Chip failure type data classification method and device, storage medium and terminal

Abstract

This invention discloses a data classification method, apparatus, storage medium, and terminal for chip failure types. The method includes acquiring design information of the chip under test (DUT), and based on the design information, acquiring the DUT's size information, logical location-to-physical location correspondence information, and failure type information; acquiring test data of the DUT, and converting the test data into test physical location information based on the size information and logical location-to-physical location correspondence information; clustering the failure positions of the DUT using an optimized clustering algorithm based on the test physical location information to obtain failure clustering results; and classifying the failure positions of the DUT based on the failure type information and the failure clustering results to obtain failure classification results for the DUT. This method improves the efficiency of clustering operations by optimizing the clustering algorithm, saves memory occupied by the operation, and greatly improves the efficiency of chip failure type acquisition; it also improves data reading and saving efficiency and saves storage space.

Description

Technical Field

[0001] This invention relates to the field of chip failure data analysis technology, and in particular to a data classification method, apparatus, storage medium, and terminal for chip failure types. Background Technology

[0002] Static Random-Access Memory (SRAM) is a type of random access memory. The physical mechanisms behind SRAM failures form specific patterns at the data level. By classifying different failure modes, the underlying causes of failures can be identified.

[0003] To obtain information about chip failure mechanisms, it is necessary to locate the position of each bit on the failed chip and determine the failure of each bit. For memory chips, since they are laid out as arrays on a silicon wafer, analyzing the chip's physical location data can determine the specific location of the failed bits. Existing chip failure bit data classification and analysis methods cluster the original data horizontally and vertically, and then merge the two clustering results. This process generates three matrices, which are then repeatedly used to run the classification algorithm. This analysis method is slow in data reading and processing, consumes a large amount of storage space, has a slow classification algorithm, high memory consumption, and is inefficient due to repeatedly using three matrices to run the algorithm. Summary of the Invention

[0004] The technical problem to be solved by this invention is that existing data classification methods for chip failure types have slow data reading and processing speeds, large storage space requirements, slow running speeds of classification algorithms, high memory usage, and low efficiency.

[0005] To address the aforementioned technical problems, this invention provides a data classification method for chip failure types, comprising:

[0006] Obtain the design information of the chip under test, and based on the design information, obtain the size information, logical location and physical location correspondence information, and failure type information of the chip under test;

[0007] The test data of the chip under test is obtained, and the test data is converted into test physical location information based on the size information and the correspondence between the logical position and the physical position.

[0008] Based on the test physical location information, the failure bits in the chip under test are clustered by an optimized clustering algorithm to obtain the failure clustering results.

[0009] Based on the failure type information and the failure clustering results, the failure bits of the chip under test are classified to obtain the failure classification results of the chip under test.

[0010] The optimized clustering algorithm includes:

[0011] Use the Nth row / column position in the target block as the target row / column position;

[0012] The failure positions in the target row / column points are clustered using a one-dimensional clustering algorithm to obtain one-dimensional clustering results;

[0013] The adjacent cluster types in the one-dimensional clustering results and the latest defined temporary clustering results are merged to obtain the target clustering result;

[0014] Determine if N equals the first preset threshold. If yes, use the latest acquired target clustering result as the block failure clustering result; otherwise, increment N by 1, define the latest acquired target clustering result as a temporary clustering result, and re-use the Nth row / column position as the target row / column position.

[0015] The initial value of N is 1, and when N is 1, the latest defined temporary clustering result is none. The chip under test includes multiple blocks, and each block includes multiple rows or columns of points. The first preset threshold is the number of rows or columns of points in the block.

[0016] Preferably, converting the test data into test physical location information based on the size information and the logical position-to-physical position correspondence information includes:

[0017] Convert the test data into matrix format data;

[0018] Based on the size information and the correspondence between the logical position and the physical position, the matrix format data is converted into test physical position information.

[0019] Preferably, the step of clustering the failure bits of the chip under test using an optimized clustering algorithm based on the test physical location information to obtain the failure clustering results includes:

[0020] Sort all blocks in the chip under test to obtain the block sorting result;

[0021] Take the Mth block in the block sorting result as the target block;

[0022] The clustering algorithm is optimized to cluster the invalid bits in the target block to obtain the block clustering result of the Mth block;

[0023] Determine whether M is equal to the preset sorting value. If so, the block clustering results of all blocks form the failure clustering results of the chip under test. Otherwise, increment M by 1 and re-select the Mth block in the block sorting results as the target block.

