Semiconductor packaging defect automatic detection system and method

Abstract

The invention discloses a semiconductor packaging defect automatic detection system and method, and relates to the technical field of semiconductor packaging detection. Comprising the following steps: extracting a d-th packaging body according to a packaging body distribution set; performing multi-feature dominant defect detection on the d-th packaging body according to the semiconductor packaging dominant defect detection channel to obtain a d-th dominant defect detection result; performing electrical performance defect detection on the d-th packaging body to obtain a d-th electrical performance defect detection result; carrying out mechanical performance defect detection analysis on the d-th packaging body to obtain a d-th mechanical performance defect detection result; performing data fusion to generate a comprehensive defect detection result of the d-th packaging body; and performing defect detection on each packaging body in the packaging body distribution set, and generating a packaging body comprehensive defect distribution set in combination with a comprehensive defect detection result of the d-th packaging body. The technical problem that the detection result is inaccurate due to the fact that semiconductor packaging defect detection is not comprehensive enough in the prior art is solved, and the technical effect of improving the accuracy of the detection result is achieved.

Description

Technical Field

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

[0002] Semiconductor packaging is the process of encapsulating a chip in a protective housing. The main purpose of packaging is to protect the chip from physical damage and environmental influences, and to provide a way to connect the chip to external circuits. Some defects may occur during the packaging process, such as poor welding, pin misalignment, and packaging material cracks. These defects may cause chip performance degradation, functional failure, or even complete failure. Therefore, timely and accurate detection of packaging defects is crucial. The existing technology for semiconductor packaging defect detection is often not comprehensive enough, resulting in technical problems such as inaccurate detection results. Summary of the invention

[0003] The embodiments of the present application provide a system and method for automatically detecting semiconductor package defects, which solve 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: An extraction module, the extraction module is used 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; a first detection module, wherein the first detection module is used to perform a multi-feature dominant defect detection on the dth package body according to a semiconductor package dominant defect detection channel to obtain a dth dominant defect detection result; a second detection module, the second detection module being used 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, the third detection module is used to perform mechanical performance defect detection and analysis on the d-th package according to the package mechanization joint detection end to obtain a d-th mechanical performance defect detection result, wherein the package mechanization joint detection end includes a package chemical performance detection end and a package mechanical performance detection end; A data fusion module, the data fusion module is used to perform data fusion according to 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 body comprehensive defect detection result; A package detection module, the package detection module is used to 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 reference matrix and the package mechanized joint detection terminal, and generate a package comprehensive defect distribution set in combination with the dth package comprehensive defect detection result.

[0006] A second aspect of an embodiment of the present application provides a method for automatically detecting semiconductor packaging defects, the method comprising: Extracting 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; Performing a multi-feature dominant defect detection on the dth package body according to a semiconductor package dominant defect detection channel to obtain a dth dominant defect detection result; Performing electrical performance benchmark calculation according to the target semiconductor packaging process solution, building a package electrical performance benchmark matrix, and performing 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 joint detection end to obtain a dth mechanical performance defect detection result, wherein the package mechanization joint detection end includes a package chemical performance detection end and a package mechanical performance detection end; Performing data fusion according to 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 body comprehensive defect detection result; According to the semiconductor package explicit defect detection channel, the package electrical performance reference matrix and the package mechanized joint detection terminal, defect detection is continued on each package in the package distribution set, and combined with the comprehensive defect detection result of the dth package, a package comprehensive defect distribution set is generated.

[0007] One or more technical solutions provided in this application have at least the following technical effects or advantages: The package distribution set includes multiple packages corresponding to the target semiconductor packaging process scheme, and the dth package is extracted from the package distribution set, 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. The semiconductor package dominant defect detection channel is used to perform multi-feature dominant defect detection on the dth package to obtain the dth dominant defect detection result. The electrical performance benchmark is calculated according to the target semiconductor packaging process scheme, 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 the dth electrical performance defect detection result. The package mechanization joint detection end is used to perform mechanization performance defect detection analysis on the dth package to obtain the dth mechanization performance defect detection result, wherein the package mechanization joint detection end includes a package chemical performance detection end and a package mechanical performance detection end. The dth dominant defect detection result, the dth electrical performance defect detection result, and the dth mechanization performance defect detection result are data fused to generate the dth package comprehensive defect detection result. According to 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 in the package distribution set, and the package comprehensive defect distribution set is generated by combining the comprehensive defect detection result of the dth package. The technical problem of inaccurate detection results caused by insufficient comprehensiveness of semiconductor package defect detection in the prior art is solved, and the technical effect of improving the accuracy of the detection results is achieved. BRIEF DESCRIPTION OF THE DRAWINGS

[0008] 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.

