Data processing methods, systems, and computer-readable storage media

Abstract

A data processing method, system, and computer-readable storage medium are disclosed. The method includes: acquiring target data; the target data includes data from multiple analysis units (21); acquiring detection data (22) of multiple substrates to be tested containing defects in each analysis unit; superimposing the defects in the multiple substrates to be tested to obtain a defect dataset (23); performing clustering calculation on the defect dataset to obtain a first defect cluster (24); and generating alarm information (25) for the location of the equipment contact point corresponding to the first defect cluster based on the first defect cluster. This method analyzes the substrates to be tested in each analysis unit, shortening the time interval between the production process and the analysis of recurring defects, which is beneficial to improving data processing efficiency; furthermore, this method can discover defects that repeatedly occur at the same location, i.e., obtain the first defect cluster, achieving the effect of timely detection of production equipment causing recurring defects, which is beneficial to improving the yield of the substrates to be tested.

Description

Technical Field

[0001] This disclosure relates to the field of data processing technology, and in particular to a data processing method, system and computer-readable storage medium. Background Technology

[0002] Currently, in the production of display panels, Automated Optical Inspection (AOI) is used to inspect the substrate (GLASS) to detect process defects. For example, improper equipment contact points or particles on the mask can cause defects to repeatedly occur at the same coordinate positions on substrates passing through the same process station, reducing yield. Summary of the Invention

[0003] This disclosure provides a data processing method, system, and computer-readable storage medium to address the shortcomings of related technologies.

[0004] According to a first aspect of the present disclosure, a data processing method is provided, comprising:

[0005] Acquire target data; the target data includes data from multiple analysis units;

[0006] Acquire detection data from multiple substrates to be tested that contain defects in each of the analysis units;

[0007] A defect dataset is obtained by superimposing the defective points from the multiple substrates to be inspected.

[0008] The first cluster set of defective points is obtained by performing clustering calculations on the defective point dataset;

[0009] Alarm information for the location of the device contact point corresponding to the first defect cluster is generated based on the first defect cluster.

[0010] Optionally, the target data is acquired, including:

[0011] Based on the production data in the source data, the substrates to be tested are sorted according to the production time of each production equipment to obtain the sorting result of the substrates to be tested.

[0012] Based on the sorting results, a preset number of substrates to be tested are sequentially divided into the same unit to obtain multiple analysis units;

[0013] Based on the detection data in the source data, the number of substrates to be tested that have passed through the detection station in each analysis unit is obtained;

[0014] The arrival rate of the substrate to be detected in each analysis unit is obtained, and the data of the analysis units whose arrival rate exceeds the preset arrival rate threshold is used as the converted target data.

[0015] Optionally, clustering is performed on the defective point dataset to obtain a first defective point cluster set, including:

[0016] Using any defective point Pi in the defective point dataset as a reference point, obtain the number m of defective points within the region formed by r as a distance threshold, and obtain the initial defective point cluster;

[0017] Traverse the defective points in the defective point dataset to obtain multiple initial defective point clusters;

[0018] Obtain multiple initial defective point clusters that satisfy the condition that the number of defective points m in the initial defective point cluster is greater than or equal to the first threshold, and obtain the first defective point cluster set.

[0019] Optionally, alarm information for the location of the device contact point corresponding to the first defect cluster is generated based on the first defect cluster, including:

[0020] Multiple first defect point clusters are filtered out, and first defect point clusters with defect points having the same location information are filtered out to obtain second defect point clusters.

[0021] Alarm information for the location of the device contact point corresponding to the second defect cluster is generated based on the second defect cluster.

[0022] Optionally, multiple first defect point clusters are filtered to remove first defect point clusters with defect points having the same location information, including:

[0023] Obtain the number of defective points in each cluster within the first defective point cluster set;

[0024] The point clusters within the first defective point cluster set are sorted according to the stated quantity;

[0025] The cluster with the most defective points in the first defective point cluster is selected as the candidate point cluster.

[0026] Obtain the intersection-union ratio (CURNR) of any point cluster within the first defective point cluster set and the candidate point cluster, and remove point clusters from the first defective point cluster set whose CURNR exceeds a preset CURNR threshold.

[0027] The candidate point cluster is moved from the first defective point cluster set into a preset second defective point cluster set;

[0028] If the first defective point cluster is a non-empty set, continue to execute the step of obtaining the point cluster with the most defective points in the first defective point cluster as a candidate point cluster; otherwise, determine that the second defective point cluster is obtained.

[0029] Optionally, the method further includes:

[0030] The corresponding production equipment is controlled according to the location of the contact point of the equipment, so that the production equipment performs a protective action.

[0031] Optionally, the location of the device contact point includes at least one of the following: coordinates of the defective point, identification code of the substrate to be tested, and identification code of the production equipment.

[0032] According to a second aspect of the present disclosure, a data processing method is provided, comprising:

[0033] In response to the operation of the subscription monitoring function, a preset value of at least one subscription parameter related to the detection of recurring defects is collected;

[0034] The target data is obtained according to the preset value, and the target data is processed to obtain the first defect cluster set corresponding to each analysis unit;

[0035] Alarm information for the location of the device contact point corresponding to the first defect cluster is generated based on the first defect cluster.

[0036] According to a third aspect of the present disclosure, a data processing system is provided, including: a distributed storage device, a data conversion device, and a data processing device;

[0037] The distributed storage device is configured to store source data.

[0038] The data conversion device includes one or more first processors configured to convert the source data into target data; the target data includes data from multiple analysis units.

[0039] The data processing device includes one or more second processors, which are configured to acquire detection data of multiple substrates to be tested containing defects in each analysis unit, and superimpose the defects in the multiple substrates to be tested to obtain a defect dataset, perform clustering calculation on the defect dataset to obtain a first defect cluster, and generate alarm information for the device contact point location corresponding to the first defect cluster based on the first defect cluster.

[0040] Optionally, the source data includes at least one of metadata, production data, and testing data; the metadata represents the association between process stations and testing stations, the production data represents the production history data of each substrate to be tested, and the testing data is the data generated when the testing station tests the substrate to be tested.

