Sample sampling method, device and storage medium for production defect analysis

Abstract

This paper discloses a sample sampling method for production defect analysis, including: constructing an original sample space according to the production inspection data of the target production object; traversing the original sample space according to the defect type to screen out the target negative sample and the original positive sample, and generating an original positive sample set; for any target negative sample, selecting a positive sample that has an association relationship with the target negative sample as a candidate positive sample and generating a candidate positive sample set, determining the importance of the candidate positive sample in the defect analysis, and determining the extraction weight of the candidate positive sample according to the importance; performing importance random sampling on the candidate positive sample set according to the extraction weight of the candidate positive sample to obtain the target positive sample; generating a new sample set for production defect analysis from multiple target positive samples and multiple target negative samples. This paper can reduce the number of samples used for production defect analysis and improve the effectiveness of samples.

Description

Technical Field

[0001] This article relates to but is not limited to the field of factory manufacturing automation technology, and in particular to a sample sampling method, device and storage medium for production defect analysis. Background Art

[0002] In the semiconductor and panel industries, due to the complicated processes and the large number of production units, relying on manpower to locate the causes of defects has extremely limited processing time and accuracy, making it difficult to meet the growing production needs.

[0003] When big data technology is applied to factory manufacturing, automatic diagnosis and analysis can be performed through data mining methods. However, if the data generated by each manufacturing process of the entire factory is used, the huge amount of data and the sparseness of negative samples (bad samples) will greatly increase the difficulty of statistical modeling and analysis, and the analysis results will also be inaccurate. Summary of the invention

[0004] In a first aspect, an embodiment of the present disclosure provides a sample sampling method for production failure analysis, comprising:

[0005] Constructing an original sample space of the target production object according to production inspection data of the target production object for production defect analysis;

[0006] Traversing the original sample space according to the defect type for production defect analysis, and screening samples belonging to the defect type as target negative samples;

[0007] The samples in the original sample space that do not belong to the bad type are taken as original positive samples, and an original positive sample set is generated from multiple original positive samples; for any target negative sample, positive samples that are associated with the target negative sample are selected from the original positive sample set as candidate positive samples, and a candidate positive sample set is generated from the candidate positive samples of all target negative samples;

[0008] Determine the importance of each candidate positive sample in the candidate positive sample set in the defect analysis, and determine the extraction weight of the candidate positive sample according to the importance; perform importance random sampling on the candidate positive sample set according to the extraction weight of the candidate positive sample, and use the sampled candidate positive samples as target positive samples;

[0009] A new sample set is generated from multiple target positive samples and multiple target negative samples as a target sample set for production defect analysis.

[0010] In a second aspect, an embodiment of the present disclosure provides a sample sampling device for production failure analysis, comprising:

[0011] An original sample space construction module, configured to construct an original sample space of a target production object according to production inspection data of the target production object for production defect analysis;

[0012] A negative sample screening module is configured to traverse the original sample space according to the defect type for production defect analysis, and screen samples belonging to the defect type as target negative samples;

[0013] A positive sample set establishment module is configured to take samples in the original sample space that do not belong to the bad type as original positive samples, and generate an original positive sample set from multiple original positive samples; for any target negative sample, select positive samples that are associated with the target negative sample from the original positive sample set as candidate positive samples, and generate a candidate positive sample set from the candidate positive samples of all target negative samples;

[0014] A positive sample sampling module is configured to determine the importance of each candidate positive sample in the candidate positive sample set during the defect analysis, and determine the extraction weight of the candidate positive sample according to the importance; perform importance random sampling on the candidate positive sample set according to the extraction weight of the candidate positive sample, and use the sampled candidate positive samples as target positive samples;

[0015] The target sample set establishment module is configured to generate a new sample set from a plurality of target positive samples and a plurality of target negative samples as a target sample set for production defect analysis.

[0016] In a third aspect, an embodiment of the present disclosure provides a sample sampling device for production defect analysis, comprising: a memory and a processor, wherein the memory stores a computer program, and when the computer program is executed by the processor, the steps of the sample sampling method for production defect analysis are implemented.

