Semiconductor packaging defect automatic detection system and method

Abstract

The present invention discloses an automatic semiconductor package defect detection system and method, relating to the field of semiconductor package detection technology. The system and method comprise: extracting the dth package based on a package distribution set; performing multi-feature dominant defect detection on the dth package based on a semiconductor package dominant defect detection channel to obtain the dth dominant defect detection result; performing electrical performance defect detection on the dth package to obtain the dth electrical performance defect detection result; performing mechanical performance defect detection and analysis on the dth package to obtain the dth mechanical performance defect detection result; performing data fusion to generate a comprehensive defect detection result for the dth package; and performing defect detection on each package within the package distribution set, combining the comprehensive defect detection result of the dth package to generate a comprehensive package defect distribution set. The system solves the technical problem of inaccurate detection results caused by insufficiently comprehensive semiconductor package defect detection in the prior art, thereby achieving the technical effect of improving the accuracy of detection results.

Description

Technical Field

[0001] The present invention relates to the technical field of semiconductor packaging detection, and in particular to a system and method for automatically detecting semiconductor packaging defects. Background Art

[0002] Semiconductor packaging is the process of encapsulating a chip within a protective housing. The primary purpose of packaging is to protect the chip from physical damage and environmental influences and to provide a means of connecting the chip to external circuitry. Defects such as poor soldering, pin misalignment, and cracks in the packaging material can occur during the packaging process. These defects can lead to chip performance degradation, functional failure, or even complete failure. Therefore, timely and accurate detection of packaging defects is crucial. Existing technologies for semiconductor packaging defect detection are often incomplete, resulting in inaccurate test results. Summary of the Invention

[0003] The embodiments of the present application provide a system and method for automatically detecting semiconductor package defects, which solves the technical problem in the prior art that semiconductor package defect detection is not comprehensive enough, resulting in inaccurate detection results.

[0004] In view of the above problems, embodiments of the present application provide a system and method for automatically detecting semiconductor packaging defects.

[0005] A first aspect of an embodiment of the present application provides a semiconductor packaging defect automatic detection system, the system comprising:

[0006] an extraction module configured to extract a dth package according to a package distribution set, wherein the package distribution set includes a plurality of packages corresponding to a target semiconductor packaging process solution, wherein d is a positive integer greater than or equal to 1, d belongs to D, and D is a total number of the plurality of packages;

[0007] a first detection module, configured to perform a multi-feature dominant defect detection on the dth package according to a semiconductor package dominant defect detection channel to obtain a dth dominant defect detection result;

[0008] a second detection module, configured to perform electrical performance benchmark calculation according to the target semiconductor packaging process solution, build a package electrical performance benchmark matrix, and perform electrical performance defect detection on the dth package based on the package electrical performance benchmark matrix to obtain a dth electrical performance defect detection result;

[0009] a third detection module, configured to perform mechanical performance defect detection and analysis on the dth package according to a package-mechanized combined detection terminal to obtain a dth mechanical performance defect detection result, wherein the package-mechanized combined detection terminal includes a package chemical performance detection terminal and a package mechanical performance detection terminal;

[0010] a data fusion module, configured to fuse data based on the dth dominant defect detection result, the dth electrical performance defect detection result, and the dth mechanical performance defect detection result to generate a dth package comprehensive defect detection result;

[0011] A package detection module is configured to continue to perform defect detection on each package in the package distribution set based on the semiconductor package dominant defect detection channel, the package electrical performance benchmark matrix, and the package mechanized joint detection terminal, and generate a package comprehensive defect distribution set in combination with the comprehensive defect detection result of the dth package.

[0012] A second aspect of an embodiment of the present application provides a method for automatically detecting semiconductor package defects, the method comprising:

[0013] Extracting a d-th package according to a package distribution set, wherein the package distribution set includes multiple packages corresponding to a target semiconductor packaging process solution, wherein d is a positive integer greater than or equal to 1, d belongs to D, and D is the total number of the multiple packages;

[0014] performing a multi-feature dominant defect detection on the dth package according to a semiconductor package dominant defect detection channel to obtain a dth dominant defect detection result;

[0015] performing an electrical performance benchmark calculation according to the target semiconductor packaging process solution, building a package electrical performance benchmark matrix, and performing an electrical performance defect detection on the dth package based on the package electrical performance benchmark matrix to obtain a dth electrical performance defect detection result;

[0016] performing mechanical performance defect detection and analysis on the dth package according to the package mechanization combined detection terminal to obtain a dth mechanical performance defect detection result, wherein the package mechanization combined detection terminal includes a package chemical performance detection terminal and a package mechanical performance detection terminal;

[0017] performing data fusion based on the dth dominant defect detection result, the dth electrical performance defect detection result, and the dth mechanical performance defect detection result to generate a dth package comprehensive defect detection result;

[0018] Continue to perform defect detection on each package in the package distribution set according to the semiconductor package explicit defect detection channel, the package electrical performance benchmark matrix and the package mechanized joint detection terminal, and generate a package comprehensive defect distribution set in combination with the comprehensive defect detection result of the dth package.

