Sample sampling method, device and storage medium for production process bad analysis

Abstract

The application discloses a sample sampling method, device and storage medium for production process defect analysis. The sample sampling method comprises the following steps: determining a feature vector of a sample according to a defect type of an analysis item of production process defect analysis, filtering production detection data of a production object to obtain a sample set of the production object on the analysis item according to the feature vector; determining a positive-negative sample ratio threshold value according to a total number of samples in the sample set and the analysis item; wherein a negative sample is a sample belonging to a defective product on the analysis item, and a positive sample is a sample not belonging to a defective product on the analysis item; and performing clustering processing on all samples in the sample set according to the feature vector to obtain a positive sample set and a negative sample set satisfying a positive-negative sample ratio requirement. The application can improve the accuracy and efficiency of defect analysis.

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, apparatus and storage medium for defect analysis in production processes. Background Technology

[0002] In recent years, with the rapid development of intelligent learning and computer hardware and software, computer intelligent learning technology has been widely used in various fields of manufacturing, such as equipment maintenance, intelligent monitoring, and defect detection.

[0003] The screen production process is an assembly line operation, and a malfunction in any piece of equipment at any production station can lead to glass defects. Because there are many pieces of equipment, it is impossible to immediately identify which machine caused the defect during the screen production process. The specific cause of a defect can only be located through subsequent defect inspection and defect analysis.

[0004] Because of the massive amount of data generated during the production process, negative samples (bad samples) may be overwhelmed by the big data, thus affecting the efficiency and accuracy of bad sample localization. Summary of the Invention

[0005] In a first aspect, embodiments of this disclosure provide a sample sampling method for defect analysis in a production process, comprising:

[0006] Based on the defect type of the analysis item in the production process defect analysis, the feature vector of the sample is determined, and the production inspection data of the production object is filtered according to the feature vector to obtain the sample set of the production object on the analysis item.

[0007] The positive-to-negative sample ratio threshold is determined based on the total number of samples in the sample set and the analysis item; wherein, negative samples are samples that belong to defective products in the analysis item, and positive samples are samples that do not belong to defective products in the analysis item.

[0008] Clustering is performed on all samples in the sample set based on the feature vector to obtain a positive sample set and a negative sample set that meet the positive-to-negative sample ratio requirement.

[0009] Secondly, embodiments of this disclosure provide a sample sampling device for defect analysis in a production process, comprising: a memory and a processor, wherein the memory stores a computer program, and the computer program, when executed by the processor, implements the steps of the sample sampling method for defect analysis in a production process described above.

[0010] Thirdly, embodiments of this disclosure provide a computer-readable storage medium storing a computer program that, when executed by a processor, implements the steps of the sample sampling method for defect analysis in the production process described above.

[0011] The sampling method, apparatus, and storage medium for defect analysis in production processes provided in this disclosure determine feature vectors of samples based on the defect types of the analysis items in the production process defect analysis. Production inspection data of the production object is filtered based on the feature vectors to obtain a sample set of the production object for the analysis item. A positive-to-negative sample ratio threshold is determined based on the total number of samples in the sample set and the analysis item. All samples in the sample set are clustered based on the feature vectors to obtain a positive sample set and a negative sample set that meet the positive-to-negative sample ratio requirement. The above sampling method discovers the correlation between samples through clustering, thereby uncovering potential negative and / or positive samples, making the division of positive and negative samples more closely reflect the actual production process, thus improving the accuracy and efficiency of defect analysis. Attached Figure Description

[0012] The accompanying drawings are used to provide an understanding of the technical solutions of this disclosure and form part of the specification. They are used together with the embodiments of this disclosure to explain the technical solutions of this disclosure and do not constitute a limitation on the technical solutions of this disclosure.

