to obtain at least one test data set, perform normalization processing on the at least one test data set, and obtain a normalization parameter set.
[0165] The pooling module 1104 is used to construct the space vector to be tested of each test data set according to the normalization parameter set.
[0166] The similarity determination module 1106 is used to obtain the vector similarity between each of the test data sets and the target data set according to the to-be-tested spatial vector of each of the test data sets and the preset spatial vector of the target data set.
[0167] The determination module 1108 is configured to determine abnormal data from the at least one test data set according to the vector similarity.
[0168] The above-mentioned data processing device obtains a test data set of at least one electrical parameter, normalizes the test data set of at least one electrical parameter, and obtains a normalized parameter set, thereby avoiding the situation where the calculation is complicated due to different types and different numerical ranges of electrical parameters; at the same time, according to the normalized parameter set, a space vector to be tested of each test data set is constructed; according to the space vector to be tested of each test data set and the space vector of the target data set, the vector similarity between each test data set and the target data set is obtained; abnormal data is determined from at least one test data set according to the vector similarity, that is, by converting the test data into a spatial vector, and by calculating the distance between the spatial vectors, the vector similarity is obtained, and the abnormal data is determined according to the vector similarity. This method does not require model training and is implemented in an algorithm based on logical rules. The code volume is small, the consumption of computing power is extremely small, it is easy to deploy and implement, and it is also easy to maintain and upgrade, saving computing power resources and improving data processing efficiency; further, in actual production scenarios, the results can basically be obtained in seconds. In addition, the rule-based algorithm makes the calculation process explainable. In scenarios where problems occur or updates are required, the relevant parameters that need to be adjusted can be accurately located, and it has a high degree of generalization.
[0169] Optionally, in an exemplary embodiment, the data processing module 1102 is further used to obtain a test data set of at least one electrical parameter of multiple wafers to be inspected, and normalize the test data set of the at least one electrical parameter to obtain a normalized parameter set.
[0170] Optionally, in an exemplary embodiment, the data processing module 1102 includes a normalization processing module, and the normalization processing module is used to determine the data characteristics of each electrical parameter based on the test type of the electrical parameter;
[0171] Based on the data characteristics, the test data set is normalized to obtain a normalized parameter set.
[0172] Optionally, in an exemplary embodiment, the normalization processing module is further used to normalize all test values of the test data sets of different electrical parameters to the same data range based on data features.
[0173] Optionally, in an exemplary embodiment, the normalization processing module is further used to sort the test values in the test data set when the data feature is that the value range of the test data set is greater than the preset value range, and determine the percentile value according to the sorting value of each test value in the test data set;
[0174] The corresponding normalization parameter is determined according to the percentile value corresponding to each test value to obtain a normalization parameter set.
[0175] Optionally, in an exemplary embodiment, the normalization processing module is further used to sort the test values in the test data set when the data feature is that the number of identical test values in the test data set is greater than a preset number, to obtain a sort value of each test value in the test data set;
[0176] A normalized parameter for each test value is determined based on the maximum value among the sorted values to obtain a normalized parameter set.
[0177] Optionally, in an exemplary embodiment, the wafer data processing device further includes a partitioning module, and the partitioning module is used to perform partitioning processing on the wafer to be inspected corresponding to the test data set to obtain a plurality of wafer partitions.
[0178] Optionally, in an exemplary embodiment, the pooling module 1104 is further used to determine the characteristic parameters of each wafer partition according to the position intervals corresponding to the multiple wafer partitions and the target positions corresponding to the test values in the normalized parameter set; the target position is the position of each test value on the wafer to be detected;
[0179] According to the characteristic parameters of each wafer partition, a corresponding space vector to be tested for each wafer to be tested is constructed.
[0180] Optionally, in an exemplary embodiment, the partitioning module is used to determine the partitioning mode according to a preset space vector of the target wafer;
[0181] The wafer to be inspected is partitioned according to the partitioning mode to obtain multiple wafer partitions.
[0182] Optionally, in an exemplary embodiment, the similarity determination module 1106 is used to perform similarity calculation based on the to-be-tested spatial vector of each test data set and the spatial vector of the target data set to obtain the vector space distance between each test data set and the target data set;
[0183] The vector similarity between each test dataset and the target dataset is determined based on the vector space distance.