[0017] In a fourth aspect, an embodiment of the present disclosure provides a computer-readable storage medium storing a computer program, which, when executed by a processor, implements the steps of the sample sampling method for production defect analysis.

[0018] The sample sampling method for production defect analysis provided by the embodiment of the present disclosure constructs an original sample space according to the production inspection data of the target production object, traverses the original sample space according to the defect type for production defect analysis, and selects samples belonging to the defect type as target negative samples; takes samples that do not belong to the defect type in the original sample space as original positive samples, and generates an original positive sample set from multiple original positive samples; for any target negative sample, selects positive samples that have an association relationship with the target negative sample from the original positive sample set as candidate positive samples, and generates a candidate positive sample set from the candidate positive samples of all target negative samples; determines the importance of each candidate positive sample in the candidate positive sample set in the defect analysis, and determines the extraction weight of the candidate positive sample according to the importance; performs importance random sampling on the candidate positive sample set according to the extraction weight of the candidate positive sample to obtain the target positive sample; generates a new sample set from multiple target positive samples and multiple target negative samples as the target sample set for production defect analysis. The above embodiment can balance the number of positive samples and negative samples by screening out positive samples related to negative samples, reduce the number of samples used for production defect analysis, and improve the effectiveness of samples. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] The accompanying drawings are used to provide an understanding of the technical solution of the present disclosure and constitute a part of the specification. Together with the embodiments of the present disclosure, they are used to explain the technical solution of the present disclosure and do not constitute a limitation on the technical solution of the present disclosure.

[0020] Figure 1 A flow chart of a sample sampling method for production defect analysis provided by an embodiment of the present disclosure;

[0021] Figure 2 A flowchart of an importance random sampling method provided by an embodiment of the present disclosure;

[0022] Figure 3 A schematic diagram of the structure of a sample sampling device for production defect analysis provided by an embodiment of the present disclosure;

[0023] Figure 4 A schematic structural diagram of another sample sampling device for production defect analysis provided in an embodiment of the present disclosure. DETAILED DESCRIPTION

[0024] In order to make the purpose, technical solutions and advantages of the present disclosure clearer, the embodiments of the present disclosure will be described in detail in conjunction with the accompanying drawings below. Note that the embodiments can be implemented in multiple different forms. A person of ordinary skill in the art can easily understand the fact that the methods and contents can be transformed into various forms without departing from the purpose and scope of the present disclosure. Therefore, the present disclosure should not be interpreted as being limited to the contents described in the following embodiments. In the absence of conflict, the embodiments in the present disclosure and the features in the embodiments can be arbitrarily combined with each other.

[0025] In the present specification, ordinal numbers such as “first”, “second” and “third” are provided to avoid confusion among constituent elements, and are not intended to limit the number.

[0026] like Figure 1 As shown, the embodiment of the present disclosure provides a sample sampling method for production defect analysis, comprising:

[0027] Step S10, constructing an original sample space of the target production object according to the production inspection data of the target production object for production defect analysis;

[0028] Step S20, traversing the original sample space according to the defect type for production defect analysis, and selecting samples belonging to the defect type as target negative samples;

[0029] Step S30, taking samples in the original sample space that do not belong to the bad type as original positive samples, and generating an original positive sample set from multiple original positive samples; for any target negative sample, selecting positive samples that are associated with the target negative sample from the original positive sample set as candidate positive samples, and generating a candidate positive sample set from the candidate positive samples of all target negative samples;

[0030] Step S40, determining the importance of each candidate positive sample in the candidate positive sample set in the defect analysis, and determining the extraction weight of the candidate positive sample according to the importance; performing importance random sampling on the candidate positive sample set according to the extraction weight of the candidate positive sample, and taking the sampled candidate positive sample as the target positive sample;

[0031] Step S50 , generating a new sample set from a plurality of target positive samples and a plurality of target negative samples as a target sample set for production defect analysis.

[0032] In some exemplary embodiments, before constructing the original sample space of the target production object, the method further includes: determining the target production object for defect analysis according to user input information.