[0041] Optionally, the one or more second processors are further configured to perform:

[0042] Based on the production data in the source data, the substrates to be tested are sorted according to the production time of each production equipment to obtain the sorting result of the substrates to be tested.

[0043] Based on the sorting results, a preset number of substrates to be tested are sequentially divided into the same unit to obtain multiple analysis units;

[0044] Based on the detection data in the source data, the number of substrates to be tested that have passed through the detection station in each analysis unit is obtained;

[0045] The arrival rate of the substrate to be detected in each analysis unit is obtained, and the data of the analysis units whose arrival rate exceeds the preset arrival rate threshold is used as the converted target data.

[0046] Optionally, the one or more second processors are further configured to perform clustering calculations on the defective point dataset to obtain a first defective point cluster set, including:

[0047] Using any defective point Pi in the defective point dataset as a reference point, the number m of defective points within the region formed by the distance threshold r is obtained, thus obtaining the initial defective point cluster;

[0048] Traverse the defective points in the defective point dataset to obtain multiple initial defective point clusters;

[0049] Obtain multiple initial defective point clusters that satisfy the condition that the number of defective points m in the initial defective point cluster is greater than or equal to the first threshold, and obtain the first defective point cluster set.

[0050] Optionally, the one or more second processors are further configured to execute alarm information for generating device contact point locations corresponding to the first defect cluster based on the first defect cluster, including:

[0051] Multiple first defect point clusters are filtered out, and first defect point clusters with defect points having the same location information are filtered out to obtain second defect point clusters.

[0052] Alarm information for the location of the device contact point corresponding to the second defect cluster is generated based on the second defect cluster.

[0053] Optionally, the one or more second processors are further configured to perform filtering on a plurality of first defect clusters, filtering first defect clusters with defective points having the same location information, including:

[0054] Obtain the number of defective points in each cluster within the first defective point cluster set;

[0055] The point clusters within the first defective point cluster set are sorted according to the stated quantity;

[0056] The cluster with the most defective points in the first defective point cluster is selected as the candidate point cluster.

[0057] Obtain the intersection-union ratio (CURNR) of any point cluster within the first defective point cluster set and the candidate point cluster, and remove point clusters from the first defective point cluster set whose CURNR exceeds a preset CURNR threshold.

[0058] The candidate point cluster is moved from the first defective point cluster set into a preset second defective point cluster set;

[0059] If the first defective point cluster is a non-empty set, continue to execute the step of obtaining the point cluster with the most defective points in the first defective point cluster as a candidate point cluster; otherwise, determine that the second defective point cluster is obtained.

[0060] Optionally, the one or more second processors are further configured to perform:

[0061] The corresponding production equipment is controlled according to the location of the contact point of the equipment, so that the production equipment performs a protective action.

[0062] Optionally, the location of the device contact point includes at least one of the following: coordinates of the defective point, identification code of the substrate to be tested, and identification code of the production equipment.

[0063] Optionally, it also includes a display device; the display device includes at least one display screen; the at least one display screen is configured to display the alarm information.

[0064] According to a fifth aspect of the present disclosure, a computer-readable storage medium is provided that, when an executable computer program in the storage medium is executed by a processor, enables the implementation of the above-described method.

[0065] The technical solutions provided by the embodiments of this disclosure may include the following beneficial effects:

[0066] As can be seen from the above embodiments, in this embodiment of the disclosure, target data can be acquired; detection data of multiple substrates to be tested containing defects in each analysis unit can be acquired; then, the defects in the multiple substrates to be tested are superimposed to obtain a defect dataset; subsequently, clustering calculation is performed on the defect dataset to obtain a first defect cluster; finally, alarm information of the device contact point location corresponding to the first defect cluster is generated based on the first defect cluster. In this way, by dividing the substrates to be tested produced by the same production equipment into analysis units, and then analyzing the substrates to be tested in each analysis unit, the time interval between the production process and the analysis of repeated defects is shortened, which is beneficial to improving data processing efficiency; and, by superimposing the defects of multiple substrates to be tested in this disclosure, defects that repeatedly occur in the same position can be found to obtain the first defect cluster, which achieves the effect of timely detection of repeated defects caused by the production equipment, which is beneficial to improving the yield of the substrates to be tested.

[0067] It should be understood that the above general description and the following detailed description are exemplary and explanatory only, and are not intended to limit this disclosure. Attached Figure Description

[0068] The accompanying drawings, which are incorporated in and form a part of this specification, illustrate embodiments consistent with this disclosure and, together with the description, serve to explain the principles of this disclosure.

[0069] Figure 1 This is a block diagram illustrating a data processing system according to an exemplary embodiment.

[0070] Figure 2 This is a flowchart illustrating a data processing method according to an exemplary embodiment.

[0071] Figure 3 This is a schematic diagram illustrating the relationship between process stations and testing stations according to an exemplary embodiment.

[0072] Figure 4 This is a flowchart illustrating the acquisition of target data according to an exemplary embodiment.

[0073] Figure 5 This is a flowchart illustrating the acquisition of a first cluster of defects according to an exemplary embodiment.

[0074] Figure 6 This is a flowchart illustrating the generation of alarm information according to an exemplary embodiment.

[0075] Figure 7 This is a flowchart illustrating the filtering of duplicate defective points in a first point cluster set according to an exemplary embodiment.

[0076] Figure 8(a) is a schematic diagram illustrating the effect of a first point cluster set comprising 5 point clusters according to an exemplary embodiment.

[0077] Figure 8(b) is a schematic diagram illustrating the effect of a first point cluster set comprising 5 point clusters according to an exemplary embodiment.

[0078] Figure 9 This is a flowchart illustrating another data processing method according to an exemplary embodiment.

[0079] Figure 10 This is a schematic diagram illustrating the effect of a functional order interface according to an exemplary embodiment.

[0080] Figure 11 This is a schematic diagram illustrating the effect of an alarm interface according to an exemplary embodiment. Detailed Implementation

[0081] Exemplary embodiments will now be described in detail, examples of which are illustrated in the accompanying drawings. When the following description relates to the drawings, unless otherwise indicated, the same numerals in different drawings denote the same or similar elements. The embodiments described below by way of example do not represent all embodiments consistent with this disclosure. Rather, they are merely examples of apparatus consistent with some aspects of this disclosure as detailed in the appended claims.