[0024] Wherein, the initial value of M is 1, and the preset sorting value is the total number of blocks in the chip under test.

[0025] Preferably, the step of classifying the failure bits of the chip under test based on the failure type information and the failure clustering results to obtain the failure classification results of the chip under test includes:

[0026] The clustering types in each block of the chip under test are classified to obtain the block classification result of each block. The block classification results of all the blocks form the failure classification result of the chip under test.

[0027] The process of classifying cluster types within a single block includes:

[0028] Determine whether the total number of failure bits in all cluster types in the block is greater than the first preset number. If so, the block is determined to be a block failure mode. Otherwise, determine whether all cluster types in the block meet the linear failure setting and set the cluster types that meet the linear failure setting as linear failure mode. Then, determine whether the remaining cluster types in the block meet the point failure setting and set the cluster types that meet the point failure setting as point failure mode, and set the cluster types that do not meet the point failure setting as other failure modes.

[0029] The linear failure is defined as the number of failure positions in the cluster type being greater than a second preset number, and the point failure is defined as the cluster type having only one failure position, having two consecutive failure positions, or having four adjacent failure positions forming a square.

[0030] Preferably, the data classification method for chip failure types further includes:

[0031] Based on the size information of the chip under test and the classification result, the classification result is displayed as an image to obtain a failure classification display diagram of the chip under test.

[0032] Preferably, when a block is determined to be a block failure mode, the density distribution of the failure bits in the block is calculated, and the block is represented in the form of a density distribution map in the failure classification display diagram of the chip under test.

[0033] Preferably, the data classification method for chip failure types further includes:

[0034] The test data is stored in the database using the sparse matrix principle. Each classification result in the failure classification results is indexed to the corresponding data in the test data, and the indexed failure classification results are stored in the database.

[0035] To address the aforementioned technical problems, the present invention also provides a data classification device for chip failure types, comprising an information acquisition module, a test physical location information acquisition module, a failure clustering result acquisition module, and a failure classification result acquisition module;

[0036] The information acquisition module is used to acquire the design information of the chip under test, and based on the design information, acquire the size information, logical position and physical position correspondence information, and failure type information of the chip under test.

[0037] The test physical location information acquisition module is used to acquire the test data of the chip under test, and convert the test data into test physical location information based on the size information and the logical position to physical position correspondence information;

[0038] The failure clustering result acquisition module is used to cluster the failure positions in the chip under test based on the test physical location information using an optimized clustering algorithm to obtain failure clustering results.

[0039] The failure classification result acquisition module is used to classify the failure positions of the chip under test based on the failure type information and the failure clustering result, so as to obtain the failure classification result of the chip under test.

[0040] The optimized clustering algorithm includes:

[0041] Use the Nth row / column position in the target block as the target row / column position;

[0042] The failure positions in the target row / column points are clustered using a one-dimensional clustering algorithm to obtain one-dimensional clustering results;

[0043] The adjacent clustering types in the one-dimensional clustering result and the latest defined temporary clustering result are merged to obtain the target clustering result;

[0044] Determine if N equals the first preset threshold. If yes, use the latest acquired target clustering result as the block failure clustering result; otherwise, increment N by 1, define the latest acquired target clustering result as a temporary clustering result, and re-use the Nth row / column position as the target row / column position.

[0045] The initial value of N is 1, and when N is 1, the latest defined temporary clustering result is none. The chip under test includes multiple blocks, and each block includes multiple rows or columns of points. The first preset threshold is the number of rows or columns of points in the block.

[0046] To address the aforementioned technical problems, the present invention also provides a storage medium storing a computer program that, when executed by a processor, implements a data classification method for chip failure types.

[0047] To address the aforementioned technical problems, the present invention also provides a terminal, comprising: a processor and a memory, wherein the memory and the processor are communicatively connected;

[0048] The memory is used to store computer programs, and the processor is used to execute the computer programs stored in the memory to enable the terminal to perform a data classification method such as chip failure type.