[0009] Figure 1 A schematic diagram of the structure of a semiconductor packaging defect automatic detection system provided in an embodiment of the present application; Figure 2 A schematic flow chart of a method for automatically detecting semiconductor packaging defects provided in an embodiment of the present application.

[0010] Explanation of reference numerals: 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

[0011] 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 semiconductor package defect automatic detection system and method.

[0012] The following will be combined with the drawings in the embodiments of the present application to clearly and completely describe the technical solutions in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of this application.

[0013] 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 explicitly listed, but may include other steps or modules that are not explicitly listed or are inherent to these processes, methods, products or devices.

[0014] Embodiment 1 like Figure 1 As shown, the present application provides a semiconductor packaging defect automatic detection system, the system is used to perform a semiconductor packaging defect automatic detection method. The system includes: An extraction module 11, the extraction module 11 is used 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; A first detection module 12, wherein the first detection module 12 is used to perform a multi-feature dominant defect detection on the dth package body according to a semiconductor package dominant defect detection channel to obtain a dth dominant defect detection result; A second detection module 13, the second detection module 13 is used 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 14, the third detection module 14 is used to perform mechanical performance defect detection and analysis on the d-th package according to the package mechanization joint detection end to obtain a d-th mechanical performance defect detection result, wherein the package mechanization joint detection end includes a package chemical performance detection end and a package mechanical performance detection end; A data fusion module 15, the data fusion module 15 is used to perform data fusion according to 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 body comprehensive defect detection result; The package detection module 16 is used to 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 reference matrix and the package mechanized joint detection terminal, and generate a package comprehensive defect distribution set in combination with the dth package comprehensive defect detection result.

[0015] Furthermore, the first detection module 12 is used to perform the following method: The semiconductor package explicit defect detection channel includes a detection data acquisition branch, an acquisition data cleaning branch and an explicit defect detection branch; Performing data collection on the d-th package according to the 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; Inputting the M package body collected data into the collected data cleaning branch, and preprocessing the M package body collected data according to the collected data denoising algorithm in the collected data cleaning branch to obtain M collected cleaning data; Performing defect recognition on the M collected cleaning data according to the explicit defect detection branch to generate M defect recognition results; Data fusion is performed on the M defect recognition results to generate the d-th dominant defect detection result.

[0016] Furthermore, the first detection module 12 is used to perform the following method: Perform package defect detection record backtracking according to the M heterogeneous detection and collection devices to obtain M sets of package defect detection record sets; Taking a preset defect detection accuracy threshold as a supervised learning constraint, respectively performing supervised learning on the M groups of package defect detection record sets to generate the explicit defect detection branch, wherein the explicit defect detection branch includes M explicit defect detectors; The M collected cleaning data are input into the M explicit defect detectors to obtain the M defect recognition results.

[0017] Furthermore, the second detection module 13 is used to perform the following method: Conducting electrical performance test record backtest of normal sample packages according to the target semiconductor packaging process scheme to obtain a normal sample electrical performance test record set; Performing standardization processing on the normal sample electrical performance test record set to obtain a standard normal electrical performance test record set; Clustering is performed according to the standard normal electrical performance detection record set to obtain multiple normal electrical performance detection intervals; Performing concentrated value calculations on the plurality of normal electrical performance detection intervals respectively to generate a plurality of normal electrical performance detection reference values; The package electrical performance reference matrix is ​​constructed according to the plurality of normal electrical performance detection reference values.