[0082] Currently, industrial production lines consist of numerous production machines. Any malfunction or abnormal operating parameters of any machine can affect product yield. When defective products are produced, production personnel need to pinpoint the cause. However, the amount of data generated by the production equipment on the production line is substantial. Using traditional relational databases for data storage and analysis presents several problems: slow read / write speeds, high costs, limited storage capacity, and difficulty in expansion. These inherent drawbacks increase the time and hardware costs of database use. Furthermore, the continuous production and constant generation of data further complicate the monitoring and location of defective products and the analysis of their causes, leading to decreased productivity and significant consumption of manpower and materials on the production line. Since multiple factories and numerous production machines are involved, the amount of raw data is enormous. For example, the raw data generated by all production equipment daily may be several hundred gigabytes, and the data generated hourly may be tens of gigabytes.

[0083] This disclosure provides a data processing system, and subsequent embodiments are also referred to as the system of this disclosure, this system, or this system. Figure 1As shown, the data processing system includes a distributed storage device 100, a data conversion device 200, and a data processing device 300. The data conversion device 200 is connected to both the distributed storage device 100 and the data processing device 300. This solution adopts a completely new data processing system workflow, which can effectively reduce database management and usage costs, while improving the efficiency of finding the causes of defective products, thereby improving production efficiency.

[0084] The distributed storage device 100 may include a source database capable of processing raw data generated by production equipment, such as a relational database (e.g., Oracle, MySQL) of systems such as YMS (Yield Management System), FDC (Fault Detection & Classification), and MES (Manufacturing Execution System), to store production data generated by multiple production equipment.

[0085] For example, production data generated by multiple production devices includes production data; this production data includes information about the production devices that multiple samples (such as a substrate to be tested) passed through during the production process and information about the types of defects that occurred. Each sample goes through multiple production devices during the production process, and each production device participates in the production process of only a portion of the multiple samples. For ease of explanation, the sample in this disclosure can be a substrate to be tested (GLASS), and subsequent embodiments can use the substrate to be tested as an example to describe various schemes.

[0086] The distributed storage device 100 stores relatively complete data. The distributed storage device can include multiple hardware storage devices distributed in different physical locations (e.g., in different factories or on different production lines), and they exchange information via wireless transmission (e.g., networks) or wired connections. This results in a distributed data structure, but logically constitutes a database based on big data technology.

[0087] The data in the distributed storage device 100 can be stored using Hive or HBase database formats. For example, using Hive, the raw data is first stored in a data lake; then, it can be preprocessed in Hive according to the application theme and scenario, such as data cleaning and transformation, to obtain data warehouses with different themes (e.g., production history, testing data, equipment data, metadata) and datasets with different scenarios (e.g., equipment analysis, parameter analysis). These datasets can then be connected to display devices, analysis devices, etc., through different API interfaces to achieve data interaction with these devices.

[0088] In one embodiment, a storage and computing scheme for massive structured data is implemented: a big data scheme for a distributed file system (DFS).

[0089] Among them, big data technologies based on Distributed File System (DFS) allow the construction of large clusters using multiple inexpensive hardware devices to process massive amounts of data. For example, Hive is a data warehouse tool based on Hadoop, used for data extraction, transformation, and loading (ETL). Hive defines a simple SQL-like query language and also allows for complex analysis tasks that the default tool cannot perform through custom MapReduce mappers and reducers. Hive does not have a dedicated data storage format or indexes; users can freely organize the tables and process the data in the database. It is evident that the parallel processing of distributed file management can meet the storage and processing requirements of massive amounts of data. Users can process simple data through SQL queries, while complex processing can be achieved using custom functions. Therefore, when analyzing massive amounts of data from a factory, it is necessary to extract the data from the factory database into a distributed file system. This avoids damage to the original data and improves data analysis efficiency.

[0090] In one embodiment, the distributed storage device 100 can be a single memory, multiple memories, or a collective term for multiple storage elements. For example, the memory may include: Random Access Memory (RAM), Double Data Rate Synchronous Dynamic Random Access Memory (DDR SRAM), and may also include non-volatile memory, such as disk storage, flash memory, etc.

[0091] The data transformation device 200 may include one or more processors, hereinafter referred to as the first processor for distinction. The data transformation device 200 can communicate with the distributed storage device 100 to obtain source data. For example, the data transformation device can use ETL data extraction tools (such as Pentaho, Sqoop, Kettle, etc.) to extract source data from a distributed database to transform the source data into target data.

[0092] The data processing device 300 may include one or more processors, hereinafter referred to as a second processor for distinction. The data processing device 300 is used to implement the data processing method as described in any of the following embodiments. For a specific implementation, please refer to a data processing method scheme in the following embodiments, which will not be described here.

[0093] In one embodiment, the data processing system further includes a display device (not shown). This display device is used to display an interface, which may include data such as alarm information from the data processing device 300. In one embodiment, the display device may be a touch screen display, or a product containing a display, such as a television, computer (all-in-one or desktop), tablet computer, mobile phone, electronic display screen, etc.

[0094] Figure 2 This is a flowchart illustrating a data processing method according to an exemplary embodiment, which can be applied to... Figure 1 The data processing system shown. See also Figure 2 A data processing method, comprising steps 21 to 24:

[0095] In step 21, target data is acquired; the target data includes data from multiple analysis units.

[0096] In this embodiment, the data conversion device 200 in this system can communicate with the distributed storage device 100 to obtain the cutting-edge data stored in the distributed storage device. This source data may include at least one of metadata, production data, and testing data; the source data collected can be selected according to the specific scenario, and is not limited here. Metadata represents the association between process stations and testing stations, with the effect as follows: Figure 3 As shown, metadata can be used to identify the testing equipment responsible for inspecting the production equipment. In practical applications, metadata can also represent the production equipment level (EQP, MACHINE, UNIT) for defect calculation and subsequent alarms. In practical applications, the above metadata is generated by the production equipment, production line, or data processing system when the production work order is imported into the production line, based on the production equipment the substrate to be inspected passes through and the relevant parameters to be inspected on the substrate, or it can be pre-set and stored. It can be set according to the specific scenario and is not limited here.