[0033] The original sample space includes multiple samples. Exemplarily, the embodiments of the present disclosure can be used in a display panel production line; for example, it can be used to determine the correlation between each process equipment and the defect type of the display panel production line during the production process of display panels (such as liquid crystal display panels, organic light emitting diode display panels, etc.). Of course, the embodiments of the present disclosure can also be used for other products. Exemplarily, the sample includes a display panel motherboard (glass), and the display panel motherboard includes multiple display panels (panels). For example, the display panel motherboard also includes a substrate, and the multiple display panels are arranged on the substrate. Exemplarily, the substrate may include: a rigid substrate such as glass (or called a hard substrate), or a flexible substrate such as PI (Polyimide); the substrate may also include: a thin film such as a buffer layer arranged on a rigid substrate or a flexible substrate.

[0034] The “bad” mentioned in the article refers to quality defects in the samples, which may lead to reduced quality or even scrapping of the samples, or may cause the samples to need to be reworked or repaired.

[0035] Exemplarily, defects can be classified into different types according to needs. For example, they can be classified according to the direct impact of the defects on sample performance, such as bright line defects, dark line defects, and hot spot defects; or they can be classified according to the specific causes of the defects, such as signal line short circuit defects, alignment defects, etc.; or they can be classified according to the general causes of the defects, such as array process defects, color film process defects, etc.; or they can be classified according to the severity of the defects, such as defects that lead to scrapping, defects that lead to reduced quality, etc.

[0036] During the production process, bad data (negative samples) are relatively sparse, and the amount of good data (positive samples) is much larger than the amount of bad data. In defect analysis, the imbalance of positive and negative data samples may affect the accuracy of data analysis.

[0037] In some exemplary embodiments, the input information includes at least one of the following information: identification information of the production object and production time information of the production object.

[0038] The identification information of the production object may include at least one of the following information: name information of the production object and type information of the production object.

[0039] The production time information of the production object may include at least one of the following information: production batch information of the production object, and offline time and / or online time of the production object.

[0040] The type information of the production object may include at least one of the following information: major category information and minor category information.

[0041] Among them, the categories included in the major information include: products, modules or single boards. The categories included in the sub-category information of the product category include: LCD panels, OLED (Organic Light-Emitting Diode) panels, etc. The categories included in the sub-category information of the module category include: backlight modules, display modules, etc. The categories included in the sub-category information of the single board category include: main control board, display signal drive control board, etc.

[0042] The production batch information is usually a set of numbers or letters plus numbers used to identify the "batch". The production objects of the same production batch refer to a group of production objects that are produced once within a certain period of time and are completely identical in quality, structure and manufacturing method.

[0043] The offline time and / or online time of the production object may be the online time and offline time of the production object for a specific production line.

[0044] In some exemplary embodiments, the production inspection data includes: production manufacturing data related to production manufacturing generated by the production object during the production process, and inspection data generated when the production object passes through the inspection unit;

[0045] The manufacturing data may include at least one of the following information: material information, equipment information, and process information. The manufacturing data may be determined based on the actual production scenario, and may include other information in addition to the above information. The present disclosure does not limit the types of information in the manufacturing data.

[0046] Different production processes have different complexities, so the number of production equipment and testing sites involved is usually different, resulting in different amounts of production test data for different process segments. For example, the display module assembly process requires the display device, connector, integrated circuit component, and FPC (Flexible Printed Circuit) component to be assembled together, and the number of equipment and testing sites involved is relatively large, so the amount of production test data is large. The cutting process of cutting a whole piece of glass into half a piece of glass involves a relatively small number of equipment and testing sites, so the amount of production test data corresponding to this process is relatively small.

[0047] In some exemplary embodiments, before traversing the original sample space according to the defect type for production defect analysis, the method further includes: determining the defect type for defect analysis according to user input information.

[0048] For example, a user may select a type of bad behavior from a selection list of multiple types of bad behaviors.

[0049] The defect types include: poor line contact, poor point contact, bad pixels in the display area, etc.

[0050] Among them, negative samples are determined according to the bad type. For example, a sample is a negative sample of bad type A, but it may also be a positive sample of bad type B. A sample may also be a negative sample of multiple bad types at the same time.