[0019] One or more technical solutions provided in this application have at least the following technical effects or advantages:

[0020] A package distribution set includes multiple packages corresponding to a target semiconductor packaging process solution. The dth package is extracted from the package distribution set, where d is a positive integer greater than or equal to 1, belongs to D, and D is the total number of the multiple packages. A multi-feature dominant defect detection channel is used to perform dominant defect detection on the dth package to obtain a dth dominant defect detection result. An electrical performance benchmark is calculated based on the target semiconductor packaging process solution, and a package electrical performance benchmark matrix is established. Then, the package electrical performance benchmark matrix is used to perform electrical performance defect detection on the dth package to obtain a dth electrical performance defect detection result. A package-mechanized combined detection terminal is used to perform mechanical performance defect detection and analysis on the dth package to obtain a dth mechanical performance defect detection result. The package-mechanized combined detection terminal includes a package chemical performance detection terminal and a package mechanical performance detection terminal. The dth dominant defect detection result, the dth electrical performance defect detection result, and the dth mechanical performance defect detection result are data fused to generate a comprehensive defect detection result for the dth package. Based on the semiconductor package explicit defect detection channel, the package electrical performance benchmark matrix, and the package mechanized joint detection terminal, defect detection is performed on other packages within the package distribution set. Combined with the comprehensive defect detection results of the dth package, a comprehensive package defect distribution set is generated. This solves the technical problem of inaccurate test results caused by insufficiently comprehensive semiconductor package defect detection in existing technologies, achieving the technical effect of improving the accuracy of test results. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0022] Figure 1 A schematic diagram of the structure of a semiconductor packaging defect automatic detection system provided in an embodiment of the present application;

[0023] Figure 2 A schematic flow chart of a method for automatically detecting semiconductor packaging defects provided in an embodiment of the present application.

[0024] Description of the accompanying drawings: extraction module 11, first detection module 12, second detection module 13, third detection module 14, data fusion module 15, package detection module 16. DETAILED DESCRIPTION

[0025] The embodiments of the present application solve the technical problem in the prior art that semiconductor package defect detection is not comprehensive enough, resulting in inaccurate detection results, by providing a system and method for automatic detection of semiconductor package defects.

[0026] The following will be combined with the accompanying drawings in the embodiments of this application to clearly and completely describe the technical solutions in the embodiments of this application. Obviously, the embodiments described are only some of the embodiments of this application, not all of them. Based on the embodiments in this application, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of this application.

[0027] It should be noted that the terms "including" and "having" and any variations thereof are intended to cover non-exclusive inclusions. For example, a process, method, system, product or server that includes a series of steps or units is not necessarily limited to those steps or units clearly listed, but may include other steps or modules that are not clearly listed or are inherent to these processes, methods, products or devices.

[0028] Example 1

[0029] like Figure 1 As shown, the present application provides a semiconductor packaging defect automatic detection system, which is used to perform a semiconductor packaging defect automatic detection method. The system includes:

[0030] An extraction module 11 is configured to extract a d-th package according to a package distribution set, wherein the package distribution set includes a plurality of packages corresponding to a target semiconductor packaging process solution, wherein d is a positive integer greater than or equal to 1, d belongs to D, and D is a total number of the plurality of packages;

[0031] a first detection module 12, configured to perform a multi-feature dominant defect detection on the dth package according to a semiconductor package dominant defect detection channel to obtain a dth dominant defect detection result;

[0032] a second detection module 13 configured to perform electrical performance benchmark calculations according to the target semiconductor packaging process solution, construct a package electrical performance benchmark matrix, and perform electrical performance defect detection on the dth package based on the package electrical performance benchmark matrix to obtain a dth electrical performance defect detection result;

[0033] a third detection module 14, configured to perform mechanical performance defect detection and analysis on the dth package according to a package-mechanized combined detection terminal to obtain a dth mechanical performance defect detection result, wherein the package-mechanized combined detection terminal includes a package chemical performance detection terminal and a package mechanical performance detection terminal;

[0034] a data fusion module 15 configured to fuse data based on the dth dominant defect detection result, the dth electrical performance defect detection result, and the dth mechanical performance defect detection result to generate a dth package comprehensive defect detection result;

[0035] The package detection module 16 is used to continue to perform defect detection on each package in the package distribution set based on the semiconductor package explicit defect detection channel, the package electrical performance benchmark matrix and the package mechanized joint detection terminal, and generate a package comprehensive defect distribution set in combination with the comprehensive defect detection result of the dth package.

[0036] Furthermore, the first detection module 12 is configured to perform the following method:

[0037] The semiconductor package explicit defect detection channel includes a detection data acquisition branch, an acquisition data cleaning branch and an explicit defect detection branch;

[0038] Performing data collection on the d-th package using M heterogeneous detection and collection devices in the detection data collection branch to obtain M package collection data, where M is a positive integer greater than 1;

[0039] Inputting the M package collected data into the collected data cleaning branch, and preprocessing the M package collected data according to the collected data denoising algorithm in the collected data cleaning branch to obtain M collected cleaned data;

[0040] Performing defect recognition on the M collected cleaning data according to the explicit defect detection branch to generate M defect recognition results;

[0041] Data fusion is performed on the M defect recognition results to generate the d-th dominant defect detection result.