[0097] The aforementioned production data represents the data generated when each substrate under test passes through various production equipment. It is also commonly referred to as production history data, such as ITO film deposition, photoresist coating, etching, film stripping, etc., including the identification code of the substrate under test (GLASSID, identification code of the production equipment, time of entry and exit from the production equipment, etc.).

[0098] The above-mentioned test data is generated when the test station tests the substrate to be tested, such as test time, identification code of the substrate to be tested, identification code of the process station, defect type (code), coordinates of defect point, etc.

[0099] In this embodiment, the data conversion device 200 can convert source data into target data. In other words, it cleans and converts source data from the source database to generate the target data required by the data processing device 300 in this system, and then uploads the target data to the database of the data processing device 300. Furthermore, the database of the data processing device can be implemented using a KUDU database.

[0100] See Figure 4 The data conversion device 200 converts source data into target data, and may include steps 41 to 44.

[0101] In step 41, the data conversion device 200 can acquire production data from the source data, which includes the production time of the substrate to be tested entering the production equipment. Then, the data conversion device 200 can sort the substrates to be tested that have passed through each production equipment according to the production time in the production data, and obtain the sorting result of the substrates to be tested. Here, the production time refers to the moment when the substrate to be tested enters each production equipment for production, that is, the production time of the substrate to be tested is a set of multiple moments that correspond one-to-one with each production equipment.

[0102] For example, in this step, the data conversion device 200 can read production data from the source data, such as production equipment EQP (e.g., CVD equipment), parallel process segments MACHINE (e.g., CVD cleaning unit) on the same equipment, and parallel process units (e.g., specific process chambers in CVD) on the same equipment, and sort them according to the time when the substrate to be tested enters and / or leaves each production equipment (or EQP, MACHINE, or UNIT), that is, the substrate to be tested with the earlier production time should be sorted before the substrate to be tested with the later production time.

[0103] In step 42, the data conversion device 200 can sequentially divide a preset number (e.g., K) of substrates to be inspected into the same unit according to the sorting result, thus obtaining multiple analysis units. Here, K is a positive integer greater than or equal to 2. It is understood that the larger the number of substrates to be inspected in each analysis unit, the more accurate the location data of recurring defects will be. Therefore, the value of K can be selected according to the specific scenario, such as based on the production speed in the production process. In one example, K is set to 28. In this step, K substrates to be inspected can be divided into one analysis unit to facilitate subsequent processing of data from one analysis unit at a time, reducing the amount of data processed each time and facilitating timely detection of recurring defects.

[0104] In step 43, the data conversion device 200 can obtain the number M of substrates to be tested that have passed through the testing station in each analysis unit based on the detection data in the source data. In practical applications, substrates to be tested may be defective after passing through the production equipment and removed from the process flow, failing to reach the testing station corresponding to the production equipment. This results in the number M of substrates to be tested in the same analysis unit reaching the testing equipment being less than or equal to K. The purpose of obtaining the above-mentioned number M in this step is: first, to identify the valid substrates to be tested in the analysis unit and obtain the corresponding source data; second, to determine whether to process the data of the analysis unit based on the arrival rate, see step 44. It is understood that obtaining the number M of substrates to be tested in step 43 can be done by obtaining all the data of each analysis unit before counting, or it can be done by obtaining the number of each substrate to be tested in each analysis unit in real time. The method of obtaining the number M can be selected according to the specific scenario, and is not limited here.

[0105] In step 44, the data conversion device 200 can obtain the arrival rate of the substrate to be tested in each of the above-mentioned analysis units. Data from analysis units whose arrival rates exceed a preset arrival rate threshold are used as the converted target data. Then, the data conversion device 200 can store the target data into the KUDU database of the data processing device. In practical applications, data from analysis units whose arrival rates are less than the aforementioned arrival rate threshold can also be stored in the KUDU database to ensure the integrity and traceability of the detection data.

[0106] Based on the above, the target data in this embodiment may include, but is not limited to, arrival rate, identification code of production equipment, identification code of substrate to be tested, serial number of analysis unit, identification code of process station and time when substrate to be tested leaves process station, and detection data of each analysis unit. These can be set according to specific scenarios and are not limited here.

[0107] In this step, the arrival rate of the substrates to be tested in each analysis unit can serve the following purposes: First, the arrival rate determines whether to process the data of the analysis unit, i.e., the target data, thereby obtaining analysis units that can characterize repeatable defects and achieving the effect of screening analysis units. Second, by setting an arrival rate threshold, when the arrival rate exceeds the threshold, the analysis unit can be used as target data for processing repeatable defects, without waiting to obtain the source data (such as test data) of all substrates to be tested before data conversion or data upload, which is beneficial to improving data upload efficiency and thus allowing the processing time of the target data of the analysis unit to be advanced, so as to facilitate timely detection of defects. Alternatively, when the arrival rate exceeds the threshold, the analysis unit can be used as target data for processing repeatable defects. In this case, the target data can be cached locally and processed after all the data of all substrates to be tested in the analysis unit is obtained. Data of analysis units with an arrival rate below the threshold can be discarded, which helps to reduce the amount of stored data and improve the efficiency of storage space utilization.

[0108] In this embodiment, the system may include a data processing device 300, which can communicate with the data conversion device 200 to obtain the target data.

[0109] In step 22, the detection data of multiple substrates to be tested containing defects in each of the analysis units are obtained.

[0110] In this embodiment, the data processing device 300 in this system can use the Spark computing engine to process the target data, leveraging Spark's parallel in-memory computing capabilities to achieve rapid data analysis and computation. The data processing device can acquire the inspection data of multiple substrates to be inspected containing defects in each analysis unit. It is understood that the data processing device 300 can determine the identification code of the inspection equipment based on the production equipment and metadata in the target data, and then extract the inspection data including the identification code of the substrate to be inspected from the inspection data corresponding to the identification code of the inspection equipment, thus obtaining the inspection data of multiple substrates to be inspected containing defects, or directly extract the inspection data of multiple substrates to be inspected containing defects from the target data.

[0111] In step 23, the defective points in the multiple substrates to be tested are superimposed to obtain a defective point dataset.