[0013] Figure 1 A flowchart illustrating a sample sampling method for defect analysis in a production process, provided as an embodiment of this disclosure;

[0014] Figure 2 A flowchart illustrating a method for separating positive and negative samples through multi-round clustering, provided in this embodiment of the disclosure;

[0015] Figure 3 A flowchart illustrating a method for clustering based on centroids, provided in this embodiment of the disclosure;

[0016] Figure 4 A schematic diagram of a sample sampling device for defect analysis in a production process, provided in an embodiment of this disclosure;

[0017] Figure 5 A schematic diagram of another sample sampling device for defect analysis in production process provided in an embodiment of this disclosure. Detailed Implementation

[0018] To make the objectives, technical solutions, and advantages of this disclosure clearer, embodiments of this disclosure will be described in detail below with reference to the accompanying drawings. Note that the implementation methods can be carried out in many different forms. Those skilled in the art will readily understand that the methods and content can be varied in various forms without departing from the spirit and scope of this disclosure. Therefore, this disclosure should not be construed as limited to the content described in the following embodiments. Without conflict, the embodiments and features in the embodiments of this disclosure can be arbitrarily combined with each other.

[0019] The ordinal numbers “first,” “second,” and “third” used in this specification are used to avoid confusion among the constituent elements, not to limit their quantity.

[0020] like Figure 1 As shown in the embodiments of this disclosure, a sample sampling method for defect analysis in a production process is provided, including:

[0021] Step S10: Determine the feature vector of the sample based on the defect type of the analysis item in the production process defect analysis, and filter the production inspection data of the production object based on the feature vector to obtain the sample set of the production object on the analysis item.

[0022] Step S20: Determine the positive-to-negative sample ratio threshold based on the total number of samples in the sample set and the analysis item; wherein, negative samples are samples that belong to defective products in the analysis item, and positive samples are samples that do not belong to defective products in the analysis item.

[0023] Step S30: Cluster all samples in the sample set according to the feature vector to obtain a positive sample set and a negative sample set that meet the positive-to-negative sample ratio requirement.

[0024] The sample sampling method for production process defect analysis provided in the above embodiments determines the feature vector of the sample based on the defect type of the analysis item in the production process defect analysis; filters the production inspection data of the production object based on the feature vector to obtain a sample set of the production object on the analysis item; determines the positive-to-negative sample ratio threshold based on the total number of samples in the sample set and the analysis item; and performs clustering processing on all samples in the sample set based on the feature vector to obtain a positive sample set and a negative sample set that meet the positive-to-negative sample ratio requirement. The above sample sampling method discovers the correlation between samples through clustering processing, thereby uncovering potential negative and / or positive samples, making the division of positive and negative samples closer to the actual production process, thus improving the accuracy and efficiency of defect analysis.

[0025] In some exemplary embodiments, production inspection data includes: production and manufacturing data generated during the production process of the production object, and inspection data generated when it passes through an inspection unit. The production and manufacturing data may include at least one of the following: material information, equipment information, and process information. The production and manufacturing data can be determined according to the actual production scenario, and may include other information besides the above-mentioned information. This document does not limit the types of information in the production and manufacturing data. Different production processes have different complexities, therefore the number of production equipment and inspection stations involved usually varies, resulting in differences in the amount of production inspection data in different process segments.

[0026] In some exemplary embodiments, before obtaining a sample set of the production object on the analysis item by filtering the production detection data of the production object according to the feature vector, the method further includes:

[0027] Data preprocessing is performed on the production testing data of the production objects.

[0028] In some exemplary embodiments, the data preprocessing includes at least one of the following processes: removing empty data, removing duplicate data, removing useless fields from the data, and filtering the data according to preset conditions.

[0029] In some exemplary embodiments, the production object includes: a display panel or a motherboard of a display panel.

[0030] During the production process of a display panel, the motherboard is cut into multiple display panels. Exemplarily, the motherboard of the display panel includes: a substrate, or a substrate and a buffer layer film covering the substrate. The substrate includes a rigid substrate or a flexible substrate.

[0031] In some exemplary embodiments, the defect types include broad categories and subcategories. Broad categories of defect types include, for example, dark line defects, bright line defects, and hot spot defects. Subcategories of dark line defects include, for example, horizontal dark line defects, vertical dark line defects, and diagonal dark line defects.

[0032] In some exemplary implementations, a defective product refers to a product whose quality has been reduced due to defects, and which may require rework, repair, or scrapping.