[0051] In some exemplary embodiments, before traversing the original sample space according to the defect types for production defect analysis, the method may further include: preprocessing the sample data of the original sample space.

[0052] In some exemplary embodiments, the preprocessing of the sample data of the original sample space includes: when the data type of the sample data is numeric, setting a threshold and an enumeration type for the sample data, comparing the sample data with the threshold, and expressing the comparison result as an enumeration type.

[0053] For example, in the process of cutting glass, the geometric dimension data of the glass is a triple (x, y, z), where x represents length, y represents width, and z represents height. When preprocessing the geometric dimension data of the glass, thresholds can be set for length, width, and height respectively. After the three-dimensional data of length, width, and height are compared with the corresponding thresholds, the triple (x, y, z) can be converted into enumeration type data (thin, thick, long, short, wide, narrow, qualified) or (unqualified, qualified). When making a subsequent bad judgment, if the preprocessed geometric dimension is not "qualified", it is judged as bad. Preprocessing can speed up the subsequent traversal of the original sample space to find negative samples.

[0054] In some exemplary embodiments, selecting, from the original positive sample set, positive samples that are associated with the target negative samples as candidate positive samples according to business logic and prior knowledge, includes:

[0055] Locate one or more process sections to be checked that may produce such defects according to the defect type; for any process section to be checked, take the production unit through which the negative sample passes in the process section to be checked as the production unit to be checked;

[0056] Determine a filtering time window according to the time when the negative sample passes through the process section to be checked, and select positive samples that pass through at least one production unit to be checked within the filtering time window from the original positive sample set as candidate positive samples;

[0057] The starting time of the filtering time window is earlier than the time when the negative sample enters the process segment, and the ending time of the filtering time window is later than the time when the negative sample leaves the process segment;

[0058] Among them, there may be duplications among candidate positive samples of different target negative samples. When generating a set of candidate positive samples, repeated candidate positive samples are counted as one candidate positive sample.

[0059] In some exemplary embodiments, determining the importance of each candidate positive sample in the candidate positive sample set in the bad analysis includes:

[0060] Determining the importance of the candidate positive sample in the bad analysis according to the correlation between the candidate positive sample and one or more target negative samples, the greater the correlation, the higher the importance;

[0061] In some exemplary embodiments, the degree of association is determined in at least one of the following ways:

[0062] Determining the relevance of the candidate positive sample according to the number of production units to be checked that the candidate positive sample has passed through: the greater the number of production units to be checked that the candidate positive sample has passed through, the greater the relevance of the candidate positive sample;

[0063] Determine the relevance of the candidate positive sample according to the time difference between the candidate positive sample and the target negative sample passing through the same production unit to be checked: the shorter the time difference between the candidate positive sample and the target negative sample passing through the same production unit to be checked, the higher the relevance of the candidate positive sample;

[0064] The relevance of the candidate positive sample is determined according to the number of target negative samples associated with the candidate positive sample: when a candidate positive sample is associated with N target negative samples, the larger the value of N is, the higher the relevance of the candidate positive sample is; N is an integer greater than 1.

[0065] In some exemplary embodiments, determining the extraction weight of the candidate positive sample according to the importance includes: the higher the importance of the candidate positive sample, the greater the extraction weight of the candidate positive sample.

[0066] By performing importance random sampling on the candidate positive sample set according to the extraction weights of the candidate positive samples, the probability of extracting the candidate positive sample with a higher extraction weight is greater.

[0067] In some exemplary embodiments, Figure 2 As shown, according to the extraction weights of the candidate positive samples, the candidate positive sample set is randomly sampled by importance, and the sampled candidate positive samples are used as target positive samples, including performing the following steps a to e:

[0068] Step a, randomly arrange all candidate positive samples in the candidate positive sample set, and form a first sample sequence A[i] by the extraction weight values ​​of N candidate positive samples; wherein the value a of the i-th element of the first sample sequence is i is the extraction weight value of the i-th candidate positive sample, 1≤i≤N, N is the total number of candidate positive samples in the candidate positive sample set;

[0069] Step b, perform cumulative sum processing on the elements in the first sample sequence A[i] to generate a second sample sequence B[i]; wherein the value of the i-th element in the second sample sequence is b i ,