[0112] In this embodiment, the data processing device 300 can directly superimpose the defective points from the multiple substrates to be tested to obtain a defective point dataset. Alternatively, the data processing device 300 can obtain the arrival rate of each analysis unit from the target data and compare the arrival rate with a preset arrival rate threshold. When the arrival rate is greater than or equal to the aforementioned arrival rate threshold, the analysis unit is determined to be processed. In this case, the data processing device 300 can superimpose the defective points from the multiple substrates to be tested to obtain a defective point dataset corresponding to each analysis unit.

[0113] The coordinates of the defective points on the substrate to be inspected are generated by the inspection station when inspecting the substrate to be inspected. For example, after the substrate to be inspected enters the inspection station, the image sensor in the inspection station acquires the image of the substrate to be inspected, and uses the preset defective point detection model to determine the defective points on the image. The defective points of each substrate to be inspected are then summarized and stored in the distributed storage device.

[0114] The overlay process can include: in the same coordinate system, for each analysis unit, the data processing device 300 can sequentially place or map the defect points of each substrate to be inspected into or onto the aforementioned coordinate system, thereby forming a two-dimensional defect point map. This two-dimensional defect point map can be equivalent to the defect point dataset D(p1, p2, ..., p) corresponding to the analysis unit. n ). Where p i (i takes values ​​from 1 to n) represents the coordinate data of a defective point in the defective point dataset, including the horizontal coordinate, the vertical coordinate, and the identification code of the substrate to be inspected.

[0115] In step 24, clustering calculations are performed on the defective point dataset to obtain the first defective point cluster.

[0116] In this embodiment, the data processing device 300 can obtain defective points that are repeated in the same area of ​​different substrates under test by superimposing the defective points of multiple substrates under test, thus obtaining defective point clusters. This achieves the effect of timely detection of production equipment with repeated defects, which is beneficial to improving the yield of substrates under test production.

[0117] In this embodiment, repeating defective points are defined as follows: Let point p... i (p i If the number of defective points in the area formed by taking (∈D) as the reference point and r (adjustable) as the distance threshold exceeds the quantity threshold, the defective points in the above area are determined to be duplicate defective points.

[0118] In this embodiment, the first defect cluster is obtained, see [link to documentation]. Figure 5 This includes steps 51 to 53: In step 51, the defective point dataset D(p1,p2,......,p) is used. nAny defective point p in ) i Using the reference point as the reference point, the number of defective points m within the region formed by the distance threshold r (adjustable) is used to obtain the initial defective point cluster R.

[0119] In step 52, the defective point dataset D(p1,p2,......,p) is traversed. n From the defective points in the data, we obtain multiple initial defective point clusters.

[0120] In step 53, multiple initial defective point clusters are obtained that satisfy the condition that the defective point data m in the initial defective point cluster set is greater than or equal to the first threshold, thus obtaining the first defective point cluster set.

[0121] For example, with point p i (p i Using (∈D) as the center and a distance threshold r (adjustable) as the radius, obtain circle C. i Calculate C i This includes the number m of substrates to which the defects belong, when m ≥ m t (e.g., 3-5, can be adjusted) then at point p i There is a cluster of recurring defects at this point, in which case p can be... i The defective points within the corresponding circle are treated as a cluster of points.

[0122] After traversing each defective point as a center, the data processing device 300 can obtain the defective point dataset D(p1, p2, ..., p...). n The defect clusters corresponding to each defect point in the dataset are used to obtain the first defect cluster set R = (R... 1, R 2, ......,R n ).

[0123] In other words, the first cluster of defective points is composed of each defective point p. i The set of defective points covered by a circle centered at a distance threshold r and with a radius of r. The distance threshold can be determined based on the alignment deviation of the substrate under test within the analysis unit, such as 3-10 pixels. This alignment deviation is the offset of a point at the same location when the image sensor at the detection station acquires images of two substrates under test, caused by the mechanical structure that moves the substrate under test.

[0124] In one embodiment, considering that the number of defective points is usually large, i.e., the dataset of defective points is large, resulting in a large amount of subsequent computation, this disclosure can use a neighborhood search method to obtain the first cluster of defective points. For example, the neighborhood search method is implemented using the kd-tree algorithm, with r (adjustable) as the distance threshold and point p as the distance threshold. i (p iTaking (∈D) as the reference point, find the number m of defective points in its neighborhood, forming a cluster R of defective points. i When R i The number of substrates to be tested, m ≥ m, of which the defects belong to t When all defective points p are in the initial point cluster R, store them in the initial point cluster R. i After completing the traversal, the first cluster of defective points can be obtained. Thus, in this embodiment, the neighborhood search algorithm enables rapid searching of nearest neighbors within a two-dimensional space of large-scale data, improving data processing speed and enhancing the efficiency of defect detection.

[0125] In step 25, alarm information for the location of the device contact point corresponding to the first defect cluster is generated based on the first defect cluster.

[0126] In this embodiment, the data processing device 300 can generate alarm information about the location of the equipment contact points corresponding to the first defect cluster based on the first defect cluster. See [link to relevant documentation]. Figure 6 This includes steps 61 to 62.

[0127] In step 61, the data processing device 300 can filter multiple first defective point clusters, filtering out first defective point clusters with defective points having the same location information to obtain a second defective point cluster. For example, the data processing device 300 can filter the initial point cluster set R = (R1, R2, ..., R... n In a point cluster set R, there are defective points with the same location information or two points that overlap in the above two-dimensional graph. For example, in the point cluster set R, point cluster R... i =(p l ,p l+1 ,p l+2 ,p l+3 ,......), point cluster R j =(……,P l-1 ,P l ,P l+1 ,P l+2 There are overlapping points (p) between the two. l ,p l+1 ,p l+2 ),in p l ∈D. Thus, by filtering defective points with the same location information across different point clusters, this step can eliminate repetitive alarms for defects occurring in the same area. Based on the above, the process by which the data processing device 300 filters defective points with the same location information is described in [reference needed]. Figure 7 This includes steps 71 to 76.

[0128] In step 71, the data processing device 300 can acquire the first defect cluster set R = (R1, R2, ..., R...). n The number of defective points in each point cluster within the initial point cluster. It is understood that the number of defective points in each point cluster can be obtained in real time in step 71; alternatively, the number of defective points in each point cluster can be stored as attribute data of each point cluster in the initial point cluster set when the initial point cluster set is obtained, and can be directly read in step 71, thereby reducing the processing steps and speeding up the processing efficiency.