[0033] In some exemplary implementations, determining a positive-to-negative sample ratio threshold based on the total number of samples in the sample set and the analysis item includes:

[0034] If the total number of samples in the sample set of the first analysis item is less than the total number of samples in the sample set of the second analysis item, then the positive-to-negative sample ratio threshold for the first analysis item is set to be less than or equal to the positive-to-negative sample ratio threshold for the second analysis item; and / or

[0035] If the number of production equipment involved in the first analysis project is greater than the number of production equipment involved in the second analysis project, then the positive-negative sample ratio threshold of the first analysis project is set to be less than or equal to the positive-negative sample ratio threshold of the second analysis project.

[0036] For example, when the total number of samples in the sample set is small, the number of negative samples can be increased by appropriately decreasing the positive-to-negative sample ratio threshold. When the total number of samples in the sample set is large, the number of positive samples can be increased by appropriately increasing the positive-to-negative sample ratio threshold. In other implementations, the positive-to-negative sample ratio threshold can also be increased or decreased according to other requirements of the defect analysis.

[0037] In some exemplary embodiments, the feature vector of the sample includes: defect detection features and production features related to the generation of the defect. By reasonably setting the types and dimensions of the sample feature vectors, clustering can be used to discover the correlation between samples, thereby uncovering potential negative and / or positive samples, making the positive and negative samples more closely resemble the actual production process, and thus improving the accuracy of defect analysis.

[0038] In some exemplary embodiments, the detection features of defects include at least one of the following: defect rate and defect location.

[0039] In some exemplary embodiments, the production characteristics associated with the occurrence of the defect include at least one of the following: production time and production equipment.

[0040] In some exemplary implementations, the positive-to-negative sample ratio requirement refers to the ratio of the number of samples in the positive sample set to the number of samples in the negative sample set being less than or equal to the positive-to-negative sample ratio threshold. During production, negative samples are relatively sparse, while the number of positive samples is far greater than the number of negative samples. In defect analysis, this imbalance between positive and negative samples may affect the accuracy of data analysis.

[0041] In some exemplary implementations, such as Figure 2 As shown, clustering is performed on all samples in the sample set based on the feature vector to obtain a positive sample set and a negative sample set that meet the positive-to-negative sample ratio requirement. This includes performing the following steps a to g:

[0042] Step a: Randomly select a first sample and a second sample from the sample set; wherein the defect rate of the first sample is lower than the defect rate of the second sample;

[0043] Step b: Use the first sample as the centroid of the positive sample group and the second sample as the centroid of the negative sample group;

[0044] Step c: Cluster all samples in the sample set using the centroids of the positive and negative sample groups, dividing the sample set into positive sample groups and negative sample groups;

[0045] Step d: Reorganize the negative samples obtained after each round of clustering to obtain a new negative sample group, and take the positive sample group obtained after the latest round of clustering as the new positive sample group;

[0046] Step e: Determine whether the ratio a / b of the number of samples a in the new positive sample group to the number of samples b in the new negative sample group is less than or equal to the positive-to-negative sample ratio threshold. If yes, proceed to step g; otherwise, proceed to step f.

[0047] Step f: Use the new positive sample group as the sample set for the next round of clustering, and return to step a;

[0048] Step g: Use the new negative sample group as the final negative sample group and the new positive sample group as the final positive sample group to end the clustering.

[0049] In some exemplary implementations, such as Figure 3 As shown, the step of clustering all samples in the sample set using the centroids of the positive and negative sample groups to divide the sample set into positive and negative sample groups includes performing the following steps c1 to c6:

[0050] Step c1: For each sample to be classified in the sample set excluding the two centroids, perform the following processing: calculate the Euclidean distance between the sample to be classified and the positive sample group based on the value of the feature vector of the centroid of the positive sample group to obtain a first distance; calculate the Euclidean distance between the sample to be classified and the negative sample group based on the value of the feature vector of the centroid of the negative sample group to obtain a second distance; if the first distance is less than the second distance, the sample to be classified belongs to the positive sample group; if the second distance is less than the first distance, the sample to be classified belongs to the negative sample group.