[0018] Furthermore, the second detection module 13 is used to perform the following method: Performing multiple rounds of electrical performance testing on the dth package according to the package electrical performance testing terminal to obtain multiple groups of electrical performance testing data groups; Performing standardization processing on the plurality of electrical performance test data groups to generate a plurality of electrical performance test standard groups; Performing concentrated value calculation according to the plurality of electrical performance detection standard groups to obtain a package electrical performance detection matrix, wherein the package electrical performance detection matrix includes a plurality of electrical performance detection element values; 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; Performing an electrical performance deviation depth evaluation based on the package electrical performance deviation matrix to generate an electrical performance defect coefficient; The package electrical performance deviation matrix and the electrical performance defect coefficient are added to the dth electrical performance defect detection result.

[0019] Furthermore, the third detection module 14 is used to perform the following method: Predicting the chemical properties of the package according to the target semiconductor packaging process scheme to obtain a multi-dimensional target chemical property prediction result, and performing standardization processing on the multi-dimensional target chemical property prediction result to generate a target chemical property vector set; Perform multiple rounds of chemical property testing on the dth package according to the package chemical property testing end to obtain multiple groups of chemical property testing data groups; Performing standardization processing and concentrated value calculation on the plurality of chemical property test data groups to obtain a d-th package chemical property vector set; Taking the target chemical property vector set as the chemical property constraint of the package, performing deviation comparison on the d-th package chemical property vector set to generate a d-th package chemical property deviation vector set; Performing a chemical property deviation depth evaluation according to the d-th package chemical property deviation vector set to generate a d-th chemical property deviation depth coefficient; 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.

[0020] Furthermore, the third detection module 14 is used to perform the following method: 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 performing standardization processing on the multi-dimensional target mechanical property prediction result to generate a target mechanical property vector set; Performing multiple rounds of mechanical property testing on the dth package according to the package mechanical property testing end to obtain multiple groups of mechanical property testing data groups; Performing standardization processing and concentrated value calculation on the plurality of sets of mechanical property test data groups to obtain a d-th package mechanical property vector set; Taking the target mechanical property vector set as the mechanical property constraint of the package, performing deviation comparison on the d-th package mechanical property vector set to generate a d-th package mechanical property deviation vector set; Performing a mechanical property deviation depth evaluation according to the d-th package mechanical property deviation vector set to generate a d-th mechanical property deviation depth coefficient; 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.

[0021] Embodiment 2 like Figure 2 As shown, an embodiment of the present application provides a method for automatically detecting semiconductor packaging defects, wherein the method comprises: Extracting 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; The package distribution set includes multiple packages corresponding to the target semiconductor packaging process solution, and 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.

[0022] Performing a multi-feature dominant defect detection on the dth package body according to a semiconductor package dominant defect detection channel to obtain a dth dominant defect detection result; The dth package is subjected to multi-feature explicit defect detection through the semiconductor package explicit defect detection channel. The multi-feature explicit defect detection can comprehensively and accurately capture various physical defects on the surface of the package, thereby obtaining the dth explicit defect detection result. The dth explicit defect detection result refers to the physical defect detection result on the surface of the package.

[0023] Further, performing multi-feature dominant defect detection on the dth package body according to the semiconductor package dominant defect detection channel to obtain a dth dominant defect detection result includes: The semiconductor package explicit defect detection channel includes a detection data acquisition branch, an acquisition data cleaning branch and an explicit defect detection branch; Performing data collection on the d-th package according to the 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; Inputting the M package body collected data into the collected data cleaning branch, and preprocessing the M package body collected data according to the collected data denoising algorithm in the collected data cleaning branch to obtain M collected cleaning data; Performing defect recognition on the M collected cleaning data according to the explicit defect detection branch to generate M defect recognition results; Data fusion is performed on the M defect recognition results to generate the d-th dominant defect detection result.

[0024] The semiconductor package explicit defect detection channel includes a detection data acquisition branch, an acquisition data cleaning branch, and an explicit defect detection branch. The detection data acquisition branch is composed of M heterogeneous detection and acquisition devices, which are responsible for collecting the original data of the package. The heterogeneous detection and acquisition devices include laser scanners, optical microscopes, infrared thermal imaging devices, etc. M heterogeneous detection and acquisition devices are used to collect data on the dth package to obtain M package acquisition data, where M is a positive integer greater than 1. The acquisition data cleaning branch preprocesses the collected original data to eliminate noise and improve data quality. The M package acquisition data are input into the acquisition data cleaning branch, and the M package acquisition data are preprocessed according to the acquisition data denoising algorithm in the acquisition data cleaning branch, such as wavelet transform algorithm, median filter algorithm, median filter algorithm, etc., to obtain M acquisition cleaning data. The explicit defect detection branch is responsible for defect recognition of the cleaned data, and the M acquisition cleaning data are input into the explicit defect detection branch for defect recognition to generate M defect recognition results. Since multiple heterogeneous detection and acquisition devices are used, each device may capture information about the package from different angles or conditions, so their recognition results may be different or complementary. The M defect recognition results are fused to generate a comprehensive d-th dominant defect detection result.