[0070] Step c, randomly select a real number m in the interval (0,1) and compare the real number m with the maximum value b max The product of p ref As a reference threshold, p ref =m*b max ; Among them, the maximum value b max is the maximum value of all elements in the second sample sequence B[i];

[0071] Step d, search for the target element b in the second sample sequence B[i] n , the value of the target element satisfies the following relationship: b n-1 <p ref ≤ b n ; 2≤n≤N; the target element b n The candidate positive samples represented are used as target positive samples;

[0072] Step e: Repeat steps c to d until the number of target positive samples obtained by sampling reaches a set value.

[0073] In other embodiments, the candidate positive sample set is randomly sampled according to the extraction weights of the candidate positive samples. Other implementation methods different from the above steps a to e can also be adopted. As long as the candidate positive samples with higher extraction weights have a greater probability of being extracted, the present disclosure does not limit the specific implementation method.

[0074] In some exemplary embodiments, all candidate positive samples may be sorted in descending order of extraction weights, and M2 candidate positive samples may be randomly selected from the top M1 candidate positive samples as target positive samples; wherein M2 is smaller than M1.

[0075] In some exemplary embodiments, all candidate positive samples may be sorted in descending order of extraction weights, and the extraction weights may be divided into N numerical intervals according to the numerical range, and all candidate positive samples may be divided into N groups according to the numerical ranges of the N extraction weights. After M extractions, M candidate positive samples are obtained as target positive samples: any extraction is performed in the following manner: a group is randomly selected from the N groups of candidate positive samples, and then a candidate positive sample with the highest extraction weight ranking is selected from the selected group of candidate positive samples as the target positive sample.

[0076] In some exemplary embodiments, generating a new sample set as a target sample set for production failure analysis from a plurality of target positive samples and a plurality of target negative samples includes:

[0077] A new sample set is generated from all target positive samples and all target negative samples as the target sample set; or a new sample set is generated from part of the target positive samples and all target negative samples as the target sample set; or a new sample set is generated from part of the target positive samples and part of the target negative samples as the target sample set;

[0078] Among them, some of the target positive samples may be some target positive samples with larger extraction weights from all the target positive samples.

[0079] The bad type of the negative samples in the target sample set is one bad type, that is, the bad types of the multiple negative samples included in the target sample set are the same.

[0080] The algorithm for production defect analysis can be an algorithm based on a specific machine learning model, such as one or more of decision tree, random forest, GBDT, XGBoost, naive Bayes, support vector machine, Adaboost, neural network model, etc., or it can be other statistical algorithm models, such as WOE&IV, Apriori, etc.

[0081] like Figure 3 As shown, the embodiment of the present disclosure provides a sample sampling device for production defect analysis, comprising:

[0082] The original sample space construction module 100 is configured to construct the original sample space of the target production object according to the production inspection data of the target production object for production defect analysis;

[0083] A negative sample screening module 200 is configured to traverse the original sample space according to the defect type for production defect analysis, and screen samples belonging to the defect type as target negative samples;

[0084] The positive sample set establishment module 300 is configured to take samples in the original sample space that do not belong to the bad type as original positive samples, and generate an original positive sample set from multiple original positive samples; for any target negative sample, select positive samples that are associated with the target negative sample from the original positive sample set as candidate positive samples, and generate a candidate positive sample set from the candidate positive samples of all target negative samples;

[0085] The positive sample sampling module 400 is configured to determine the importance of each candidate positive sample in the candidate positive sample set in the defect analysis, and determine the extraction weight of the candidate positive sample according to the importance; perform importance random sampling on the candidate positive sample set according to the extraction weight of the candidate positive sample, and use the sampled candidate positive samples as target positive samples;

[0086] The target sample set establishing module 500 is configured to generate a new sample set from a plurality of target positive samples and a plurality of target negative samples as a target sample set for production defect analysis.

[0087] In some exemplary embodiments, the sample sampling device may further include: a first input information processing module configured to determine a target production object for defect analysis according to user input information.

[0088] In some exemplary embodiments, the input information includes at least one of the following information: identification information of the production object and production time information of the production object.