[0051] Suppose that the feature vector of each sample in the sample set consists of m features (m-dimensional features), and the m feature values ​​of the centroid A are: x1, x2, ..., xm. m The m feature values ​​of the sample B to be classified are: y1, y2, ..., y m The Euclidean distance d between the centroid A and the sample B to be classified can be expressed as:

[0052]

[0053] Step c2: Construct a new positive sample as the new centroid of the positive sample group, and construct a new negative sample as the new centroid of the negative sample group; the feature vector of the new positive sample includes multiple features, and the value of each feature is the average value of the feature values ​​of all positive samples in the positive sample group in that dimension; the feature vector of the new negative sample includes multiple features, and the value of each feature is the average value of the feature values ​​of all negative samples in the negative sample group in that dimension.

[0054] The following example, using the new centroid of the positive sample group, illustrates the method for determining the feature vector values. Assume the total number of positive samples in the positive sample group is n, and the feature vector of each positive sample includes m features. The value of the feature vector of the i-th positive sample is: x i1 ,x i2 ,…,x im , 1≤i≤n. The feature vector of the new centroid of the positive sample group includes m features: y1, y2, ..., y m Then the value y of the j-th feature of the new centroid of the positive sample group. j This can be expressed as:

[0055]

[0056] 1≤j≤m.

[0057] The method for determining the eigenvector of the new centroid of the negative sample group is similar to that for the new centroid of the positive sample group, and will not be repeated here.

[0058] Step c3: Calculate the Euclidean distance between the new centroid and the original centroid of the positive sample group to obtain the third distance, and calculate the Euclidean distance between the new centroid and the original centroid of the negative sample group to obtain the fourth distance;

[0059] Step c4: Determine whether the third distance and the fourth distance are both less than the first distance threshold. If yes, proceed to step c6; otherwise, proceed to step c5.

[0060] Step c5: Replace the original centroid with the new centroid of the positive sample group, and replace the original centroid with the new centroid of the negative sample group. Return to step c2.

[0061] Step c6: End this round of clustering processing.

[0062] The above-described centroid-based clustering is an improved K-Means clustering process, where K=2. Through multiple rounds of iterative clustering, the number of positive and negative samples can be adjusted to meet the required ratio, thus providing sufficient negative and positive samples for defect localization in defect analysis.

[0063] like Figure 4 As shown, this disclosure provides a sample sampling device for defect analysis in a production process, comprising:

[0064] The sample set establishment module 100 is configured to determine the feature vector of the sample based on the defect type of the analysis item of the production process defect analysis, and filter the production inspection data of the production object according to the feature vector to obtain the sample set of the production object on the analysis item.

[0065] The setting module 200 is configured to determine a positive-to-negative sample ratio threshold based on the total number of samples in the sample set and the analysis item; wherein, negative samples are samples that belong to defective products in the analysis item, and positive samples are samples that do not belong to defective products in the analysis item.

[0066] The positive and negative sample partitioning module 300 is configured to perform clustering processing on all samples in the sample set according to the feature vector to obtain a positive sample set and a negative sample set that meet the positive and negative sample ratio requirements.

[0067] The sample sampling device for production process defect analysis provided in the above embodiments includes a sample set establishment module that determines the feature vector of the sample based on the defect type of the analysis item in the production process defect analysis, and filters the production inspection data of the production object according to the feature vector to obtain a sample set of the production object on the analysis item; a setting module that determines the positive-to-negative sample ratio threshold based on the total number of samples in the sample set and the analysis item; and a positive-to-negative sample division module that performs clustering processing on all samples in the sample set according to the feature vector to obtain a positive sample set and a negative sample set that meet the positive-to-negative sample ratio requirement. The above sample sampling device discovers the correlation between samples through clustering processing, thereby uncovering potential negative and / or positive samples, making the division of positive and negative samples closer to the actual production process, thus improving the accuracy and efficiency of defect analysis.

[0068] In some exemplary embodiments, production inspection data includes: production and manufacturing data related to the production process generated by the production object, and inspection data generated when it passes through the inspection unit.

[0069] In some exemplary embodiments, the apparatus further includes a preprocessing module configured to preprocess production inspection data of the production object.