[0025] Furthermore, defect recognition is performed on the M collected cleaning data according to the explicit defect detection branch to generate M defect recognition results, including: Perform package defect detection record backtracking according to the M heterogeneous detection and collection devices to obtain M sets of package defect detection record sets; Taking a preset defect detection accuracy threshold as a supervised learning constraint, respectively performing supervised learning on the M groups of package defect detection record sets to generate the explicit defect detection branch, wherein the explicit defect detection branch includes M explicit defect detectors; The M collected cleaning data are input into the M explicit defect detectors to obtain the M defect recognition results.

[0026] According to the detection history of M heterogeneous detection and acquisition devices, the package defect detection record is traced back to obtain M sets of package defect detection record sets, which record the historical package defect detection results. In order to generate the explicit defect detection branch, supervised learning is required. Specifically, data with clear defect labels are extracted from the M sets of package defect detection record sets, and a preset defect detection accuracy threshold is set to evaluate the performance of the model during the training process. The M explicit defect detectors are trained separately using labeled training data. Each explicit defect detector corresponds to a heterogeneous detection and acquisition device, and learns how to identify explicit defects from the collected data of the device. During the training process, the parameters and structure of the explicit defect detector are continuously adjusted to optimize its performance until the preset defect detection accuracy threshold is reached or exceeded. The M collected cleaned data are respectively input into the corresponding M explicit defect detectors. Each explicit defect detector processes the input collected cleaned data and outputs a defect recognition result. Finally, M defect recognition results will be obtained.

[0027] Performing electrical performance benchmark calculation according to the target semiconductor packaging process solution, building a package electrical performance benchmark matrix, and performing 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; The target semiconductor packaging process solution includes the type, structure, material and electrical performance parameters of the package. The electrical performance benchmark is calculated according to the target semiconductor packaging process solution to obtain the benchmark values ​​of each key electrical performance parameter, thereby building a package electrical performance benchmark matrix, which will serve as the basis for electrical performance defect detection. The dth package is tested for electrical performance defects and compared with the package electrical performance benchmark matrix to obtain the dth electrical performance defect detection result.

[0028] Furthermore, the electrical performance benchmark is calculated according to the target semiconductor packaging process solution, and a package electrical performance benchmark matrix is ​​constructed, including: Conducting electrical performance test record backtest of normal sample packages according to the target semiconductor packaging process scheme to obtain a normal sample electrical performance test record set; Performing standardization processing on the normal sample electrical performance test record set to obtain a standard normal electrical performance test record set; Clustering is performed according to the standard normal electrical performance detection record set to obtain multiple normal electrical performance detection intervals; Performing concentrated value calculations on the plurality of normal electrical performance detection intervals respectively to generate a plurality of normal electrical performance detection reference values; The package electrical performance reference matrix is ​​constructed according to the plurality of normal electrical performance detection reference values.

[0029] The electrical performance test records of the packages that were previously produced according to the target semiconductor packaging process plan and verified to be normal are traced back to obtain a normal sample electrical performance test record set, which contains normal electrical performance parameters. The collected normal sample electrical performance test record set is cleaned to remove outliers, missing values ​​or duplicate data, and then the cleaned data is standardized to eliminate the dimensional differences between different parameters, thereby obtaining a standard normal electrical performance test record set. A clustering algorithm, such as K-means clustering, hierarchical clustering, etc., is selected to cluster the standard normal electrical performance test record set, and similar data points are divided into the same category to form multiple normal electrical performance test intervals. For each normal electrical performance test interval obtained by clustering, its central value (such as mean, median, etc.) is calculated as the representative value of the interval, and these central values ​​are used as the normal electrical performance test benchmark value, reflecting the expected level of each electrical performance parameter under normal process conditions. Multiple normal electrical performance test reference values ​​are organized into a matrix form according to the corresponding electrical performance parameters and intervals. This matrix is ​​the package electrical performance reference matrix, which includes the expected range and reference value of each electrical performance parameter under different process conditions.