[0049] Compared with the prior art, one or more embodiments of the above solutions may have the following advantages or beneficial effects:

[0050] The chip failure type data classification method provided in this invention improves clustering efficiency and saves memory by optimizing the clustering algorithm. Chip failure location classification based on clustering results significantly improves the efficiency of chip failure type acquisition. The block failure model is displayed in the failure classification diagram of the chip under test using a density distribution, which more accurately reflects the actual data. The use of a sparse matrix concept combined with failure mode storage classification results and original data improves data reading and saving efficiency and saves storage space. Further optimization of the clustering algorithm, performing clustering only horizontally or vertically, reduces memory usage. Furthermore, merging adjacent cluster types during clustering reduces the number of calculations and the use of process variables, improving overall computational efficiency.

[0051] Other features and advantages of the invention will be set forth in the description which follows, and will be apparent in part from the description, or may be learned by practicing the invention. The objects and other advantages of the invention may be realized and obtained by means of the structures particularly pointed out in the description, claims, and drawings. Attached Figure Description

[0052] The accompanying drawings are provided to further illustrate the invention and form part of the specification. They are used in conjunction with the embodiments of the invention to explain the invention and do not constitute a limitation thereof. In the drawings:

[0053] Figure 1 A flowchart illustrating the data classification method for chip failure types according to Embodiment 1 of the present invention is shown.

[0054] Figure 2 An example diagram of the failure classification display diagram in Embodiment 1 of the present invention is shown;

[0055] Figure 3 A schematic diagram of the structure of the data classification device for chip failure types according to Embodiment 2 of the present invention is shown;

[0056] Figure 4 A schematic diagram of the terminal structure of Embodiment 4 of the present invention is shown. Detailed Implementation

[0057] The embodiments of the present invention will be described in detail below with reference to the accompanying drawings and examples, so that the process of how the present invention uses technical means to solve technical problems and achieve technical effects can be fully understood and implemented accordingly. It should be noted that, as long as there is no conflict, the various embodiments and features in the various embodiments of the present invention can be combined with each other, and the resulting technical solutions are all within the protection scope of the present invention.

[0058] The physical mechanisms underlying Static Random Access Memory (SRAM) failures form specific patterns at the data level. By classifying different failure modes, the underlying causes of failures can be identified. However, processing chip data faces challenges such as excessive data volume, large storage space requirements, and slow analysis speed. Current data classification methods for chip failure types store raw data directly in a database, resulting in slow data reading and processing speeds and significant storage space consumption. Horizontal and vertical clustering are performed separately to obtain two corresponding matrices, which are then merged. This process is slow and memory-intensive. Finally, the three matrices are repeatedly used during computation, leading to low computational efficiency.

[0059] Example 1

[0060] To address the technical problems existing in the prior art, embodiments of the present invention provide a data classification method for chip failure types.

[0061] Figure 1 This diagram illustrates a flowchart of a data classification method for chip failure types according to Embodiment 1 of the present invention; see reference. Figure 1 As shown, the data classification method for chip failure types in this embodiment of the invention includes the following steps.

[0062] Step S101: Obtain the design information of the chip under test, and obtain the size information, logical location and physical location correspondence information, and failure type information of the chip under test based on the design information.

[0063] Specifically, the design information of the chip under test (DUT) is obtained through downloading or other means. This design information includes chip size information, logical location-to-physical location correspondence information, and the chip's failure type information. Further, the chip size information includes the chip dimensions, the positional dimensions of each block within the chip, and the horizontal and vertical addresses of each point within each block. The logical location-to-physical location correspondence information includes the relationship between the logical and physical locations of the DUT, primarily used to determine the physical location of chip data on the chip based on the logical location information of the chip data. The failure type information includes failure modes and corresponding parameters; specific failure modes include block failure modes, linear failure modes, point failure settings, and other failure modes. The block failure mode parameters are the block failure setting definition and a first preset number of parameters; the linear failure mode parameters are the linear failure setting and a second preset number of parameters; and the point failure model parameters are the point failure model failure settings. Therefore, based on the design information of the DUT, information such as the chip size, logical location-to-physical location correspondence information, and failure type information can be obtained.

[0064] Step S102: Obtain the test data of the chip under test, and convert the test data into test physical location information based on the size information and the correspondence between logical and physical locations.