[0070] In some exemplary embodiments, the data preprocessing includes at least one of the following processes: removing empty data, removing duplicate data, removing useless fields from the data, and filtering the data according to preset conditions.

[0071] In some exemplary embodiments, the production object includes: a display panel or a motherboard of a display panel.

[0072] In some exemplary embodiments, the setting module is configured to determine a positive-to-negative sample ratio threshold based on the total number of samples in the sample set and the analysis item in the following manner: if the total number of samples in the sample set of the first analysis item is less than the total number of samples in the sample set of the second analysis item, then the positive-to-negative sample ratio threshold of the first analysis item is set to be less than or equal to the positive-to-negative sample ratio threshold of the second analysis item; and / or if the number of production equipment involved in the first analysis item is greater than the number of production equipment involved in the second analysis item, then the positive-to-negative sample ratio threshold of the first analysis item is set to be less than or equal to the positive-to-negative sample ratio threshold of the second analysis item.

[0073] In some exemplary embodiments, the feature vector of the sample includes: detection features of the defect and production features related to the generation of the defect.

[0074] In some exemplary embodiments, the detection features of defects include at least one of the following: defect rate and defect location.

[0075] In some exemplary embodiments, the production characteristics associated with the occurrence of the defect include at least one of the following: production time and production equipment.

[0076] In some exemplary implementations, the clustering process includes K-Means clustering.

[0077] In some exemplary implementations, the positive-to-negative sample ratio requirement refers to the ratio of the number of samples in the positive sample set to the number of samples in the negative sample set being less than or equal to the positive-to-negative sample ratio threshold. During production, negative samples are relatively sparse, while the number of positive samples is far greater than the number of negative samples. In defect analysis, this imbalance between positive and negative samples may affect the accuracy of data analysis.

[0078] In some exemplary embodiments, the positive and negative sample partitioning module is configured to perform clustering processing on all samples in the sample set based on the feature vector by executing steps a to g to obtain a positive sample set and a negative sample set that satisfy the positive-to-negative sample ratio requirement:

[0079] Step a: Randomly select a first sample and a second sample from the sample set; wherein the defect rate of the first sample is lower than the defect rate of the second sample;

[0080] Step b: Use the first sample as the centroid of the positive sample group and the second sample as the centroid of the negative sample group;

[0081] Step c: Cluster all samples in the sample set using the centroids of the positive and negative sample groups, dividing the sample set into positive sample groups and negative sample groups;

[0082] Step d: Reorganize the negative samples obtained after each round of clustering to obtain a new negative sample group, and take the positive sample group obtained after the latest round of clustering as the new positive sample group;

[0083] Step e: Determine whether the ratio a / b of the number of samples a in the new positive sample group to the number of samples b in the new negative sample group is less than or equal to the positive-to-negative sample ratio threshold. If yes, proceed to step g; otherwise, proceed to step f.

[0084] In each clustering process, the sum of the number of samples in the new positive sample group (a) and the number of samples in the new negative sample group (b) remains unchanged and is always equal to the number of samples in the original sample set (before the first round of clustering).

[0085] Step f: Use the new positive sample group as the sample set for the next round of clustering, and return to step a;

[0086] Step g: Use the new negative sample group as the final negative sample group and the new positive sample group as the final positive sample group to end the clustering.

[0087] In some exemplary embodiments, the positive and negative sample partitioning module is configured to cluster all samples in the sample set using the centroids of the positive and negative sample groups by performing the following steps c1 to c6, thereby dividing the sample set into a positive sample group and a negative sample group:

[0088] Step c1: For each sample to be classified in the sample set excluding the two centroids, perform the following processing: calculate the Euclidean distance between the sample to be classified and the positive sample group based on the value of the feature vector of the centroid of the positive sample group to obtain a first distance; calculate the Euclidean distance between the sample to be classified and the negative sample group based on the value of the feature vector of the centroid of the negative sample group to obtain a second distance; if the first distance is less than the second distance, the sample to be classified belongs to the positive sample group; if the second distance is less than the first distance, the sample to be classified belongs to the negative sample group.