[0184] Optionally, in an exemplary embodiment, the pooling module 1104 is further used to perform statistical processing on all normalized test values of each wafer partition to obtain characteristic parameters of each wafer partition;
[0185] According to each characteristic parameter, a corresponding space vector to be tested for each test data set is constructed.
[0186] Optionally, in an exemplary embodiment, the data processing device further includes a fault analysis module, which is used to obtain corresponding production information according to the wafers to be inspected corresponding to the multiple vector similarities;
[0187] Perform fault analysis on the wafer pattern of the wafer to be inspected based on the production information to identify wafer production problems.
[0188] Each module in the above-mentioned wafer data processing device can be implemented in whole or in part by software, hardware and a combination thereof. Each of the above-mentioned modules can be embedded in or independent of the processor in the computer device in the form of hardware, or can be stored in the memory of the computer device in the form of software, so that the processor can call and execute the operations corresponding to each of the above modules.
[0189] In an exemplary embodiment, a computer device is provided. The computer device may be a terminal, and its internal structure diagram may be as shown in FIG. Fig.12 As shown. The computer device includes a processor, a memory, an input / output interface, a communication interface, a display unit and an input device. Among them, the processor, the memory and the input / output interface are connected through a system bus, and the communication interface, the display unit and the input device are connected to the system bus through the input / output interface. Among them, the processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and a computer program. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The input / output interface of the computer device is used to exchange information between the processor and the external device. The communication interface of the computer device is used to communicate with an external terminal in a wired or wireless manner, and the wireless manner can be implemented through WIFI, a mobile cellular network, near field communication (Near Field Communication, NFC) or other technologies. When the computer program is executed by the processor, a wafer data processing method is implemented. The display unit of the computer device is used to form a visually visible picture, which can be a display screen, a projection device or a virtual reality imaging device. The display screen can be a liquid crystal display screen or an electronic ink display screen, and the input device of the computer device can be a touch layer covering the display screen, or a button, trackball or touchpad set on the computer device shell, or an external keyboard, touchpad or mouse.
[0190] Those skilled in the art will understand that Fig.12 The structure shown in the figure is only a block diagram of a part of the structure related to the solution of the present application, and does not constitute a limitation on the computer device to which the solution of the present application is applied. The specific computer device may include more or fewer components than those shown in the figure, or combine certain components, or have a different arrangement of components.
[0191] In one embodiment, a chip is further provided, including a memory and a processor, wherein a computer program is stored in the memory, and the processor implements the steps in the above-mentioned method embodiments when executing the computer program.
[0192] In one embodiment, a computer-readable storage medium is provided, on which a computer program is stored. When the computer program is executed by a processor, the steps in the above-mentioned method embodiments are implemented.
[0193] In one embodiment, a computer program product is provided, including a computer program, which implements the steps in the above method embodiments when executed by a processor.
[0194] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, stored data, displayed data, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties, and the collection, use and processing of relevant data must comply with relevant regulations.
[0195] Those of ordinary skill in the art can understand that all or part of the processes in the above-mentioned embodiment methods can be completed by instructing the relevant hardware through a computer program, and the computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to the memory, database or other medium used in the embodiments provided in the present application can include at least one of non-volatile memory and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetic random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. As an illustration and not limitation, RAM can be in various forms, such as static random access memory (SRAM) or dynamic random access memory (DRAM). The database involved in each embodiment provided in this application may include at least one of a relational database and a non-relational database. Non-relational databases may include distributed databases based on blockchains, etc., but are not limited to this. The processor involved in each embodiment provided in this application may be a general-purpose processor, a central processing unit, a graphics processor, a digital signal processor, a programmable logic device, a data processing logic device based on quantum computing, an artificial intelligence (AI) processor, etc., but are not limited to this.
[0196] The technical features of the above embodiments may be combined arbitrarily. To make the description concise, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this application.
[0197] The above-described embodiments only express several implementation methods of the present application, and the descriptions thereof are relatively specific and detailed, but they cannot be understood as limiting the scope of the present application. It should be pointed out that, for a person of ordinary skill in the art, several variations and improvements can be made without departing from the concept of the present application, and these all belong to the protection scope of the present application. Therefore, the protection scope of the present application shall be subject to the attached claims.
Claims
1. A data processing method, characterized in that: The method comprises: Acquire at least one test data set, and perform normalization processing on the at least one test data set to obtain a normalized parameter set; According to the normalized parameter set, constructing a space vector to be tested of each test data set; According to the to-be-tested spatial vector of each of the test data sets and the spatial vector of the target data set, obtaining the vector similarity between each of the test data sets and the target data set; Abnormal data is determined from the at least one test data set according to the vector similarity.