[0065] Specifically, to obtain the chip failure type, the chip must first be tested, and then the test data must be acquired as the data basis for analyzing the chip failure mode. Therefore, the chip under test (DUT) needs to be tested to obtain its test data. The DUT's test data represents the logical location data indicating the failure. The acquired test data is usually in text document format. For ease of processing, the test data needs to be converted to a standard format, specifically a matrix format. Further constraints on the standard format can be determined based on the actual situation. Then, based on the DUT's size information and the correspondence between logical and physical locations, each logical point in the test data is converted into its corresponding physical point, thereby obtaining the DUT's physical location information.

[0066] Step S103: Based on the test physical location information, the failure bits of the chip under test are clustered using an optimized clustering algorithm to obtain failure clustering results.

[0067] A chip typically comprises multiple blocks, each containing multiple points. If the chip under test (DUT) is a faulty chip, the points within it are further categorized into faulty and non-faulty points. Specifically, clustering the faulty points in the DUT includes the following steps: S31, sorting all blocks in the DUT to obtain a block sorting result. S32, selecting the Mth block from the sorted block results as the target block; where M is initially 1 and is a positive integer. S33, clustering the target block using an optimized clustering algorithm to obtain the clustering result for the Mth block. S34, determining if M equals the preset sorting value; if so, the clustering result of all blocks in the DUT forms the faulty chip clustering result; otherwise, incrementing M by 1 and proceeding to step S31 to cluster the next block in the DUT. Here, M is initially 1 and is a positive integer, and the preset sorting value is the total number of blocks in the DUT.

[0068] To simplify the computational workload of failure bit statistics, we first use an optimized clustering algorithm to cluster each block in the chip under test, and then classify the failure bits of the chip under test based on the clustering results. This greatly simplifies the statistical process of failure bits and reduces the computational workload. The specific process of further optimizing the clustering algorithm to cluster the target blocks to obtain the block clustering results includes the following steps:

[0069] S331, take the Nth row position in the target block as the target row / column position.

[0070] Specifically, the points in the target block are divided into interlaced rows and columns, and the point in the Nth row of the target block is taken as the target row point. It should be noted that the point in the Nth column of the target block can also be selected as the target column point; in the corresponding steps, rows are changed to columns, and columns are changed to rows. Whether the clustering is performed by row or by column can be set based on the actual situation of the chip under test, and is not limited here.

[0071] S332 uses a one-dimensional clustering algorithm to perform a one-dimensional clustering operation on the invalid positions in the target row to obtain the one-dimensional clustering result.

[0072] Specifically, one-dimensional clustering is performed on the failed positions in the target row in sequence to obtain multiple cluster types. These multiple cluster types form the one-dimensional clustering result of the target row. Preferably, the one-dimensional clustering algorithm specifically calculates the distance between failed positions on the target row, sets a threshold, and defines adjacent failed positions as belonging to the same class if the distance is less than the threshold, and defines them as new cluster types if the distance is greater than the threshold.

[0073] S333 merges adjacent cluster types in the one-dimensional clustering result and the latest defined temporary clustering result to obtain the target clustering result.

[0074] Specifically, it is determined whether there are adjacent cluster types in the one-dimensional clustering result of the target row and the temporary clustering result. If so, all two adjacent cluster types in the one-dimensional clustering result of the target row and the temporary clustering result are merged, and the type names of the corresponding cluster types are adjusted. Then, based on the merging of adjacent cluster types, the cluster types in the one-dimensional clustering result of the target row and the cluster types in the temporary clustering result are integrated to obtain the target clustering result. For example, assuming the cluster types in the first row of the target block include clusters 1-3, and the cluster types in the second row include clusters 4-8, determine whether there are any invalid bits in the second row that belong to the same column as the invalid bits in the first row. If so, it means that the two invalid bits in the same column belong to adjacent cluster types. For example, assuming the third column of the first row is an invalid bit and belongs to cluster 1, and the third column of the second row is also an invalid bit and belongs to cluster 3, then cluster 1 and cluster 4 are adjacent cluster types. Merge the invalid bits in cluster 4 into cluster 1, and shift the cluster values ​​after cluster 4 one position forward. In this way, all adjacent cluster types in the first row and the second row are merged, and the two rows are integrated to obtain the target clustering result after merging the first and second rows.