[0089] The identification information of the production object may include at least one of the following information: name information of the production object and type information of the production object.

[0090] The production time information of the production object may include at least one of the following information: production batch information of the production object, and offline time and / or online time of the production object.

[0091] In some exemplary embodiments, the production inspection data includes: production manufacturing data related to production manufacturing generated during the production process of the production object, and inspection data generated when the production object passes through the inspection unit.

[0092] In some exemplary embodiments, the sample sampling device may further include: a second input information processing module configured to determine the defect type for defect analysis according to the user's input information.

[0093] In some exemplary embodiments, the sample sampling device may further include: a data preprocessing module configured to preprocess the sample data of the original sample space.

[0094] In some exemplary embodiments, the positive sample set establishment module is configured to locate one or more process sections to be checked that may produce such defects according to the defect type; for any process section to be checked, the production unit through which the negative sample passes in the process section to be checked is used as the production unit to be checked; a filtering time window is determined according to the time when the negative sample passes through the process section to be checked, and positive samples that pass through at least one production unit to be checked within the filtering time window are selected from the original positive sample set as candidate positive samples.

[0095] In some exemplary embodiments, the start time of the filtering time window is earlier than the time when the negative sample enters the process segment, and the end time of the filtering time window is later than the time when the negative sample leaves the process segment.

[0096] In some exemplary embodiments, the positive sample sampling module is configured to determine the importance of the candidate positive sample in the bad analysis according to the correlation between the candidate positive sample and one or more target negative samples, and the greater the correlation, the higher the importance.

[0097] In some exemplary embodiments, the positive sample sampling module is configured to determine the degree of association in at least one of the following ways:

[0098] Determining the relevance of the candidate positive sample according to the number of production units to be checked that the candidate positive sample has passed through: the greater the number of production units to be checked that the candidate positive sample has passed through, the greater the relevance of the candidate positive sample;

[0099] Determine the relevance of the candidate positive sample according to the time difference between the candidate positive sample and the target negative sample passing through the same production unit to be checked: the shorter the time difference between the candidate positive sample and the target negative sample passing through the same production unit to be checked, the higher the relevance of the candidate positive sample;

[0100] The relevance of the candidate positive sample is determined according to the number of target negative samples associated with the candidate positive sample: when a candidate positive sample is associated with N target negative samples, the larger the value of N is, the higher the relevance of the candidate positive sample is; N is an integer greater than 1.

[0101] In some exemplary embodiments, the positive sample sampling module is configured to perform the following processing steps a to e:

[0102] Step a, randomly arrange all candidate positive samples in the candidate positive sample set, and form a first sample sequence A[i] by the extraction weight values ​​of N candidate positive samples; wherein the value a of the i-th element of the first sample sequence is i is the extraction weight value of the i-th candidate positive sample, 1≤i≤N, N is the total number of candidate positive samples in the candidate positive sample set;

[0103] Step b, perform cumulative sum processing on the elements in the first sample sequence A[i] to generate a second sample sequence B[i]; wherein the value of the i-th element in the second sample sequence is b i ,

[0104] Step c, randomly select a real number m in the interval (0,1) and compare the real number m with the maximum value b max The product of p ref As a reference threshold, p ref =m*b max ; Among them, the maximum value b max is the maximum value of all elements in the second sample sequence B[i];

[0105] Step d, search for the target element b in the second sample sequence B[i] n , the value of the target element satisfies the following relationship: b n-1 <p ref ≤ b n ; 2≤n≤N; the target element b n The candidate positive samples represented are used as target positive samples;

[0106] Step e: Repeat steps c to d until the number of target positive samples obtained by sampling reaches a set value.

[0107] like Figure 4 As shown, an embodiment of the present disclosure provides a sample sampling device for production defect analysis, including: a memory and a processor, the memory stores a computer program, and when the computer program is executed by the processor, the steps of the sample sampling method for production defect analysis are implemented.

[0108] An embodiment of the present disclosure provides a computer-readable storage medium storing a computer program, which implements the steps of the sample sampling method for production defect analysis when executed by a processor.