[0030] Further, 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: Performing multiple rounds of electrical performance testing on the dth package according to the package electrical performance testing terminal to obtain multiple groups of electrical performance testing data groups; Performing standardization processing on the plurality of electrical performance test data groups to generate a plurality of electrical performance test standard groups; Performing concentrated value calculation according to the plurality of electrical performance detection standard groups to obtain a package electrical performance detection matrix, wherein the package electrical performance detection matrix includes a plurality of electrical performance detection element values; 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; Performing an electrical performance deviation depth evaluation based on the package electrical performance deviation matrix to generate an electrical performance defect coefficient; The package electrical performance deviation matrix and the electrical performance defect coefficient are added to the dth electrical performance defect detection result.

[0031] The package electrical performance detection end is used to detect the electrical performance of the package. The package electrical performance detection end is used to perform multiple rounds of electrical performance detection on the d package, including the measurement of multiple key electrical performance parameters. Multiple rounds of electrical performance detection refer to multiple detections of an electrical performance detection indicator in each round, and multiple rounds of detection are performed until all electrical performance detections are completed. The electrical performance data obtained by the detection are recorded to form multiple groups of electrical performance detection data groups. The collected multiple groups of electrical performance detection data groups are cleaned to remove abnormal values, duplicate data or data that does not meet the requirements, and the cleaned data are standardized to generate multiple groups of electrical performance detection standard groups. The standardized multiple groups of electrical performance detection data groups are calculated with centralized values, such as calculating the mean and median of the parameters in each round of detection, and these centralized values ​​are organized into a matrix form, namely the package electrical performance detection matrix. Using the package electrical performance benchmark matrix as a reference, the deviation of each element value in the package electrical performance detection matrix is ​​calculated, and all the calculated deviation values ​​are organized into a new matrix, namely the package electrical performance deviation matrix. The package electrical performance deviation matrix reflects the electrical performance deviation of the dth package relative to the benchmark in multiple rounds of detection. Based on the package electrical performance deviation matrix, an in-depth evaluation of the electrical performance deviation is performed, such as analyzing the size, direction, trend, etc. of the deviation, and generating an electrical performance defect coefficient. The electrical performance defect coefficient is used to quantify the degree of defect in the electrical performance of the dth package. The package electrical performance deviation matrix and the electrical performance defect coefficient are added to the dth electrical performance defect detection result.

[0032] Performing mechanical performance defect detection and analysis on the dth package according to the package mechanization joint detection end to obtain a dth mechanical performance defect detection result, wherein the package mechanization joint detection end includes a package chemical performance detection end and a package mechanical performance detection end; The package mechanization joint detection end includes a package chemical property detection end and a package mechanical property detection end, which are used to detect the chemical property and mechanical property of the package. According to the package mechanization joint detection end, the d-th package is subjected to mechanization property defect detection and analysis, and the d-th mechanization property defect detection result can be obtained, and the d-th mechanization property defect detection result includes chemical property defects and mechanical property defects.

[0033] Further, the packaging mechanization joint detection end performs mechanization performance defect detection and analysis on the d-th package to obtain the d-th mechanization performance defect detection result, including: Predicting the chemical properties of the package according to the target semiconductor packaging process scheme to obtain a multi-dimensional target chemical property prediction result, and performing standardization processing on the multi-dimensional target chemical property prediction result to generate a target chemical property vector set; Perform multiple rounds of chemical property testing on the dth package according to the package chemical property testing end to obtain multiple groups of chemical property testing data groups; Performing standardization processing and concentrated value calculation on the plurality of chemical property test data groups to obtain a d-th package chemical property vector set; Taking the target chemical property vector set as the chemical property constraint of the package, performing deviation comparison on the d-th package chemical property vector set to generate a d-th package chemical property deviation vector set; Performing a chemical property deviation depth evaluation according to the d-th package chemical property deviation vector set to generate a d-th chemical property deviation depth coefficient; 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.