[0042] Furthermore, the first detection module 12 is configured to perform the following method:

[0043] Performing package defect detection record backtracking according to the M heterogeneous detection and collection devices to obtain M sets of package defect detection record sets;

[0044] Using a preset defect detection accuracy threshold as a supervised learning constraint, supervised learning is performed on the M sets of package defect detection record sets to generate the explicit defect detection branch, wherein the explicit defect detection branch includes M explicit defect detectors;

[0045] The M collected and cleaned data are input into the M explicit defect detectors to obtain the M defect recognition results.

[0046] Furthermore, the second detection module 13 is configured to perform the following method:

[0047] Backtracking the electrical performance test records of normal sample packages according to the target semiconductor packaging process plan to obtain a normal sample electrical performance test record set;

[0048] Performing standardization processing on the normal sample electrical performance test record set to obtain a standard normal electrical performance test record set;

[0049] Clustering is performed according to the standard normal electrical performance test record set to obtain multiple normal electrical performance test intervals;

[0050] Performing centralized value calculations on the plurality of normal electrical performance detection intervals respectively to generate a plurality of normal electrical performance detection reference values;

[0051] The package electrical performance reference matrix is constructed according to the multiple normal electrical performance detection reference values.

[0052] Furthermore, the second detection module 13 is configured to perform the following method:

[0053] performing multiple rounds of electrical performance testing on the dth package according to the package electrical performance testing terminal to obtain multiple sets of electrical performance testing data;

[0054] Performing standardization processing on the plurality of electrical performance test data groups to generate a plurality of electrical performance test standard groups;

[0055] Performing concentrated value calculation according to the plurality of electrical performance test standard groups to obtain a package electrical performance test matrix, wherein the package electrical performance test matrix includes a plurality of electrical performance test element values;

[0056] performing deviation calculation on the package electrical performance detection matrix based on the package electrical performance reference matrix to generate a package electrical performance deviation matrix;

[0057] Performing an electrical performance deviation depth evaluation based on the package electrical performance deviation matrix to generate an electrical performance defect coefficient;

[0058] The package electrical performance deviation matrix and the electrical performance defect coefficient are added to the dth electrical performance defect detection result.

[0059] Furthermore, the third detection module 14 is configured to perform the following method:

[0060] Predicting the chemical properties of the package according to the target semiconductor packaging process plan to obtain a multi-dimensional target chemical property prediction result, and normalizing the multi-dimensional target chemical property prediction result to generate a target chemical property vector set;

[0061] performing multiple rounds of chemical property testing on the dth package according to the package chemical property testing terminal to obtain multiple groups of chemical property testing data;

[0062] Performing standardization processing and central value calculation on the multiple chemical property test data sets to obtain a d-th package chemical property vector set;

[0063] Taking the target chemical property vector set as the package chemical property constraint, performing deviation comparison on the d-th package chemical property vector set to generate a d-th package chemical property deviation vector set;

[0064] Performing a chemical property deviation depth evaluation based on the dth package chemical property deviation vector set to generate a dth chemical property deviation depth coefficient;

[0065] The dth package chemical property deviation vector set and the dth chemical property deviation depth coefficient are added to the dth chemical property defect detection result.

[0066] Furthermore, the third detection module 14 is configured to perform the following method:

[0067] Predicting the mechanical properties of the package according to the target semiconductor packaging process solution to obtain a multi-dimensional target mechanical property prediction result, and normalizing the multi-dimensional target mechanical property prediction result to generate a target mechanical property vector set;

[0068] performing multiple rounds of mechanical property testing on the dth package according to the package mechanical property testing terminal to obtain multiple sets of mechanical property testing data;

[0069] Performing standardization processing and concentrated value calculation on the multiple sets of mechanical property test data to obtain a d-th package mechanical property vector set;

[0070] Taking the target mechanical property vector set as the package mechanical property constraint, performing deviation comparison on the d-th package mechanical property vector set to generate a d-th package mechanical property deviation vector set;

[0071] Performing a mechanical property deviation depth evaluation based on the dth package mechanical property deviation vector set to generate a dth mechanical property deviation depth coefficient;

[0072] The dth package mechanical property deviation vector set and the dth mechanical property deviation depth coefficient are added to the dth mechanical property defect detection result.

[0073] Example 2

[0074] like Figure 2 As shown, an embodiment of the present application provides a method for automatically detecting semiconductor package defects, wherein the method includes:

[0075] Extracting a d-th package according to a package distribution set, wherein the package distribution set includes multiple packages corresponding to a target semiconductor packaging process solution, wherein d is a positive integer greater than or equal to 1, d belongs to D, and D is the total number of the multiple packages;

[0076] The package distribution set includes multiple packages corresponding to the target semiconductor packaging process solution. The dth package is extracted from the package distribution set, where d is a positive integer greater than or equal to 1, d belongs to D, and D is the total number of the multiple packages.

[0077] performing a multi-feature dominant defect detection on the dth package according to a semiconductor package dominant defect detection channel to obtain a dth dominant defect detection result;

[0078] The dth package is subjected to multi-feature dominant defect detection through the semiconductor package dominant defect detection channel. The multi-feature dominant defect detection can comprehensively and accurately capture various physical defects on the surface of the package, thereby obtaining the dth dominant defect detection result. The dth dominant defect detection result refers to the physical defect detection result on the surface of the package.