[0075] It should be noted that as the value of N increases, the defined temporary clustering results are constantly changing. Therefore, the temporary clustering results used in this step are all the latest defined temporary clustering results. When N is 1, the temporary clustering results are set to be non-existent, meaning there is no need to merge the clustering results of this row with other clustering results.

[0076] S334, determine whether N is equal to the first preset threshold. If so, use the latest obtained target clustering result as the block failure clustering result. Otherwise, increment N by 1, define the latest obtained target clustering result as the temporary clustering result, and re-use the Nth row point as the target row point.

[0077] Specifically, a first preset threshold is set as the number of rows in the target block. That is, if it is determined that the current target row is the last row of the target block, the latest obtained target clustering result can be used as the block failure clustering result of the target block. If the value of N is less than the number of rows in the target block, it means that the target block has not yet been clustered to the last row. Then, N is incremented by 1, and the process proceeds to step S331 to redefine the target row position.

[0078] It should be noted that when clustering is performed by column, the first preset threshold is the number of columns of points in the target block.

[0079] Step S104: Based on the failure type information and failure clustering results, classify the failure bits of the chip under test to obtain the failure classification results of the chip under test.

[0080] Specifically, classifying the failure bits of the chip under test is essentially classifying the failure bits of each block within the chip. Further, the clustering type within each block of the chip under test is classified to obtain the block classification result for each block. The block classification results of all blocks then form the failure classification result of the chip under test.

[0081] Furthermore, the process of classifying the cluster types in a single block specifically includes: first, determining whether the total number of all failure bits in all cluster types in the block is greater than a first preset number; if so, the block is determined to be a block failure mode; that is, when the number of failure bits in all points of the block exceeds the first preset number, it means that there are too many failure bits in the block, and there is no need to classify them in detail. At this time, the block as a whole is a block failure mode; the value of the first preset number can be limited based on the actual situation.

[0082] If the block is determined not to be a block-based failure mode, it is further determined whether all cluster types in the block meet the linear failure setting. If they do, the cluster types that meet the linear failure setting are set as linear failure modes; if they do not meet the linear failure setting, the corresponding cluster types are left for further determination. The linear failure setting is defined as the number of failure positions in a cluster type being greater than a second preset number. That is, if the number of failure positions in a cluster type exceeds a certain value, the cluster type is determined to be a linear failure mode. The value of the second preset number can be limited based on actual circumstances. Furthermore, linear failure modes can be divided into multiple types. For example, if an entire row or column in a block consists entirely of failure positions, then this linear failure mode is a first-type linear failure mode. If a row or column in a block contains only one linear failure mode and contains points that are not failure positions, then this linear failure mode is a second-type linear failure mode. Other types of linear failure modes are third-type linear failure modes.

[0083] Finally, the remaining cluster types in the block whose failure types have not been determined are judged. That is, it is judged whether the remaining cluster types in the block meet the point failure setting. If they meet the setting, the cluster types that meet the point failure setting are set as point failure mode. If they do not meet the setting, the corresponding cluster types are set as other failure modes. The point failure setting is that the cluster type has only one failure position, has two consecutive failure positions, or has four adjacent failure positions that form a square.

[0084] Furthermore, when determining that a block is in a block failure mode, it is also necessary to calculate the density distribution of the failure sites in that block so that it can be displayed on the failure classification display diagram.

[0085] The failure classification results of the chip under test can also be displayed graphically. That is, the data classification method for chip failure types in this embodiment can also include the following steps.

[0086] Step S105: Based on the size information of the chip under test and the classification results, display the classification results as an image to obtain a failure classification display diagram of the chip under test.

[0087] Figure 2 An example diagram of the failure classification display diagram in Embodiment 1 of the present invention is shown; see reference. Figure 2 As shown, the classification results can be displayed using Python's matplotlib package. The blocks of blocky failure modes are represented as density distribution maps in the failure classification display diagram of the chip under test, to more accurately represent the information reflected by the actual data.

[0088] The data classification method for chip failure types in this embodiment may further include the following step S106.

[0089] Step S106: Store the test data in the database using the sparse matrix principle, establish an index relationship between each classification result in the failure classification results and the corresponding data in the test data, and store the indexed failure classification results in the database.