[0034] According to 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 obtain multi-dimensional target chemical performance prediction results, in which each dimension represents a chemical performance parameter. The multi-dimensional target chemical performance prediction results are standardized to eliminate the dimensional differences between different parameters so that each parameter can be compared and analyzed on the same scale. The standardized multi-dimensional target chemical performance prediction results form a vector set, namely the target chemical performance vector set, which will be used as a reference benchmark for chemical performance evaluation. The package chemical performance detection end is used to perform multiple rounds of chemical performance detection on the d package, and the chemical performance data obtained from each round of detection is recorded to form multiple groups of chemical performance detection data sets. The multiple groups of chemical performance detection data sets are subjected to the same standardization as the target chemical performance vector set, and the standardized chemical performance detection data are subjected to centralized value calculation, and these centralized values ​​are organized into a vector form, namely the d package chemical performance vector set, and the d package chemical performance vector set reflects the chemical performance of the d package. Taking the target chemical property vector set as the chemical property constraint of the package, the deviation comparison of the chemical property vector set of the d package is performed. The comparison results are combined into a new vector set, namely the chemical property deviation vector set of the d package. According to the chemical property deviation vector set of the d package, the depth evaluation of the chemical property deviation is performed to generate the chemical property deviation depth coefficient, which is used to quantify the degree and importance of the defect in the chemical property of the d package. The chemical property deviation vector set of the d package and the chemical property deviation depth coefficient are added to the d mechanical performance defect detection results.

[0035] Further, the packaging mechanization joint detection end performs mechanization performance defect detection and analysis on the d-th package to obtain the d-th mechanization performance defect detection result, including: 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 performing standardization processing on the multi-dimensional target mechanical property prediction result to generate a target mechanical property vector set; Performing multiple rounds of mechanical property testing on the dth package according to the package mechanical property testing end to obtain multiple groups of mechanical property testing data groups; Performing standardization processing and concentrated value calculation on the plurality of sets of mechanical property test data groups to obtain a d-th package mechanical property vector set; Taking the target mechanical property vector set as the mechanical property constraint of the package, performing deviation comparison on the d-th package mechanical property vector set to generate a d-th package mechanical property deviation vector set; Performing a mechanical property deviation depth evaluation according to the d-th package mechanical property deviation vector set to generate a d-th mechanical property deviation depth coefficient; 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.

[0036] According to the target semiconductor packaging process scheme, 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, where each dimension represents a mechanical performance parameter. The multi-dimensional target mechanical performance prediction results are standardized to eliminate the dimensional differences between different parameters so that each parameter can be compared and analyzed on the same scale. The standardized multi-dimensional target mechanical performance prediction results form a vector set, namely the target mechanical performance vector set, which will be used as a reference benchmark for mechanical performance evaluation. The package mechanical performance detection end is used to perform multiple rounds of mechanical performance detection on the d package, and the mechanical performance data obtained from each round of detection is recorded to form multiple groups of mechanical performance detection data sets. The multiple groups of mechanical performance detection data sets are subjected to the same standardized processing as the target mechanical performance vector set, and the standardized mechanical performance detection data are subjected to centralized value calculation, and these centralized values ​​are formed into a vector form, namely the d package mechanical performance vector set, which reflects the mechanical performance performance of the d package. Taking the target mechanical property vector set as the mechanical property constraint of the package, the deviation comparison is performed on the mechanical property vector set of the dth package, and the comparison results are combined into a new vector set, namely the mechanical property deviation vector set of the dth package. The depth evaluation of the mechanical property deviation is performed based on the mechanical property deviation vector set of the dth package, and the mechanical property deviation depth coefficient is generated. The mechanical property deviation depth coefficient is used to quantify the degree and importance of the defects in the mechanical properties of the dth package. The mechanical property deviation vector set of the dth package and the mechanical property deviation depth coefficient are added to the dth mechanical performance defect detection result.

[0037] Performing data fusion according to 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 body comprehensive defect detection result; Assign weight values ​​to the dth dominant defect detection result, the dth electrical performance defect detection result, and the dth mechanical performance defect detection result, respectively, to ensure that the degree of influence of different defect types on the overall performance of the package can be reflected. After weighted summation, the dth package comprehensive defect detection result is obtained, which reflects the overall defect situation.