[0109] It will be appreciated by those skilled in the art that all or some of the steps, systems, and functional modules / units in the methods disclosed above may be implemented as software, firmware, hardware, and appropriate combinations thereof. In hardware implementations, the division between the functional modules / units mentioned in the above description does not necessarily correspond to the division of physical components; for example, a physical component may have multiple functions, or a function or step may be performed by several physical components in cooperation. Some or all components may be implemented as software executed by a processor, such as a digital signal processor or a microprocessor, or implemented as hardware, or implemented as an integrated circuit, such as an application-specific integrated circuit. Such software may be distributed on a computer-readable medium, which may include a computer storage medium (or non-transitory medium) and a communication medium (or temporary medium). As known to those skilled in the art, the term computer storage medium includes volatile and non-volatile, removable and non-removable media implemented in any method or technology for storing information (such as computer-readable instructions, data structures, program modules, or other data). Computer storage media include, but are not limited to, RAM, ROM, EEPROM, flash memory or other memory technology, CD-ROM, digital versatile disks (DVD) or other optical disk storage, magnetic cassettes, magnetic tapes, magnetic disk storage or other magnetic storage devices, or any other medium that can be used to store the desired information and can be accessed by a computer. In addition, it is well known to those of ordinary skill in the art that communication media typically contain computer-readable instructions, data structures, program modules, or other data in a modulated data signal such as a carrier wave or other transport mechanism, and may include any information delivery media.

[0110] Although the embodiments disclosed in the present disclosure are as above, the contents described are only embodiments adopted to facilitate understanding of the present disclosure and are not intended to limit the present disclosure. Any technician in the field to which the present disclosure belongs can make any modifications and changes in the form and details of implementation without departing from the spirit and scope disclosed in the present disclosure, but the scope of patent protection of the present disclosure shall still be subject to the scope defined in the attached claims.

Claims

1. A method for sampling samples for production failure analysis, comprising: Constructing an original sample space of the target production object according to production inspection data of the target production object for production defect analysis; Traversing the original sample space according to the defect type for production defect analysis, and screening samples belonging to the defect type as target negative samples; Taking samples in the original sample space that do not belong to the bad type as original positive samples, and generating an original positive sample set from multiple original positive samples; For any target negative sample, a positive sample associated with the target negative sample is selected from the original positive sample set as a candidate positive sample, and a candidate positive sample set is generated from the candidate positive samples of all target negative samples; Determine the importance of each candidate positive sample in the candidate positive sample set in the defect analysis, and determine the extraction weight of the candidate positive sample according to the importance: the higher the importance of the candidate positive sample, the greater the extraction weight of the candidate positive sample; According to the extraction weights of the candidate positive samples, the candidate positive sample set is randomly sampled according to importance, so that the candidate positive samples with higher extraction weights have a greater probability of being extracted, and the sampled candidate positive samples are used as target positive samples; A new sample set is generated from multiple target positive samples and multiple target negative samples as a target sample set for production defect analysis.

2. The sample sampling method according to claim 1, characterized in that: Before constructing the original sample space of the target production object, the method further includes: determining the target production object for defect analysis according to user input information.

3. The sample sampling method according to claim 2, characterized in that: The input information includes at least one of the following information: identification information of the production object and production time information of the production object.

4. The sample sampling method according to claim 1, characterized in that: Before traversing the original sample space according to the defect type for production defect analysis, the method further includes: determining the defect type for defect analysis according to user input information.

5. The sample sampling method according to claim 1, characterized in that: The selecting, from the original positive sample set, positive samples that are associated with the target negative samples as candidate positive samples comprises: Locate one or more process sections to be checked that may produce such defects according to the defect type; for any process section to be checked, take the production unit through which the negative sample passes in the process section to be checked as the production unit to be checked; A filtering time window is determined according to the time when the negative sample passes through the process segment to be checked, and positive samples that pass through at least one production unit to be checked within the filtering time window are selected from the original positive sample set as candidate positive samples.

6. The sample sampling method according to claim 5, characterized in that: The start time of the filtering time window is earlier than the time when the negative sample enters the process segment, and the end time of the filtering time window is later than the time when the negative sample leaves the process segment.