[0090] Specifically, after acquiring the test data, it needs to be stored in a database based on the principle of sparse matrices for subsequent data retrieval. Using sparse matrix principles to store the test data allows for matrix operations with less memory and faster computation speed. Then, each classification result in the failure classification results of the chip under test is indexed with its corresponding data in the test data, so that each classification result can be retrieved from its original test data through the index. Finally, the indexed failure classification results are stored in the database. Utilizing the concept of sparse matrices combined with failure mode storage classification results and test data effectively improves data retrieval efficiency.

[0091] The chip failure type data classification method provided in this invention improves clustering efficiency and saves memory by optimizing the clustering algorithm; it classifies chip failure positions based on clustering results, greatly improving the efficiency of chip failure type acquisition; and it displays the block failure model in the failure classification display diagram of the chip under test through density distribution, which can more accurately represent the situation reflected by the actual data. The method uses the concept of sparse matrix to combine failure mode storage classification results with original data, improving data reading and saving efficiency and saving storage space. Further optimization of the clustering algorithm, performing clustering only horizontally or vertically, reduces memory usage, and merging adjacent cluster types during the clustering process reduces the number of calculations and the use of process variables, thus improving overall computational efficiency.

[0092] Example 2

[0093] To address the technical problems existing in the prior art, embodiments of the present invention provide a data classification device for chip failure types.

[0094] Figure 3 A schematic diagram of the data classification device for chip failure types according to Embodiment 2 of the present invention is shown; Reference Figure 3 As shown, the chip failure type data classification device of this embodiment includes an information acquisition module, a test physical location information acquisition module, a failure clustering result acquisition module, and a failure classification result acquisition module.

[0095] The information acquisition module is used to acquire the design information of the chip under test, and based on the design information, acquire the chip's size information, logical location and physical location correspondence information, and failure type information.

[0096] The test physical location information acquisition module is used to acquire test data of the chip under test and convert the test data into test physical location information based on size information and the correspondence between logical location and physical location.

[0097] The failure clustering result acquisition module is used to cluster the failure bits of the chip under test based on the test physical location information and by optimizing the clustering algorithm to obtain the failure clustering results.

[0098] The failure classification result acquisition module is used to classify the failure bits of the chip under test based on failure type information and failure clustering results, so as to obtain the failure classification result of the chip under test.

[0099] The optimized clustering algorithm is as follows:

[0100] Use the Nth row / column position in the target block as the target row / column position;

[0101] One-dimensional clustering results are obtained by clustering the invalid positions in the target row / column points using a one-dimensional clustering algorithm.

[0102] The one-dimensional clustering results and the latest temporary clustering results are combined to obtain the target clustering result.

[0103] Determine if N equals the first preset threshold. If yes, use the latest acquired target clustering result as the invalid clustering result for the block. Otherwise, increment N by 1, define the latest acquired target clustering result as a temporary clustering result, and re-use the Nth row / column position as the target row / column position.

[0104] The initial value of N is 1, and when N is 1, the latest defined temporary clustering result is none. The chip under test includes multiple blocks, each block includes multiple rows or columns of points, and the first preset threshold is the number of rows or columns of points in the block.

[0105] The chip failure type data classification device provided in this invention improves clustering efficiency and saves memory by optimizing the clustering algorithm; it classifies chip failure positions based on clustering results, greatly improving the efficiency of chip failure type acquisition; it displays the block failure model in the failure classification display diagram of the chip under test through density distribution, which can more accurately express the situation reflected by the actual data; it adopts the concept of sparse matrix to combine failure mode storage classification results with original data, improving data reading and saving efficiency and saving storage space. Further optimization of the clustering algorithm, performing clustering only horizontally or vertically, reduces memory usage, and merging adjacent cluster types during the clustering process reduces the number of calculations and the use of process variables, improving overall computational efficiency.

[0106] Example 3

[0107] To address the aforementioned technical problems in the prior art, this embodiment of the invention also provides a storage medium storing a computer program that, when executed by a processor, can implement all the steps in the data classification method for chip failure types in Embodiment 1.

[0108] The specific steps of the data classification method for chip failure types and the beneficial effects obtained by applying the readable storage medium provided in this embodiment of the invention are the same as in Embodiment 1, and will not be repeated here.