[0079] Furthermore, performing a multi-feature dominant defect inspection on the dth package according to the semiconductor package dominant defect inspection channel to obtain a dth dominant defect inspection result includes:

[0080] The semiconductor package explicit defect detection channel includes a detection data acquisition branch, an acquisition data cleaning branch and an explicit defect detection branch;

[0081] Performing data collection on the d-th package using M heterogeneous detection and collection devices in the detection data collection branch to obtain M package collection data, where M is a positive integer greater than 1;

[0082] Inputting the M package collected data into the collected data cleaning branch, and preprocessing the M package collected data according to the collected data denoising algorithm in the collected data cleaning branch to obtain M collected cleaned data;

[0083] Performing defect recognition on the M collected cleaning data according to the explicit defect detection branch to generate M defect recognition results;

[0084] Data fusion is performed on the M defect recognition results to generate the d-th dominant defect detection result.

[0085] The semiconductor package explicit defect detection channel consists of a test data acquisition branch, a data cleaning branch, and a explicit defect detection branch. The test data acquisition branch, consisting of M heterogeneous test and acquisition devices, is responsible for collecting raw package data. These heterogeneous devices include laser scanners, optical microscopes, and infrared thermal imaging devices. Data from the dth package is collected using these M heterogeneous test and acquisition devices, resulting in M package data collections, where M is a positive integer greater than 1. The data cleaning branch preprocesses the collected raw data to eliminate noise and improve data quality. The M package data collections are input into the data cleaning branch. Using data noise reduction algorithms within the data cleaning branch, such as wavelet transform, median filtering, and median filtering, the data collections are preprocessed to produce M cleaned data. The explicit defect detection branch is responsible for defect identification on the cleaned data. The M cleaned data collections are then input into the explicit defect detection branch for defect identification, generating M defect identification results. Because multiple heterogeneous detection and acquisition devices are used, each device may capture package information from different angles or conditions, so their recognition results may differ or complement each other. The M defect recognition results are fused to generate a comprehensive d-th dominant defect detection result.

[0086] Furthermore, defect identification is performed on the M collected cleaning data according to the explicit defect detection branch to generate M defect identification results, including:

[0087] Performing package defect detection record backtracking according to the M heterogeneous detection and collection devices to obtain M sets of package defect detection record sets;

[0088] Using a preset defect detection accuracy threshold as a supervised learning constraint, supervised learning is performed on the M sets of package defect detection record sets to generate the explicit defect detection branch, wherein the explicit defect detection branch includes M explicit defect detectors;

[0089] The M collected and cleaned data are input into the M explicit defect detectors to obtain the M defect recognition results.

[0090] Based on the inspection history of M heterogeneous inspection and acquisition devices, package defect inspection records are backtracked to obtain M sets of package defect inspection records, which record historical package defect inspection results. To generate the explicit defect detection branch, supervised learning is required. Specifically, data with clear defect labels is extracted from the M sets of package defect inspection records, and a preset defect detection accuracy threshold is set to evaluate the model performance during the training process. Using labeled training data, M explicit defect detectors are trained separately. Each explicit defect detector corresponds to a heterogeneous inspection and acquisition device and learns to identify explicit defects from the data collected by that device. During the training process, the parameters and structure of the explicit defect detectors are continuously adjusted to optimize their performance until the preset defect detection accuracy threshold is reached or exceeded. The M cleaned data are then fed into the corresponding M explicit defect detectors. Each explicit defect detector processes the cleaned data and outputs a defect recognition result. Ultimately, M defect recognition results are obtained.

[0091] performing an electrical performance benchmark calculation according to the target semiconductor packaging process solution, building a package electrical performance benchmark matrix, and performing an electrical performance defect detection on the dth package based on the package electrical performance benchmark matrix to obtain a dth electrical performance defect detection result;

[0092] The target semiconductor packaging process plan includes the package type, structure, materials, and electrical performance parameters. Based on the target semiconductor packaging process plan, an electrical performance benchmark calculation is performed to obtain benchmark values for each key electrical performance parameter. This is used to construct a package electrical performance benchmark matrix, which serves as the basis for electrical performance defect detection. The dth package is then tested for electrical performance defects and compared with the package electrical performance benchmark matrix to obtain the dth electrical performance defect detection result.

[0093] Furthermore, electrical performance benchmark calculations are performed according to the target semiconductor packaging process solution to build a package electrical performance benchmark matrix, including:

[0094] Backtracking the electrical performance test records of normal sample packages according to the target semiconductor packaging process plan to obtain a normal sample electrical performance test record set;

[0095] Performing standardization processing on the normal sample electrical performance test record set to obtain a standard normal electrical performance test record set;

[0096] Clustering is performed according to the standard normal electrical performance test record set to obtain multiple normal electrical performance test intervals;

[0097] Performing centralized value calculations on the plurality of normal electrical performance detection intervals respectively to generate a plurality of normal electrical performance detection reference values;

[0098] The package electrical performance reference matrix is constructed according to the multiple normal electrical performance detection reference values.