[0129] In step 72, the data processing device 300 can sort the clusters within the first defective cluster set according to their quantity, for example, sorting them from largest to smallest.

[0130] In step 73, the data processing device 300 can obtain the cluster with the most defective points in the first defective point cluster set as a candidate cluster.

[0131] In step 74, the data processing device 300 can obtain the intersection-union ratio (IUR) of any point cluster and a candidate point cluster within the first defective point cluster set, and remove point clusters from the first defective point cluster set whose IUR exceeds a preset IUR threshold. For example, candidate point cluster R i With point cluster R j intersection ratio When T >= t (t∈[0,1]), the point cluster R j Remove from R. If the value is less than t, retain the point cluster R. j .

[0132] In step 75, the data processing device 300 can move candidate point clusters from the first defective point cluster set into a preset second defective point cluster set. Inside.

[0133] In step 76, if the first defective point cluster set is a non-empty set, the data processing device 300 can continue to execute the step of obtaining the point cluster with the most defective points in the first defective point cluster set as a candidate point cluster, that is, it can only end after traversing all point clusters in the first defective point cluster set, i.e., re-execute steps 73 to 76; otherwise, it determines that a second defective point cluster set has been obtained. The results are shown in Figure 8. Figure 8(a) shows the effect of five clusters in the first defect cluster set, and Figure 8(b) shows the effect of three clusters in the second defect cluster set. Comparing Figure 8(a) and Figure 8(b), clusters R1 and R5 are filtered out.

[0134] In step 62, the data processing device 300 can generate alarm information for the location of the equipment contact point corresponding to the second defect cluster based on the defect cluster.

[0135] In this embodiment, the data processing device 300 obtains the cluster of points with repeated defects by acquiring the second defect cluster. Then, the data processing device 300 can query the identification code of the substrate to be inspected based on the defect points in the cluster, and obtain the identification code of the corresponding production equipment based on the production data of the substrate to be inspected to determine the production equipment with malfunction. That is, the method provided in this embodiment locates the production equipment causing the repeated defects. Furthermore, the data processing device 300 can determine the location information of the defect points based on the cluster of points with repeated defects, and determine the abnormality of a certain component of the production equipment based on the location information and the production equipment, or in other words, obtain the location of the equipment contact point corresponding to the defect dataset.

[0136] In this embodiment, after obtaining the location of the equipment contact point, the data processing device 300 can collect alarm information for the corresponding equipment contact point location from the defect cluster. The equipment contact point location can include at least one of the following: defect coordinates, the identification code of the substrate to be inspected, and the identification code of the production equipment. Here, the alarm information can also include defect cause data and the time of defect occurrence, which can be set according to the specific scenario and is not limited here. Then, the data processing device 300 can send the information to the user through a specified method (such as SMS, email, push notification, etc.) to promptly detect problems and provide early warnings to prevent losses and maintain yield.

[0137] For example, a user can log in to the data processing system, find the BP Common Defect monitoring function in the function subscription interface, and click subscribe. Then, they can input the process sites to be monitored, as well as calculation parameters (the number of analysis units (Batch) K, arrival rate, defect matching accuracy, number of defective GLASSes, Intersection over Union (IOU) threshold, and other key parameters). Finally, they can select the abnormal alarm method and confirm the subscription takes effect. After a user subscribes to messages from the data processing system, the data processing device can send alarm information to subscribed users via mobile portal push notifications based on the subscription configuration information when generating alarm information, thus promptly reminding users.

[0138] Figure 9 This is a flowchart illustrating a data processing method according to an exemplary embodiment, which can be applied to... Figure 1 The data processing system shown. See also Figure 9 A data processing method, comprising steps 91 to 93:

[0139] In step 91, in response to the operation of the subscription monitoring function, a preset value for at least one subscription parameter related to recurring defect detection is acquired. In this step, the data processing system includes a display device. This display device may be a touch screen displaying the function subscription interface, as shown below. Figure 10As shown. After logging into the data processing system, the user can find the subscription monitoring function (such as the BP Common Defect monitoring function) in the function subscription interface, and then click to subscribe. In the function subscription interface, the user enters the preset value of at least one subscription parameter related to recurring defect detection. For example, the user can input the process station to be monitored, and the calculated parameters can include key parameters such as the number of analysis units (Batch) K, arrival rate (e.g., 80%), defect matching accuracy (e.g., 0.5mm), number of defective GLASSes (e.g., 3), Intersection over Union (IOU) threshold (e.g., 0.8), and the abnormal alarm method (e.g., mobile portal message), etc., and then confirm the subscription is effective. In this way, the touch display can collect the preset values ​​of at least one of the above subscription parameters and send them to the data processing device.

[0140] In step 92, target data is acquired according to the preset values, and the target data is processed to obtain the first defect cluster set corresponding to each analysis unit. In this step, the data processing device can acquire the defect cluster set corresponding to each analysis unit according to the above preset values. For the specific acquisition process, please refer to [link to relevant documentation]. Figure 2 The details of the illustrated embodiments will not be repeated here.

[0141] In step 93, alarm information for the device contact point location corresponding to the first defect cluster is generated based on the defect cluster. In this step, the data processing device can generate alarm information, as detailed in step 25. After generating the alarm information, it can also send the alarm information to subscribed users via mobile portal message push according to a preset alarm method, thereby promptly reminding users. In this example, the data processing device can send the alarm information to a touch screen display, which will then display the alarm information, as shown in the example. Figure 11 As shown.

[0142] Based on the above data processing methods, this disclosure provides a data processing system, see [link to relevant documentation]. Figure 1 It includes: a distributed storage device 100, a data conversion device 200, and a data processing device 300;

[0143] The distributed storage device is configured to store source data.

[0144] The data conversion device includes one or more first processors configured to convert the source data into target data; the target data includes data from multiple analysis units.