[0038] According to the semiconductor package explicit defect detection channel, the package electrical performance reference matrix and the package mechanized joint detection terminal, defect detection is continued on each package in the package distribution set, and combined with the comprehensive defect detection result of the dth package, a package comprehensive defect distribution set is generated.

[0039] According to 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, which represents the defects of the package distribution set.

[0040] In summary, the embodiments of the present application have at least the following technical effects: The package distribution set includes multiple packages corresponding to the target semiconductor packaging process scheme, and the dth package is extracted from the package distribution set, 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. The semiconductor package dominant defect detection channel is used to perform multi-feature dominant defect detection on the dth package to obtain the dth dominant defect detection result. The electrical performance benchmark is calculated according to the target semiconductor packaging process scheme, 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 the dth electrical performance defect detection result. The package mechanization joint detection end is used to perform mechanization performance defect detection analysis on the dth package to obtain the dth mechanization performance defect detection result, wherein the package mechanization joint detection end includes a package chemical performance detection end and a package mechanical performance detection end. The dth dominant defect detection result, the dth electrical performance defect detection result, and the dth mechanization performance defect detection result are data fused to generate the dth package comprehensive defect detection result. According to 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 in the package distribution set, and the package comprehensive defect distribution set is generated by combining the comprehensive defect detection result of the dth package. The technical problem of inaccurate detection results caused by insufficient comprehensiveness of semiconductor package defect detection in the prior art is solved, and the technical effect of improving the accuracy of the detection results is achieved.

[0041] It should be noted that the above-mentioned sequence of the embodiments of the present application is only for description and does not represent the advantages and disadvantages of the embodiments. And the above-mentioned specific embodiments of this specification are described. Other embodiments are within the scope of the attached claims. In some cases, the actions or steps recorded 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 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.

[0042] The above description is only a preferred embodiment of the present application and is not intended to limit the present application. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present application shall be included in the protection scope of the present application.

[0043] This specification and drawings are merely exemplary illustrations of the present application and are deemed to cover any and all modifications, variations, combinations or equivalents within the scope of the present application. Obviously, a person skilled in the art may make various modifications and variations to the present application without departing from the scope of the present application. Thus, if these modifications and variations of the present application fall within the scope of the present application and its equivalents, the present application intends to include these modifications and variations.

Claims

1. Semiconductor packaging defect automatic detection system, characterized in that: The system comprises: An extraction module, the extraction module is used 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; a first detection module, wherein the first detection module is used to perform a multi-feature dominant defect detection on the dth package body according to a semiconductor package dominant defect detection channel to obtain a dth dominant defect detection result; a second detection module, the second detection module being used 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, the third detection module is used to perform mechanical performance defect detection and analysis on the d-th package according to the package mechanization joint detection end to obtain a d-th mechanical performance defect detection result, wherein the package mechanization joint detection end includes a package chemical performance detection end and a package mechanical performance detection end; A data fusion module, the data fusion module is used to perform data fusion according to 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 body comprehensive defect detection result; A package detection module, the package detection module is used to 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 reference matrix and the package mechanized joint detection terminal, and generate a package comprehensive defect distribution set in combination with the dth package comprehensive defect detection result.

2. The system according to claim 1, characterized in that The first detection module comprises: 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 collection module, the data collection module is used to collect data on the d-th package according to M heterogeneous detection and collection devices in the detection data collection branch to obtain M package collection data, wherein M is a positive integer greater than 1; A data processing module, the data processing module is used to input the M package body collected data into the collected data cleaning branch, and pre-process the M package body collected data according to the collected data denoising algorithm in the collected data cleaning branch to obtain M collected cleaning data; A defect recognition module, the defect recognition module is used to perform defect recognition on the M collected cleaning data according to the explicit defect detection branch to 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, characterized in that The defect recognition module comprises: A first backtracking module, the first backtracking module is used to perform package defect detection record backtracking according to the M heterogeneous detection and collection devices to obtain M sets of package defect detection record sets; A supervised learning module, wherein the supervised learning module is used to perform supervised learning on the M groups of package defect detection record sets respectively with 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 explicit defect detection module is used to input the M collected and cleaned data into the M explicit defect detectors to obtain the M defect recognition results.