[0099] A normalized sample electrical performance test record set containing normal electrical performance parameters is obtained by reviewing the electrical performance test records of previously produced and verified packages according to the target semiconductor packaging process plan. This set of normalized sample electrical performance test records is cleaned to remove outliers, missing values, or duplicate data. The cleaned data is then normalized to eliminate dimensional differences between parameters, thereby obtaining a standard normalized electrical performance test record set. A clustering algorithm, such as K-means or hierarchical clustering, is selected to cluster the standard normalized electrical performance test record set, grouping similar data points into the same category to form multiple normalized electrical performance test intervals. For each clustered normalized electrical performance test interval, the central value (e.g., mean, median, etc.) is calculated as the representative value for that interval. These central values are then used as normalized electrical performance test benchmark values, reflecting the expected levels of each electrical performance parameter under normal process conditions. Multiple normal electrical performance test benchmark values are organized into a matrix according to the corresponding electrical performance parameters and intervals. This matrix is the package electrical performance benchmark matrix, which contains the expected range and benchmark values of each electrical performance parameter under different process conditions.

[0100] Furthermore, performing electrical performance defect detection on the dth package based on the package electrical performance reference matrix to obtain a dth electrical performance defect detection result includes:

[0101] performing multiple rounds of electrical performance testing on the dth package according to the package electrical performance testing terminal to obtain multiple sets of electrical performance testing data;

[0102] Performing standardization processing on the plurality of electrical performance test data groups to generate a plurality of electrical performance test standard groups;

[0103] Performing concentrated value calculation according to the plurality of electrical performance test standard groups to obtain a package electrical performance test matrix, wherein the package electrical performance test matrix includes a plurality of electrical performance test element values;

[0104] performing deviation calculation on the package electrical performance detection matrix based on the package electrical performance reference matrix to generate a package electrical performance deviation matrix;

[0105] Performing an electrical performance deviation depth evaluation based on the package electrical performance deviation matrix to generate an electrical performance defect coefficient;

[0106] The package electrical performance deviation matrix and the electrical performance defect coefficient are added to the dth electrical performance defect detection result.

[0107] The package electrical performance testing terminal is used to test the electrical performance of the package. Multiple rounds of electrical performance testing are performed on the dth package using this terminal, including the measurement of multiple key electrical performance parameters. Multiple rounds of electrical performance testing involve testing each electrical performance parameter multiple times in each round, and these rounds are repeated until all electrical performance parameters are tested. The electrical performance data obtained from these tests is recorded to form multiple sets of electrical performance test data. The collected multiple sets of electrical performance test data are cleaned to remove outliers, duplicate data, or data that does not meet the requirements. The cleaned data are then standardized to generate multiple sets of electrical performance test standard sets. Centralized value calculations are performed on the standardized multiple sets of electrical performance test data, such as the mean and median of the parameters in each test round. These centralized values are organized into a matrix, namely the package electrical performance test matrix. Using the package electrical performance benchmark matrix as a reference, deviations are calculated for each element in the package electrical performance test matrix. All calculated deviations are organized into a new matrix, the package electrical performance deviation matrix. This matrix reflects the electrical performance deviation of the dth package relative to the benchmark across multiple rounds of testing. Based on this matrix, an in-depth evaluation of electrical performance deviations is performed, analyzing the magnitude, direction, and trend of the deviations. This coefficient is used to quantify the degree of electrical performance defects in the dth package. The package electrical performance deviation matrix and coefficient are then added to the dth electrical performance defect test results.

[0108] performing mechanical performance defect detection and analysis on the dth package according to the package mechanization combined detection terminal to obtain a dth mechanical performance defect detection result, wherein the package mechanization combined detection terminal includes a package chemical performance detection terminal and a package mechanical performance detection terminal;

[0109] The package-mechanized combined inspection terminal includes a package chemical performance inspection terminal and a package mechanical performance inspection terminal, which are used to inspect the chemical and mechanical properties of the package. The package-mechanized combined inspection terminal performs mechanical performance defect inspection and analysis on the dth package, obtaining the dth mechanical performance defect inspection result. The dth mechanical performance defect inspection result includes chemical and mechanical performance defects.

[0110] Furthermore, the package-mechanized joint detection terminal performs mechanical performance defect detection and analysis on the dth package to obtain the dth mechanical performance defect detection result, including:

[0111] Predicting the chemical properties of the package according to the target semiconductor packaging process plan to obtain a multi-dimensional target chemical property prediction result, and normalizing the multi-dimensional target chemical property prediction result to generate a target chemical property vector set;

[0112] performing multiple rounds of chemical property testing on the dth package according to the package chemical property testing terminal to obtain multiple groups of chemical property testing data;

[0113] Performing standardization processing and central value calculation on the multiple chemical property test data sets to obtain a d-th package chemical property vector set;

[0114] Taking the target chemical property vector set as the package chemical property constraint, performing deviation comparison on the d-th package chemical property vector set to generate a d-th package chemical property deviation vector set;

[0115] Performing a chemical property deviation depth evaluation based on the dth package chemical property deviation vector set to generate a dth chemical property deviation depth coefficient;

[0116] The dth package chemical property deviation vector set and the dth chemical property deviation depth coefficient are added to the dth chemical property defect detection result.