[0089] Step c2: Construct a new positive sample as the new centroid of the positive sample group, and construct a new negative sample as the new centroid of the negative sample group; the feature vector of the new positive sample includes multiple features, and the value of each feature is the average value of the feature values ​​of all positive samples in the positive sample group in that dimension; the feature vector of the new negative sample includes multiple features, and the value of each feature is the average value of the feature values ​​of all negative samples in the negative sample group in that dimension.

[0090] Step c3: Calculate the Euclidean distance between the new centroid and the original centroid of the positive sample group to obtain the third distance, and calculate the Euclidean distance between the new centroid and the original centroid of the negative sample group to obtain the fourth distance;

[0091] Step c4: Determine whether the third distance and the fourth distance are both less than the first distance threshold. If yes, proceed to step c6; otherwise, proceed to step c5.

[0092] Step c5: Replace the original centroid with the new centroid of the positive sample group, and replace the original centroid with the new centroid of the negative sample group. Return to step c2.

[0093] Step c6: End this round of clustering processing.

[0094] The above-described centroid-based clustering is an improved K-Means clustering process, where K=2. Through multiple rounds of iterative clustering, the number of positive and negative samples can be adjusted to meet the required ratio, thus providing sufficient negative and positive samples for defect localization in defect analysis.

[0095] like Figure 5 As shown, this disclosure provides a sample sampling device for defect analysis in a production process, including: a memory and a processor. The memory stores a computer program, and when the computer program is executed by the processor, it implements the steps of the sample sampling method for defect analysis in a production process described above.

[0096] This disclosure provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the steps of the sample sampling method described above for defect analysis in a production process.

[0097] It will be understood by those skilled in the art that all or some of the steps, systems, or apparatuses disclosed above, and their functional modules / units, can be implemented as software, firmware, hardware, or suitable combinations thereof. In hardware implementations, the division between functional modules / units mentioned above 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 collaboratively by several physical components. Some or all components may be implemented as software executed by a processor, such as a digital signal processor or microprocessor, or as hardware, or as an integrated circuit, such as an application-specific integrated circuit (ASIC). Such software may be distributed on a computer-readable medium, which may include computer storage media (or non-transitory media) and communication media (or transient media). As is known to those skilled in the art, the term computer storage media 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 technologies, CD-ROM, digital versatile disc (DVD) or other optical disc storage, magnetic cartridges, magnetic tape, disk storage or other magnetic storage devices, or any other medium that can be used to store desired information and can be accessed by a computer. Furthermore, it is well known to those skilled in the art that communication media typically contain computer-readable instructions, data structures, program modules, or other data in modulated data signals such as carrier waves or other transmission mechanisms, and may include any information delivery medium.

[0098] While the embodiments disclosed herein are as described above, the content is merely for the purpose of facilitating understanding of this disclosure and is not intended to limit this disclosure. Any person skilled in the art to which this disclosure pertains may make any modifications and changes in the form and details of the implementation without departing from the spirit and scope disclosed herein; however, the scope of patent protection of this disclosure shall still be determined by the scope defined in the appended claims.

Claims

1. A sample sampling method for defect analysis in production processes, comprising: Based on the defect type of the analysis item in the production process defect analysis, the feature vector of the sample is determined, and the production inspection data of the production object is filtered according to the feature vector to obtain the sample set of the production object on the analysis item. The positive-to-negative sample ratio threshold is determined based on the total number of samples in the sample set and the analysis item; wherein, negative samples are samples that belong to defective products in the analysis item, and positive samples are samples that do not belong to defective products in the analysis item. Clustering is performed on all samples in the sample set based on the feature vector to obtain a positive sample set and a negative sample set that meet the positive-to-negative sample ratio requirement. The positive-to-negative sample ratio requirement means that the ratio of the number of samples in the positive sample set to the number of samples in the negative sample set is less than or equal to the positive-to-negative sample ratio threshold, including performing the following steps a to g: Step a: Randomly select a first sample and a second sample from the sample set; wherein the defect rate of the first sample is lower than the defect rate of the second sample; Step b: Use the first sample as the centroid of the positive sample group and the second sample as the centroid of the negative sample group; Step c: Cluster all samples in the sample set using the centroids of the positive and negative sample groups, dividing the sample set into positive sample groups and negative sample groups; Step d: Reorganize the negative samples obtained after each round of clustering to obtain a new negative sample group, and take the positive sample group obtained after the latest round of clustering as the new positive sample group. Step e: Determine whether the ratio a / b of the number of samples a in the new positive sample group to the number of samples b in the new negative sample group is less than or equal to the positive-to-negative sample ratio threshold. If yes, proceed to step g; otherwise, proceed to step f. Step f: Use the new positive sample group as the sample set for the next round of clustering, and return to step a; Step g: Use the new negative sample group as the final negative sample group and the new positive sample group as the final positive sample group to end the clustering.