4. The system according to claim 1, characterized in that The second detection module includes: A second backtracking module, the second backtracking module is used to backtrack the electrical performance test records of the normal sample package according to the target semiconductor packaging process scheme to obtain a normal sample electrical performance test record set; A first standardization processing module, the first standardization processing module is used to perform standardization processing according to the normal sample electrical performance detection record set to obtain a standard normal electrical performance detection 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, the first calculation module is used to perform centralized value calculations on the multiple normal electrical performance detection intervals respectively to generate multiple 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, the electrical performance detection module is used to perform multiple rounds of electrical performance detection on the d-th package according to the package electrical performance detection terminal to obtain multiple groups of electrical performance detection data groups; A second standardization processing module, the second standardization processing module is used to perform standardization processing according to the multiple groups of electrical performance detection data groups to generate multiple groups of electrical performance detection standard groups; A second calculation module, the second calculation module is used to perform concentrated value calculation according to the multiple groups of electrical performance detection standard groups to obtain a package electrical performance detection matrix, wherein the package electrical performance detection matrix includes multiple electrical performance detection element values; A deviation calculation module, the deviation calculation module is used 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, the first deviation depth evaluation module is used 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 used 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 comprises: A chemical property prediction module, the chemical property prediction module is used to predict the chemical properties of the package according to the target semiconductor packaging process scheme, obtain a multi-dimensional target chemical property prediction result, and perform standardization processing on the multi-dimensional target chemical property prediction result to generate a target chemical property vector set; A chemical property detection module, the chemical property detection module is used to perform multiple rounds of chemical property detection on the dth package according to the package chemical property detection end to obtain multiple groups of chemical property detection data groups; A first processing module, the first processing module is used to perform standardization processing and centralized value calculation according to the multiple groups of chemical property detection data groups to obtain a d-th package chemical property vector set; a first deviation comparison module, the first deviation comparison module being used to perform deviation comparison on the d-th package chemical property vector set by taking the target chemical property vector set as the package chemical property constraint, and generating a d-th package chemical property deviation vector set; A second deviation depth evaluation module, the second deviation depth evaluation module is used to perform a chemical property deviation depth evaluation according to the d-th package chemical property deviation vector set to generate a d-th chemical property deviation depth coefficient; A second adding module, the second adding module is used 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, characterized in that The third detection module comprises: A mechanical property prediction module, the mechanical property prediction module is used to predict the mechanical properties of the package according to the target semiconductor packaging process scheme, obtain a multi-dimensional target mechanical property prediction result, and perform standardization processing on the multi-dimensional target mechanical property prediction result to generate a target mechanical property vector set; A mechanical property detection module, the mechanical property detection module is used to perform multiple rounds of mechanical property detection on the dth package according to the package mechanical property detection end to obtain multiple groups of mechanical property detection data groups; A second processing module, the second processing module is used to perform standardization processing and concentrated value calculation according to the multiple groups of mechanical property detection data groups to obtain a d-th package mechanical property vector set; a second deviation comparison module, the second deviation comparison module being used to perform deviation comparison on the d-th package mechanical property vector set by taking the target mechanical property vector set as the package mechanical property constraint, and generating a d-th package mechanical property deviation vector set; A third deviation depth evaluation module, the third deviation depth evaluation module is used to perform a mechanical property deviation depth evaluation according to the dth package mechanical property deviation vector set to generate a dth mechanical property deviation depth coefficient; A third adding module is used 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 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; Performing a multi-feature dominant defect detection on the dth package body according to a semiconductor package dominant defect detection channel to obtain a dth dominant defect detection result; Performing electrical performance benchmark calculation according to the target semiconductor packaging process solution, building a package electrical performance benchmark matrix, and performing 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 joint detection end to obtain a dth mechanical performance defect detection result, wherein the package mechanization joint detection end includes a package chemical performance detection end and a package mechanical performance detection end; Performing data fusion according to 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 body 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 reference matrix and the package mechanized joint detection terminal, and generate a package comprehensive defect distribution set in combination with the dth package comprehensive defect detection result.

Images