[0117] Based on the target semiconductor packaging process plan, the chemical properties of the package, including key chemical performance indicators such as oxidation resistance, corrosion resistance, and stability, are predicted to produce multidimensional target chemical performance prediction results, where each dimension represents a chemical performance parameter. These multidimensional target chemical performance prediction results are normalized to eliminate dimensional differences between the different parameters, allowing for comparison and analysis on a consistent scale. After normalization, the multidimensional target chemical performance prediction results form a vector set, namely the target chemical performance vector set, which serves as a reference for chemical performance evaluation. Multiple rounds of chemical performance testing are performed on the dth package using a package chemical performance testing terminal. The chemical performance data from each round is recorded to form multiple chemical performance test data sets. The multiple chemical performance test data sets are then normalized using the same standardization process as the target chemical performance vector set. Centralized values are then calculated on the normalized chemical performance test data, and these centralized values are organized into a vector form, namely the dth package chemical performance vector set, which reflects the chemical performance of the dth package. Using the target chemical property vector set as the package chemical property constraint, a deviation comparison is performed on the chemical property vector set of the dth package. The comparison results are combined into a new vector set, the dth package chemical property deviation vector set. A depth evaluation of chemical property deviations is performed based on the dth package chemical property deviation vector set, generating a chemical property deviation depth coefficient. The chemical property deviation depth coefficient is used to quantify the extent and significance of chemical property defects in the dth package. The dth package chemical property deviation vector set and the chemical property deviation depth coefficient are added to the dth mechanical performance defect detection results.

[0118] Furthermore, the package-mechanized joint detection terminal performs mechanical performance defect detection and analysis on the dth package to obtain the dth mechanical performance defect detection result, including:

[0119] Predicting the mechanical properties of the package according to the target semiconductor packaging process solution to obtain a multi-dimensional target mechanical property prediction result, and normalizing the multi-dimensional target mechanical property prediction result to generate a target mechanical property vector set;

[0120] performing multiple rounds of mechanical property testing on the dth package according to the package mechanical property testing terminal to obtain multiple sets of mechanical property testing data;

[0121] Performing standardization processing and concentrated value calculation on the multiple sets of mechanical property test data to obtain a d-th package mechanical property vector set;

[0122] Taking the target mechanical property vector set as the package mechanical property constraint, performing deviation comparison on the d-th package mechanical property vector set to generate a d-th package mechanical property deviation vector set;

[0123] Performing a mechanical property deviation depth evaluation based on the dth package mechanical property deviation vector set to generate a dth mechanical property deviation depth coefficient;

[0124] The dth package mechanical property deviation vector set and the dth mechanical property deviation depth coefficient are added to the dth mechanical property defect detection result.

[0125] Based on the target semiconductor packaging process plan, the mechanical properties of the package, including key mechanical performance indicators such as strength, hardness, toughness, and wear resistance, are predicted to obtain multi-dimensional target mechanical performance prediction results, with each dimension representing a mechanical performance parameter. These multi-dimensional target mechanical performance prediction results are normalized to eliminate dimensional differences between the different parameters, allowing for comparison and analysis on a consistent scale. After normalization, the multi-dimensional target mechanical performance prediction results form a vector set, namely the target mechanical performance vector set, which serves as a reference for mechanical performance evaluation. Multiple rounds of mechanical performance testing are performed on the dth package using a package mechanical performance testing terminal. The mechanical performance data from each round is recorded to form multiple sets of mechanical performance test data. The multiple sets of mechanical performance test data are then normalized using the same standardization process as the target mechanical performance vector set. The standardized mechanical performance test data are then aggregated and combined into a vector, namely the dth package mechanical performance vector set, which reflects the mechanical performance of the dth package. Using the target mechanical property vector set as the package's mechanical property constraint, a deviation comparison is performed on the dth package's mechanical property vector set. The comparison results are combined into a new vector set, the dth package's mechanical property deviation vector set. A depth evaluation of mechanical property deviations is performed based on the dth package's mechanical property deviation vector set, generating a mechanical property deviation depth coefficient. The mechanical property deviation depth coefficient is used to quantify the extent and significance of the mechanical property defects of the dth package. The dth package's mechanical property deviation vector set and the mechanical property deviation depth coefficient are added to the dth mechanical performance defect detection results.

[0126] performing data fusion based on the dth dominant defect detection result, the dth electrical performance defect detection result, and the dth mechanical performance defect detection result to generate a dth package comprehensive defect detection result;

[0127] Assign weights to the dth visible defect detection result, the dth electrical performance defect detection result, and the dth mechanical performance defect detection result to ensure that they reflect the impact of different defect types on the overall performance of the package. The weighted summation yields the dth package comprehensive defect detection result, which reflects the overall defect situation.

[0128] Continue to perform defect detection on each package in the package distribution set according to the semiconductor package explicit defect detection channel, the package electrical performance benchmark matrix and the package mechanized joint detection terminal, and generate a package comprehensive defect distribution set in combination with the comprehensive defect detection result of the dth package.

[0129] Based on the semiconductor package explicit defect detection channel, the package electrical performance benchmark matrix and the package mechanized joint detection terminal, defect detection is continued on each package in the package distribution set, and the comprehensive defect detection results of all packages are summarized to form a package comprehensive defect distribution set. The package comprehensive defect distribution set represents the defects of the package distribution set.