7. The sample sampling method according to claim 1, characterized in that: Determining the importance of each candidate positive sample in the candidate positive sample set in the bad analysis includes: Determining the importance of the candidate positive sample in the bad analysis according to the correlation between the candidate positive sample and one or more target negative samples, the greater the correlation, the higher the importance; The method for determining the degree of association includes at least one of the following: Determining the relevance of the candidate positive sample according to the number of production units to be checked that the candidate positive sample has passed through: the greater the number of production units to be checked that the candidate positive sample has passed through, the greater the relevance of the candidate positive sample; Determine the relevance of the candidate positive sample according to the time difference between the candidate positive sample and the target negative sample passing through the same production unit to be checked: the shorter the time difference between the candidate positive sample and the target negative sample passing through the same production unit to be checked, the higher the relevance of the candidate positive sample; The relevance of the candidate positive sample is determined according to the number of target negative samples associated with the candidate positive sample: when a candidate positive sample is associated with N target negative samples, the larger the value of N is, the higher the relevance of the candidate positive sample is; N is an integer greater than 1.

8. The sample sampling method according to any one of claims 1 to 7, characterized in that: According to the extraction weights of the candidate positive samples, the candidate positive sample set is randomly sampled by importance, and the sampled candidate positive samples are used as target positive samples, including performing the following steps a to e: Step a, randomly arrange all candidate positive samples in the candidate positive sample set, and form a first sample sequence A[i] by the extraction weight values ​​of N candidate positive samples; wherein the value a of the i-th element of the first sample sequence is i is the extraction weight value of the i-th candidate positive sample, 1≤i≤N, N is the total number of candidate positive samples in the candidate positive sample set; Step b, accumulating and processing the elements in the first sample sequence A[i] to generate a second sample sequence B[i]; wherein the value of the i-th element in the second sample sequence is b i , Step c, randomly select a real number m in the interval (0,1) and compare the real number m with the maximum value b max The product of p ref As a reference threshold, p ref =m*b max ; Among them, the maximum value b max is the maximum value of all elements in the second sample sequence B[i]; Step d, search for the target element b in the second sample sequence B[i] n , the value of the target element satisfies the following relationship: b n-1 <p ref ≤ b n ; 2≤n≤N; the target element b n The candidate positive samples represented are used as target positive samples; Step e: Repeat steps c to d until the number of target positive samples obtained by sampling reaches a set value.

9. A sample sampling device for production failure analysis, comprising: An original sample space construction module, configured to construct an original sample space of a target production object according to production inspection data of the target production object for production defect analysis; A negative sample screening module is configured to traverse the original sample space according to the defect type for production defect analysis, and screen samples belonging to the defect type as target negative samples; A positive sample set establishment module is configured to take samples in the original sample space that do not belong to the bad type as original positive samples, and generate an original positive sample set from multiple original positive samples; For any target negative sample, a positive sample associated with the target negative sample is selected from the original positive sample set as a candidate positive sample, and a candidate positive sample set is generated from the candidate positive samples of all target negative samples; A positive sample sampling module is configured to determine the importance of each candidate positive sample in the candidate positive sample set during the bad analysis, and determine the extraction weight of the candidate positive sample according to the importance: the higher the importance of the candidate positive sample, the greater the extraction weight of the candidate positive sample; According to the extraction weights of the candidate positive samples, the candidate positive sample set is randomly sampled according to importance, so that the candidate positive samples with higher extraction weights have a greater probability of being extracted, and the sampled candidate positive samples are used as target positive samples; The target sample set establishment module is configured to generate a new sample set from a plurality of target positive samples and a plurality of target negative samples as a target sample set for production defect analysis.

10. A sample sampling device for production failure analysis, comprising: A memory and a processor, wherein the memory stores a computer program, and when the computer program is executed by the processor, the steps of the sample sampling method for production defect analysis described in any one of claims 1 to 8 are implemented.

11. A computer-readable storage medium storing a computer program, wherein when the computer program is executed by a processor, the computer program implements the steps of the sample sampling method for production defect analysis according to any one of claims 1 to 8.

Images