[0109] It should be noted that storage media include various media that can store program code, such as ROM, RAM, magnetic disks, or optical disks.

[0110] Example 4

[0111] To address the aforementioned technical problems in the prior art, embodiments of the present invention also provide a terminal.

[0112] Figure 4 A schematic diagram of the terminal structure of Embodiment 4 of the present invention is shown, with reference to... Figure 4 In this embodiment, the terminal includes a processor and a memory that are interconnected. The memory is used to store computer programs, and the processor is used to execute the computer programs stored in the memory, so that when the terminal is executed, it can implement all the steps in the data classification method for chip failure types in Embodiment 1.

[0113] The specific steps of the data classification method for chip failure types and the beneficial effects obtained by the terminal provided by the embodiment of the present invention are the same as those in Embodiment 1, and will not be repeated here.

[0114] It should be noted that the memory may include random access memory (RAM) and may also include non-volatile memory, such as at least one disk storage device. Similarly, the processor can be a general-purpose processor, including a central processing unit (CPU), a network processor (NP), etc.; it can also be a digital signal processor (DSP), an application-specific integrated circuit (ASIC), a field-programmable gate array (FPGA), or other programmable logic devices, discrete gate or transistor logic devices, or discrete hardware components.

[0115] While the embodiments disclosed in this invention are as described above, the content is merely for the purpose of facilitating understanding of the invention and is not intended to limit the invention. Any person skilled in the art to which this invention pertains may make any modifications and changes in form and detail of the implementation without departing from the spirit and scope disclosed herein; however, the scope of protection of this invention shall still be determined by the scope defined in the appended claims.

Claims

1. A data classification method for chip failure types, comprising: Obtain the design information of the chip under test, and based on the design information, obtain the size information, logical location and physical location correspondence information, and failure type information of the chip under test; The test data of the chip under test is obtained, and the test data is converted into test physical location information based on the size information and the correspondence between the logical position and the physical position. Based on the test physical location information, the failure bits in the chip under test are clustered by an optimized clustering algorithm to obtain the failure clustering results. Based on the failure type information and the failure clustering results, the failure bits of the chip under test are classified to obtain the failure classification results of the chip under test. The optimized clustering algorithm includes: Use the Nth row / column position in the target block as the target row / column position; The failure positions in the target row / column points are clustered using a one-dimensional clustering algorithm to obtain one-dimensional clustering results; The adjacent cluster types in the one-dimensional clustering results and the latest defined temporary clustering results are merged to obtain the target clustering result; Determine if N equals the first preset threshold. If yes, use the latest acquired target clustering result as the block failure clustering result; otherwise, increment N by 1, define the latest acquired target clustering result as a temporary clustering result, and re-use the Nth row / column position as the target row / column position. The initial value of N is 1, and when N is 1, the latest defined temporary clustering result is none. The chip under test includes multiple blocks, and each block includes multiple rows or columns of points. The first preset threshold is the number of rows or columns of points in the block. The step of classifying the failure bits of the chip under test based on the failure type information and the failure clustering results to obtain the failure classification results of the chip under test includes: The clustering types in each block of the chip under test are classified to obtain the block classification result of each block. The block classification results of all the blocks form the failure classification result of the chip under test. The process of classifying cluster types within a single block includes: Determine whether the total number of failure bits in all cluster types in the block is greater than the first preset number. If so, the block is determined to be a block failure mode. Otherwise, determine whether all cluster types in the block meet the linear failure setting and set the cluster types that meet the linear failure setting as linear failure mode. Then, determine whether the remaining cluster types in the block meet the point failure setting and set the cluster types that meet the point failure setting as point failure mode, and set the cluster types that do not meet the point failure setting as other failure modes. The linear failure is defined as the number of failure positions in the cluster type being greater than a second preset number, and the point failure is defined as the cluster type having only one failure position, having two consecutive failure positions, or having four adjacent failure positions forming a square.

2. The classification method according to claim 1, characterized in that, Converting the test data into test physical location information based on the size information and the logical position-to-physical position correspondence information includes: Convert the test data into matrix format data; Based on the size information and the correspondence between the logical position and the physical position, the matrix format data is converted into test physical position information.