[0130] In summary, the embodiments of the present application have at least the following technical effects:

[0131] A package distribution set includes multiple packages corresponding to a target semiconductor packaging process solution. The dth package is extracted from the package distribution set, where d is a positive integer greater than or equal to 1, belongs to D, and D is the total number of the multiple packages. A multi-feature dominant defect detection channel is used to perform dominant defect detection on the dth package to obtain a dth dominant defect detection result. An electrical performance benchmark is calculated based on the target semiconductor packaging process solution, and a package electrical performance benchmark matrix is established. Then, the package electrical performance benchmark matrix is used to perform electrical performance defect detection on the dth package to obtain a dth electrical performance defect detection result. A package-mechanized combined detection terminal is used to perform mechanical performance defect detection and analysis on the dth package to obtain a dth mechanical performance defect detection result. The package-mechanized combined detection terminal includes a package chemical performance detection terminal and a package mechanical performance detection terminal. The dth dominant defect detection result, the dth electrical performance defect detection result, and the dth mechanical performance defect detection result are data fused to generate a comprehensive defect detection result for the dth package. Based on the semiconductor package explicit defect detection channel, the package electrical performance benchmark matrix, and the package mechanized joint detection terminal, defect detection is performed on other packages within the package distribution set. Combined with the comprehensive defect detection results of the dth package, a comprehensive package defect distribution set is generated. This solves the technical problem of inaccurate test results caused by insufficiently comprehensive semiconductor package defect detection in existing technologies, achieving the technical effect of improving the accuracy of test results.

[0132] It should be noted that the above-mentioned order of the embodiments of the present application is for descriptive purposes only and does not represent the superiority or inferiority of the embodiments. The above description is of specific embodiments of this specification. Other embodiments are within the scope of the appended claims. In some cases, the actions or steps described in the claims can be performed in an order different from that in the embodiments and still achieve the desired results. In addition, the processes depicted in the accompanying drawings do not necessarily require the specific order or continuous order shown to achieve the desired results. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0133] The above description is only a preferred embodiment of the present application and is not intended to limit the present application. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principles of the present application shall be included in the scope of protection of the present application.

[0134] This specification and drawings are merely illustrative of the present application and are intended to cover any and all modifications, variations, combinations, or equivalents within the scope of this application. Obviously, those skilled in the art may make various modifications and variations to this application without departing from the scope of this application. Thus, this application is intended to include such modifications and variations as fall within the scope of this application and its equivalents.

Claims

1. Semiconductor packaging defect automatic detection system, characterized in that: The system comprises: an extraction module configured to extract a dth package according to a package distribution set, wherein the package distribution set includes a plurality of packages corresponding to a target semiconductor packaging process solution, wherein d is a positive integer greater than or equal to 1, d belongs to D, and D is a total number of the plurality of packages; a first detection module, configured to perform a multi-feature dominant defect detection on the dth package according to a semiconductor package dominant defect detection channel to obtain a dth dominant defect detection result; a second detection module, configured to perform electrical performance benchmark calculation according to the target semiconductor packaging process solution, build a package electrical performance benchmark matrix, and perform electrical performance defect detection on the dth package based on the package electrical performance benchmark matrix to obtain a dth electrical performance defect detection result; a third detection module, configured to perform mechanical performance defect detection and analysis on the dth package according to a package-mechanized combined detection terminal to obtain a dth mechanical performance defect detection result, wherein the package-mechanized combined detection terminal includes a package chemical performance detection terminal and a package mechanical performance detection terminal; a data fusion module, configured to fuse data based on the dth dominant defect detection result, the dth electrical performance defect detection result, and the dth mechanical performance defect detection result to generate a dth package comprehensive defect detection result; A package detection module is configured to continue to perform defect detection on each package in the package distribution set based on the semiconductor package dominant defect detection channel, the package electrical performance benchmark matrix, and the package mechanized joint detection terminal, and generate a package comprehensive defect distribution set in combination with the comprehensive defect detection result of the dth package.

2. The system according to claim 1, wherein The first detection module includes: A channel composition module, wherein the channel composition module is used for the semiconductor package dominant defect detection channel and includes a detection data acquisition branch, an acquisition data cleaning branch, and a dominant defect detection branch; a data acquisition module, configured to acquire data from the d-th package using M heterogeneous detection and acquisition devices within the detection data acquisition branch, to obtain M package acquisition data, where M is a positive integer greater than 1; a data processing module, the data processing module being configured to input the M package collected data into the collected data cleaning branch, and pre-process the M package collected data according to the collected data denoising algorithm in the collected data cleaning branch to obtain M collected cleaned data; a defect recognition module, configured to perform defect recognition on the M collected and cleaned data according to the explicit defect detection branch, and generate M defect recognition results; A defect recognition result fusion module is used to perform data fusion on the M defect recognition results to generate the d-th dominant defect detection result.

3. The system according to claim 2, wherein: The defect identification module includes: a first backtracking module, configured to backtrack package defect detection records based on the M heterogeneous detection and collection devices to obtain M sets of package defect detection record sets; a supervised learning module, configured to perform supervised learning on the M sets of package defect detection record sets using a preset defect detection accuracy threshold as a supervised learning constraint, to generate the explicit defect detection branch, wherein the explicit defect detection branch includes M explicit defect detectors; The overt defect detection module is used to input the M collected and cleaned data into the M overt defect detectors to obtain the M defect recognition results.