[0145] The data processing device includes one or more second processors, which are configured to acquire detection data of multiple substrates to be tested containing defects in each analysis unit, and superimpose the defects in the multiple substrates to be tested to obtain a defect dataset, perform clustering calculation on the defect dataset to obtain a first defect cluster, and generate alarm information for the device contact point location corresponding to the first defect cluster based on the first defect cluster.

[0146] Optionally, the source data includes at least one of metadata, production data, and testing data; the metadata represents the association between process stations and testing stations, the production data represents the production history data of each substrate to be tested, and the testing data is the data generated when the testing station tests the substrate to be tested.

[0147] Optionally, the one or more second processors are further configured to perform:

[0148] Based on the production data in the source data, the substrates to be tested are sorted according to the production time of each production equipment to obtain the sorting result of the substrates to be tested.

[0149] Based on the sorting results, a preset number of substrates to be tested are sequentially divided into the same unit to obtain multiple analysis units;

[0150] Based on the detection data in the source data, the number of substrates to be tested that have passed through the detection station in each analysis unit is obtained;

[0151] The arrival rate of the substrate to be detected in each analysis unit is obtained, and the data of the analysis units whose arrival rate exceeds the preset arrival rate threshold is used as the converted target data.

[0152] Optionally, the one or more second processors are further configured to perform clustering calculations on the defective point dataset to obtain a first defective point cluster set, including:

[0153] Using any defective point Pi in the defective point dataset as a reference point, obtain the number m of defective points within the region formed by r as a distance threshold, and obtain the initial defective point cluster;

[0154] Traverse the defective points in the defective point dataset to obtain multiple initial defective point clusters;

[0155] Obtain multiple initial defective point clusters that satisfy the condition that the number of defective points m in the initial defective point cluster is greater than or equal to the first threshold, and obtain the first defective point cluster set.

[0156] Optionally, the one or more second processors are further configured to execute alarm information for generating device contact point locations corresponding to the first defect cluster based on the first defect cluster, including:

[0157] Multiple first defect point clusters are filtered out, and first defect point clusters with defect points having the same location information are filtered out to obtain second defect point clusters.

[0158] Alarm information for the location of the device contact point corresponding to the second defect cluster is generated based on the second defect cluster.

[0159] Optionally, the one or more second processors are further configured to perform filtering on a plurality of first defect clusters, filtering first defect clusters with defective points having the same location information, including:

[0160] Obtain the number of defective points in each cluster within the first defective point cluster set;

[0161] The point clusters within the first defective point cluster set are sorted according to the stated quantity;

[0162] The cluster with the most defective points in the first defective point cluster is selected as the candidate point cluster.

[0163] Obtain the intersection-union ratio (CURNR) of any point cluster within the first defective point cluster set and the candidate point cluster, and remove point clusters from the first defective point cluster set whose CURNR exceeds a preset CURNR threshold.

[0164] The candidate point cluster is moved from the first defective point cluster set into a preset second defective point cluster set;

[0165] If the first defective point cluster is a non-empty set, continue to execute the step of obtaining the point cluster with the most defective points in the first defective point cluster as a candidate point cluster; otherwise, determine that the second defective point cluster is obtained.

[0166] Optionally, the one or more second processors are further configured to perform:

[0167] The corresponding production equipment is controlled according to the location of the contact point of the equipment, so that the production equipment performs a protective action.

[0168] Optionally, the location of the device contact point includes at least one of the following: coordinates of the defective point, identification code of the substrate to be tested, and identification code of the production equipment.

[0169] Optionally, it also includes a display device; the display device includes at least one display screen; the at least one display screen is configured to display the alarm information.

[0170] It is understood that the system provided in this disclosure corresponds to the method described above, and the specific details can be found in the various embodiments of the method, which will not be repeated here. In an exemplary embodiment, a computer-readable storage medium is also provided, such as a memory including instructions, wherein the executable computer program described above can be executed by a processor. The readable storage medium may be a ROM, random access memory (RAM), CD-ROM, magnetic tape, floppy disk, or optical data storage device, etc.

[0171] Other embodiments of this disclosure will readily occur to those skilled in the art upon consideration of the specification and practice of the disclosure herein. This disclosure is intended to cover any variations, uses, or adaptations that follow the general principles of this disclosure and include common knowledge or customary techniques in the art not disclosed herein. The specification and examples are to be considered exemplary only, and the true scope and spirit of this disclosure are indicated by the following claims.

[0172] It should be understood that this disclosure is not limited to the precise structures described above and shown in the accompanying drawings, and various modifications and changes can be made without departing from its scope. The scope of this disclosure is limited only by the appended claims.

Claims

1. A data processing method, characterized in that, include: Obtain the target data; The target data includes data from multiple analysis units; Acquire detection data from multiple substrates to be tested that contain defects in each of the analysis units; A defect dataset is obtained by superimposing the defective points from the multiple substrates to be inspected. The first cluster set of defective points is obtained by performing clustering calculations on the defective point dataset; Based on the first defect cluster, generate alarm information for the location of the device contact point corresponding to the defect cluster; The production equipment is controlled according to the location of the equipment contact point, so that the production equipment performs protective actions; Obtain the target data, including: The production data is read from the source data of the distributed storage device, and the substrates to be tested are sorted according to the production time of each production device to obtain the sorting result of the substrates to be tested. Based on the sorting results, a preset number of substrates to be tested are sequentially divided into the same unit to obtain multiple analysis units; Based on the detection data in the source data, the number of substrates to be tested that have passed through the detection station in each analysis unit is obtained; The arrival rate of the substrate to be detected in each analysis unit is obtained, and the data of the analysis units whose arrival rate exceeds the preset arrival rate threshold is used as the converted target data.

2. The data processing method according to claim 1, characterized in that, Clustering the defective data set yields the first defective cluster set, which includes: Using any defective point Pi in the defective point dataset as a reference point, the number m of defective points within the region formed by the distance threshold r is obtained, thus obtaining the initial defective point cluster; Traverse the defective points in the defective point dataset to obtain multiple initial defective point clusters; Obtain multiple initial defective point clusters that satisfy the condition that the number of defective points m in the initial defective point cluster is greater than or equal to the first threshold, and obtain the first defective point cluster set.