3. The classification method according to claim 1, characterized in that, Based on the test physical location information, the failure bits of the chip under test are clustered using an optimized clustering algorithm to obtain the failure clustering results. The steps include: Sort all blocks in the chip under test to obtain the block sorting result; Take the Mth block in the block sorting result as the target block; The clustering algorithm is optimized to cluster the invalid bits in the target block to obtain the block clustering result of the Mth block; Determine whether M is equal to the preset sorting value. If so, the block clustering results of all blocks form the failure clustering results of the chip under test. Otherwise, increment M by 1 and re-select the Mth block in the block sorting results as the target block. Wherein, the initial value of M is 1, and the preset sorting value is the total number of blocks in the chip under test.

4. The classification method according to claim 3, characterized in that, Also includes: Based on the size information of the chip under test and the classification result, the classification result is displayed as an image to obtain a failure classification display diagram of the chip under test.

5. The classification method according to claim 4, characterized in that, When a block is determined to be in a block failure mode, the density distribution of the failure bits in that block is calculated, and the block is represented as a density distribution map in the failure classification display diagram of the chip under test.

6. The classification method according to claim 1, characterized in that, Also includes: The test data is stored in the database using the sparse matrix principle. Each classification result in the failure classification results is indexed to the corresponding data in the test data, and the indexed failure classification results are stored in the database.

7. A data classification device for chip failure types, characterized in that, It includes an information acquisition module, a test physical location information acquisition module, a failure clustering result acquisition module, and a failure classification result acquisition module; The information acquisition module is used to acquire the design information of the chip under test, and based on the design information, acquire the size information, logical position and physical position correspondence information, and failure type information of the chip under test. The test physical location information acquisition module is used to acquire the test data of the chip under test, and convert the test data into test physical location information based on the size information and the logical position to physical position correspondence information; The failure clustering result acquisition module is used to cluster the failure positions in the chip under test based on the test physical location information using an optimized clustering algorithm to obtain failure clustering results. The failure classification result acquisition module is used to classify the failure positions of the chip under test based on the failure type information and the failure clustering result, so as to obtain the failure classification result of the chip under test. The optimized clustering algorithm includes: Use the Nth row / column position in the target block as the target row / column position; The failure positions in the target row / column points are clustered using a one-dimensional clustering algorithm to obtain one-dimensional clustering results; The adjacent clustering types in the one-dimensional clustering result and the latest defined temporary clustering result are merged to obtain the target clustering result; Determine if N equals the first preset threshold. If yes, use the latest acquired target clustering result as the block failure clustering result; otherwise, increment N by 1, define the latest acquired target clustering result as a temporary clustering result, and re-use the Nth row / column position as the target row / column position. The initial value of N is 1, and when N is 1, the latest defined temporary clustering result is none. The chip under test includes multiple blocks, and each block includes multiple rows or columns of points. The first preset threshold is the number of rows or columns of points in the block. The step of classifying the failure bits of the chip under test based on the failure type information and the failure clustering results to obtain the failure classification results of the chip under test includes: The clustering types in each block of the chip under test are classified to obtain the block classification result of each block. The block classification results of all the blocks form the failure classification result of the chip under test. The process of classifying cluster types within a single block includes: Determine whether the total number of failure bits in all cluster types in the block is greater than the first preset number. If so, the block is determined to be a block failure mode. Otherwise, determine whether all cluster types in the block meet the linear failure setting and set the cluster types that meet the linear failure setting as linear failure mode. Then, determine whether the remaining cluster types in the block meet the point failure setting and set the cluster types that meet the point failure setting as point failure mode, and set the cluster types that do not meet the point failure setting as other failure modes. The linear failure is defined as the number of failure positions in the cluster type being greater than a second preset number, and the point failure is defined as the cluster type having only one failure position, having two consecutive failure positions, or having four adjacent failure positions forming a square.

8. A storage medium having a computer program stored thereon, characterized in that, When executed by the processor, the program implements the data classification method for any one of the chip failure types in claims 1 to 6.

9. A terminal, characterized in that, include: A processor and a memory, wherein the memory and the processor are communicatively connected; The memory is used to store a computer program, and the processor is used to execute the computer program stored in the memory to cause the terminal to perform a data classification method for any one of the chip failure types as claimed in claims 1 to 6.

Images