4. The system according to claim 1, wherein: The second detection module includes: a second backtracking module, configured to backtrack electrical performance test records of normal sample packages according to the target semiconductor packaging process solution to obtain a normal sample electrical performance test record set; A first standardization processing module, configured to perform standardization processing on the normal sample electrical performance test record set to obtain a standard normal electrical performance test record set; A clustering module, the clustering module is used to perform clustering according to the standard normal electrical performance detection record set to obtain multiple normal electrical performance detection intervals; a first calculation module, configured to perform centralized value calculations on the plurality of normal electrical performance detection intervals to generate a plurality of normal electrical performance detection reference values; A matrix construction module is used to construct the package electrical performance reference matrix according to the multiple normal electrical performance detection reference values.

5. The system according to claim 1, wherein: The second detection module includes: an electrical performance detection module, configured to perform multiple rounds of electrical performance detection on the dth package according to the package electrical performance detection terminal to obtain multiple sets of electrical performance detection data; A second standardization processing module, configured to perform standardization processing on the plurality of electrical performance test data groups to generate a plurality of electrical performance test standard groups; a second calculation module, configured to perform centralized value calculation based on the plurality of electrical performance test standard groups to obtain a package electrical performance test matrix, wherein the package electrical performance test matrix includes a plurality of electrical performance test element values; a deviation calculation module, configured to perform deviation calculation on the package electrical performance detection matrix based on the package electrical performance reference matrix to generate a package electrical performance deviation matrix; a first deviation depth evaluation module, configured to perform an electrical performance deviation depth evaluation based on the package electrical performance deviation matrix to generate an electrical performance defect coefficient; A first adding module is configured to add the package electrical performance deviation matrix and the electrical performance defect coefficient to the dth electrical performance defect detection result.

6. The system according to claim 1, wherein: The third detection module includes: a chemical property prediction module, configured to predict the chemical properties of the package according to the target semiconductor packaging process solution, obtain multi-dimensional target chemical property prediction results, and perform standardization processing on the multi-dimensional target chemical property prediction results to generate a target chemical property vector set; a chemical property detection module, configured to perform multiple rounds of chemical property detection on the dth package according to the package chemical property detection terminal to obtain multiple groups of chemical property detection data; a first processing module, configured to perform standardization processing and centralized value calculation on the plurality of chemical property detection data sets to obtain a dth package chemical property vector set; a first deviation comparison module, configured to perform deviation comparison on the dth package chemical property vector set using the target chemical property vector set as a package chemical property constraint, to generate a dth package chemical property deviation vector set; a second deviation depth evaluation module, configured to perform a chemical property deviation depth evaluation based on the dth package chemical property deviation vector set to generate a dth chemical property deviation depth coefficient; A second adding module is configured to add the dth package chemical property deviation vector set and the dth chemical property deviation depth coefficient to the dth chemical property defect detection result.

7. The system according to claim 1, wherein: The third detection module includes: a mechanical property prediction module, configured to predict the mechanical properties of a package according to the target semiconductor packaging process solution, obtain multi-dimensional target mechanical property prediction results, and perform standardization processing on the multi-dimensional target mechanical property prediction results to generate a target mechanical property vector set; a mechanical property detection module, configured to perform multiple rounds of mechanical property detection on the dth package according to the package mechanical property detection terminal to obtain multiple sets of mechanical property detection data; a second processing module, configured to perform standardization processing and centralized value calculation on the plurality of mechanical property test data sets to obtain a dth package mechanical property vector set; a second deviation comparison module, configured to perform deviation comparison on the d-th package mechanical property vector set using the target mechanical property vector set as a package mechanical property constraint, to generate a d-th package mechanical property deviation vector set; a third deviation depth evaluation module, configured to perform a mechanical property deviation depth evaluation based on the dth package mechanical property deviation vector set to generate a dth mechanical property deviation depth coefficient; A third adding module is configured to add the dth package mechanical property deviation vector set and the dth mechanical property deviation depth coefficient to the dth mechanical property defect detection result.

8. A method for automatically detecting semiconductor packaging defects, characterized in that: The method comprises: Extracting a d-th package according to a package distribution set, wherein the package distribution set includes multiple packages corresponding to a target semiconductor packaging process solution, wherein d is a positive integer greater than or equal to 1, d belongs to D, and D is the total number of the multiple packages; performing a multi-feature dominant defect detection on the dth package according to a semiconductor package dominant defect detection channel to obtain a dth dominant defect detection result; performing an electrical performance benchmark calculation according to the target semiconductor packaging process solution, building a package electrical performance benchmark matrix, and performing an electrical performance defect detection on the dth package based on the package electrical performance benchmark matrix to obtain a dth electrical performance defect detection result; performing mechanical performance defect detection and analysis on the dth package according to the package mechanization combined detection terminal to obtain a dth mechanical performance defect detection result, wherein the package mechanization combined detection terminal includes a package chemical performance detection terminal and a package mechanical performance detection terminal; performing data fusion based on the dth dominant defect detection result, the dth electrical performance defect detection result, and the dth mechanical performance defect detection result to generate a dth package comprehensive defect detection result; Continue to perform defect detection on each package in the package distribution set according to the semiconductor package explicit defect detection channel, the package electrical performance benchmark matrix and the package mechanized joint detection terminal, and generate a package comprehensive defect distribution set in combination with the comprehensive defect detection result of the dth package.

Images