3. The data processing method according to claim 2, characterized in that, Based on the first defect cluster, alarm information is generated for the location of the device contact point corresponding to the defect cluster, including: Multiple first defect point clusters are filtered out, and first defect point clusters with defect points having the same location information are filtered to obtain second defect point clusters. Alarm information for the location of the device contact point corresponding to the second defect cluster is generated based on the second defect cluster.

4. The data processing method according to claim 3, characterized in that, Multiple clusters of first defects are filtered to identify clusters of first defects with the same location information, including: Obtain the number of defective points in each cluster within the first defective point cluster set; The point clusters within the first defective point cluster set are sorted according to the stated quantity; The cluster with the most defective points in the first defective point cluster is selected as the candidate point cluster. Obtain the intersection-union ratio (CURNR) of any point cluster within the first defective point cluster set and the candidate point cluster, and remove point clusters from the first defective point cluster set whose CURNR exceeds a preset CURNR threshold. The candidate point cluster is moved from the first defective point cluster set into a preset second defective point cluster set; If the first defective point cluster is a non-empty set, continue to execute the step of obtaining the point cluster with the most defective points in the first defective point cluster as a candidate point cluster; otherwise, determine that the second defective point cluster is obtained.

5. The data processing method according to claim 1, characterized in that, The location of the device contact point includes at least one of the following: coordinates of the defective point, identification code of the substrate to be tested, and identification code of the production equipment.

6. A data processing method, characterized in that, include: In response to the operation of the subscription monitoring function, a preset value of at least one subscription parameter related to the detection of recurring defects is collected; The target data is obtained according to the preset value, and the target data is processed to obtain the first defect cluster set corresponding to each analysis unit; Based on the first defect cluster, generate alarm information for the location of the device contact point corresponding to the defect cluster; The production equipment is controlled according to the location of the equipment contact point, so that the production equipment performs protective actions; Obtain the target data, including: The production data is read from the source data of the distributed storage device, and the substrates to be tested are sorted according to the production time of each production device to obtain the sorting result of the substrates to be tested. Based on the sorting results, a preset number of substrates to be tested are sequentially divided into the same unit to obtain multiple analysis units; Based on the detection data in the source data, the number of substrates to be tested that have passed through the detection station in each analysis unit is obtained; The arrival rate of the substrate to be detected in each analysis unit is obtained, and the data of the analysis units whose arrival rate exceeds the preset arrival rate threshold is used as the converted target data.

7. A data processing system, characterized in that, include: Distributed storage devices, data conversion devices, and data processing devices; The distributed storage device is configured to store source data. The data conversion device includes one or more first processors configured to convert the source data into target data; the target data includes data from multiple analysis units. The data processing device includes one or more second processors, which are configured to acquire detection data of multiple substrates to be tested containing defects in each analysis unit, and superimpose the defects in the multiple substrates to be tested to obtain a defect dataset, and perform clustering calculation on the defect dataset to obtain a first defect cluster. Based on the first defect cluster, alarm information is generated for the location of the equipment contact point corresponding to the defect cluster; based on the location of the equipment contact point, the corresponding production equipment is controlled to perform a protection action; The one or more first processors are further configured to execute: The production data is read from the source data of the distributed storage device, and the substrates to be tested are sorted according to the production time of each production device to obtain the sorting result of the substrates to be tested. Based on the sorting results, a preset number of substrates to be tested are sequentially divided into the same unit to obtain multiple analysis units; Based on the detection data in the source data, the number of substrates to be tested that have passed through the detection station in each analysis unit is obtained; The arrival rate of the substrate to be detected in each analysis unit is obtained, and the data of the analysis units whose arrival rate exceeds the preset arrival rate threshold is used as the converted target data.

8. The data processing system according to claim 7, characterized in that, The source data includes at least one of metadata, production data, and testing data; the metadata represents the relationship between process stations and testing stations, the production data represents the production history data of each substrate to be tested, and the testing data is the data generated when the testing station tests the substrate to be tested.

9. The data processing system according to claim 7, characterized in that, The one or more second processors are further configured to perform clustering calculations on the defective point dataset to obtain a first defective point cluster set, including: Using any defective point Pi in the defective point dataset as a reference point, the number m of defective points within the region formed by the distance threshold r is obtained, thus obtaining the initial defective point cluster; Traverse the defective points in the defective point dataset to obtain multiple initial defective point clusters; Obtain multiple initial defective point clusters that satisfy the condition that the number of defective points m in the initial defective point cluster is greater than or equal to the first threshold, and obtain the first defective point cluster set.

10. The data processing system according to claim 9, characterized in that, The one or more second processors are further configured to execute alarm information based on the first defect cluster, generating alarm information for the location of the device contact point corresponding to the defect cluster, including: Multiple first defect point clusters are filtered out, and first defect point clusters with defect points having the same location information are filtered out to obtain second defect point clusters. Alarm information for the location of the device contact point corresponding to the second defect cluster is generated based on the second defect cluster.

11. The data processing system according to claim 10, characterized in that, The one or more second processors are further configured to perform filtering on multiple first defect clusters, filtering first defect clusters with defective points having the same location information, including: Obtain the number of defective points in each cluster within the first defective point cluster set; The point clusters within the first defective point cluster set are sorted according to the stated quantity; The cluster with the most defective points in the first defective point cluster is selected as the candidate point cluster. Obtain the intersection-union ratio (CURNR) of any point cluster within the first defective point cluster set and the candidate point cluster, and remove point clusters from the first defective point cluster set whose CURNR exceeds a preset CURNR threshold. The candidate point cluster is moved from the first defective point cluster set into a preset second defective point cluster set; If the first defective point cluster is a non-empty set, continue to execute the step of obtaining the point cluster with the most defective points in the first defective point cluster as a candidate point cluster; otherwise, determine that the second defective point cluster is obtained.

12. The data processing system according to claim 7, characterized in that, The location of the device contact point includes at least one of the following: coordinates of the defective point, identification code of the substrate to be tested, and identification code of the production equipment.

13. The data processing system according to claim 7, characterized in that, It also includes a display device; the display device includes at least one display screen; the at least one display screen is configured to display the alarm information.

14. A computer-readable storage medium, characterized in that, When the executable computer program in the storage medium is executed by a processor, it can implement the method as described in any one of claims 1 to 6.

Images