2. The method according to claim 1, characterized in that The acquiring of at least one test data set and normalizing the at least one test data set to obtain a normalized parameter set includes: A test data set of at least one electrical parameter of a plurality of wafers to be inspected is obtained, and the test data set of at least one electrical parameter is normalized to obtain a normalized parameter set.
3. The method according to claim 2, characterized in that The step of normalizing the test data set of the at least one electrical parameter to obtain a normalized parameter set includes: Determining data characteristics of each electrical parameter based on the test type of the electrical parameter; Based on the data features, the test data set is normalized to obtain a normalized parameter set.
4. The method according to claim 3, characterized in that: The method of normalizing the test data set based on the data features to obtain a normalized parameter set further includes: Based on the data features, all test values of the test data sets with different electrical parameters are normalized to the same data range.
5. The method according to claim 3, characterized in that: The step of normalizing the test data set based on the data features to obtain a normalized parameter set includes: In a case where the data feature is that the value range of the test data set is greater than a preset value range, sorting the test values in the test data set, and determining a percentile value according to the sorting value of each test value in the test data set; The corresponding normalization parameter is determined according to the percentile value corresponding to each test value to obtain a normalization parameter set.
6. The method according to claim 3, characterized in that The step of normalizing the test data set based on the data features to obtain a normalized parameter set includes: When the data feature is that the number of identical test values in the test data set is greater than a preset number, sorting the test values in the test data set to obtain a sorting value of each test value in the test data set; A normalized parameter for each test value is determined based on the maximum value among the ranked values to obtain a normalized parameter set.
7. The method according to claim 2, characterized in that The step of constructing a space vector to be tested of each test data set according to the normalized parameter set includes: Partitioning the wafer to be inspected corresponding to the test data set to obtain a plurality of wafer partitions; Determine the characteristic parameters of each wafer partition according to the position intervals corresponding to the plurality of wafer partitions and the target positions corresponding to the test values in the normalized parameter set; the target position is the position of each test value on the wafer to be tested; According to the characteristic parameters of each wafer partition, a corresponding space vector to be tested of each test data set is constructed.
8. The method according to claim 7, characterized in that The step of constructing a corresponding space vector to be tested of each test data set according to the characteristic parameters of each wafer partition includes: Performing statistical processing on all normalized test values of each wafer partition to obtain characteristic parameters of each wafer partition; According to each of the characteristic parameters, a corresponding space vector to be tested of each of the wafers to be tested is constructed.
9. The method according to claim 7, characterized in that: The partitioning process of the wafer to be inspected corresponding to the test data set to obtain a plurality of wafer partitions includes: Determining a partitioning mode according to a preset spatial vector of the target wafer; The wafer to be inspected is partitioned according to the partitioning mode to obtain a plurality of wafer partitions.
10. The method according to claim 1, characterized in that The obtaining, according to the to-be-tested spatial vector of each of the test data sets and the preset spatial vector of the target data set, the vector similarity between each of the test data sets and the target data set comprises: Performing similarity calculation based on the to-be-tested spatial vector of each of the test data sets and the preset spatial vector of the target data set to obtain the vector space distance between each of the test data sets and the target data set; The vector similarity between each of the test data sets and the target data set is determined according to the vector space distance.
11. The method according to claim 1, characterized in that: The method further comprises: Determine the corresponding abnormal wafer according to the abnormal data and obtain the corresponding process information; The wafer process problem is determined according to the process information and the test data set of the abnormal wafer.
12. A data processing device, characterized in that: The device comprises: A data processing module, used to obtain at least one test data set, and perform normalization processing on the at least one test data set to obtain a normalized parameter set; A pooling module, used for constructing a space vector to be tested of each test data set according to the normalization parameter set; A similarity determination module, used to obtain the vector similarity between each of the test data sets and the target data set according to the to-be-tested spatial vector of each of the test data sets and the preset spatial vector of the target data set; A determination module is used to determine abnormal data from the at least one test data set according to the vector similarity.
13. A chip comprising a memory and a processor, wherein the memory stores a computer program, characterized in that: When the processor executes the computer program, the steps of the method according to any one of claims 1 to 11 are implemented.
14. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 11 are implemented.
15. A computer program product comprising a computer program, characterized in that When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 11 are implemented.