2. The sample sampling method as described in claim 1, characterized in that: The production inspection data includes: production and manufacturing data related to the production process generated by the production object, and inspection data generated when it passes through the inspection unit.

3. The sample sampling method as described in claim 1, characterized in that: Determining the positive-to-negative sample ratio threshold based on the total number of samples in the sample set and the analysis item includes: If the total number of samples in the sample set of the first analysis item is less than the total number of samples in the sample set of the second analysis item, then the positive-to-negative sample ratio threshold for the first analysis item is set to be less than or equal to the positive-to-negative sample ratio threshold for the second analysis item; and / or If the number of production equipment involved in the first analysis project is greater than the number of production equipment involved in the second analysis project, then the positive-negative sample ratio threshold of the first analysis project is set to be less than or equal to the positive-negative sample ratio threshold of the second analysis project.

4. The sample sampling method as described in claim 1, characterized in that: The feature vector of the sample includes: the detection features of the defect and the production features related to the generation of the defect.

5. The sample sampling method as described in claim 4, characterized in that: The detection characteristics of defects include at least one of the following: defect rate, defect location; The production characteristics associated with the occurrence of the defect include at least one of the following: production time and production equipment.

6. The sample sampling method as described in claim 1, characterized in that: The step of clustering all samples in the sample set using the centroids of the positive and negative sample groups to divide the sample set into positive and negative sample groups includes performing the following steps c1 to c6: Step c1: For each sample to be classified in the sample set excluding the two centroids, perform the following processing: calculate the Euclidean distance between the sample to be classified and the positive sample group based on the value of the feature vector of the centroid of the positive sample group to obtain a first distance; calculate the Euclidean distance between the sample to be classified and the negative sample group based on the value of the feature vector of the centroid of the negative sample group to obtain a second distance; if the first distance is less than the second distance, the sample to be classified belongs to the positive sample group; if the second distance is less than the first distance, the sample to be classified belongs to the negative sample group. Step c2: Construct a new positive sample as the new centroid of the positive sample group, and construct a new negative sample as the new centroid of the negative sample group; the feature vector of the new positive sample includes multiple features, and the value of each feature is the average value of the corresponding feature values ​​of all positive samples in the positive sample group; the feature vector of the new negative sample includes multiple features, and the value of each feature is the average value of the corresponding feature values ​​of all negative samples in the negative sample group. Step c3: Calculate the Euclidean distance between the new centroid and the original centroid of the positive sample group to obtain the third distance, and calculate the Euclidean distance between the new centroid and the original centroid of the negative sample group to obtain the fourth distance; Step c4: Determine whether the third distance and the fourth distance are both less than the first distance threshold. If yes, proceed to step c6; otherwise, proceed to step c5. Step c5: Replace the original centroid with the new centroid of the positive sample group, and replace the original centroid with the new centroid of the negative sample group. Return to step c2. Step c6: End this round of clustering.

7. A sample sampling device for defect analysis in a production process, comprising: A memory and a processor, wherein the memory stores a computer program, which, when executed by the processor, implements the steps of the sample sampling method for defect analysis in a production process as described in any one of claims 1-6.

8. A computer-readable storage medium storing a computer program that, when executed by a processor, implements the steps of the sample sampling method for defect analysis in a production process according to any one of claims 1-6.

Images