Defect source identification method and apparatus

Abstract

A defect source identification method and apparatus are provided. The method includes obtaining image data of at least one object, wherein the image data includes defect features of at least one defect located on a surface of the object; obtaining, based on an identification of the object, a plurality of feature profile data of a plurality of operation tools operating on the surface of the object, wherein each feature profile data includes contact profile data representing a region profile on the surface of the object that is contacted by the operation tool when the operation tool operates on the object; and identifying, based on the defect features and the feature profile data, a corresponding operation tool from the plurality of operation tools that causes one or more defects of the at least one defect.

Description

Technical Field

[0001] This invention relates to chip packaging, and more particularly to the identification of fault machine sources that cause chip defects due to the operation of machines during the chip packaging process. Background Technology

[0002] Yield loss is a persistent problem in chip manufacturing. Chip defects causing yield loss can include scratches or stress damage, leaving defects on the chip surface. Actual inspection reveals that the primary cause of these problems is damage or contamination of the machine's operating end or tools, leading to defects when the chip is handled, such as by touching it. Therefore, identifying the faulty machine causing the chip problems or malfunctions is of paramount importance.

[0003] Conventionally, to identify faulty machines, engineers obtain photographs of the tools used to operate the chips and compare them with defects on the chip surface to determine the suspected tools that might be causing these defects—the sources of failure. However, this conventional manual inspection method is time-consuming to confirm the relevant sources of failure. Moreover, because chips undergo multiple processes and are operated by various tools on the production line, it is difficult to preliminarily determine which tool in which process might be faulty. Therefore, this manual inspection method is difficult and requires a significant amount of manpower. Consequently, only a sample of chips is usually inspected, which can lead to the missed detection of problematic chips. Summary of the Invention

[0004] This invention provides an improved fault source identification scheme. It innovatively introduces the contact contour of each operating tool to characterize each tool and evaluates defects based on the contact contour, thereby quickly identifying the operating tool causing the defect. According to the scheme of this invention, a feature contour map specific to each type of operating tool is predetermined, wherein each feature contour includes a contact contour, which represents the contour of the area on the surface of the object that comes into contact with the operating tool when the tool operates the object. Then, during chip defect analysis, potentially problematic tools can be automatically identified by overlaying the tool-specific feature contour image with a chip surface defect image in a single chip image and confirming whether the defect on the chip surface is located within the corresponding contact area in the feature contour image. Attached Figure Description

[0005] Figure 1A A unit U is schematically shown;

[0006] Figure 1B An exemplary manipulator tip for unit U is schematically shown;

[0007] Figure 2A A schematic diagram of the characteristic profile of the suction head for surface SF1 operation of unit U in FIG1 is shown;

[0008] Figure 2B A schematic feature outline diagram of a plurality of operating tools for unit U is shown as an example;

[0009] Figure 3 The diagram illustrates four nodes on the production line;

[0010] Figure 4 The process for identifying defect sources according to an embodiment of the present invention is illustrated;

[0011] Figure 5 The process for identifying defect sources according to another embodiment of the present invention is illustrated;

[0012] Figure 6 The defect diagram and projection diagram of the cell after defect sampling are shown as an example;

[0013] Figure 7 A flowchart illustrating the automatic identification of defect sources based on yet another example is shown;

[0014] Figure 8A A block diagram of a monitoring device is shown based on an example.

[0015] Figure 8B A block diagram of a controller based on another example is shown. Detailed Implementation

[0016] Before explaining any embodiments of the invention in detail, it should be understood that the application of the invention is not limited to the construction details set forth in the following description or shown in the drawings. The invention can have other embodiments and can be practiced or implemented in various ways. Furthermore, it should be understood that the wording and terminology used herein are for descriptive purposes and should not be considered limiting. In this disclosure, a “cell” (also called an “object”) can refer to a die, or an integrated circuit or circuit board that packages or integrates one or more dies or other electronic units. Furthermore, a handling tool refers to any device that operates on the surface of a chip, such as a pick-up tool for picking up a chip, where the pick-up tool can make direct contact with the chip surface; a handling tool can also be a device that operates indirectly relative to the chip surface, such as a device for implanting solder balls on the back of a chip. According to embodiments of the invention, for the parts or areas on the chip surface that have direct or indirect contact with the handling tool, a touch profile map (Touch Profile) of the handling tool is created. Obviously, chip failures will also necessarily occur within this touch profile area, and the invention uses the touch profile to diagnose defect sources.

[0017] Figure 1A A unit U is schematically shown, such as Figure 1A As shown, unit U here is a circuit board that integrates two cores C1 and C2. For the sake of simplicity, other circuits or components on the circuit board are ignored and not shown. Figure 1B An example of a set of manipulators TooL for cell U is schematically shown, for example, having two manipulator tips for manipulating cores C1 and C2 at surface SF1 of cell U. If the tips are contaminated, for example, due to contaminants, the tips may damage or scratch the two cores C1 and C2 when manipulating cell U. Because the tips have a pre-designed shape and a predetermined positioning relationship relative to cell U, the locations where the tips may damage the two cores C1 and C2 are fixed. Figure 2A The feature profile diagram WF of the two suction heads of tool TooL for operating on surface SF1 of unit U in Figure 1 is shown. (See diagram for reference.) Figure 2A As shown, WF represents the overall profile of unit U in the plane direction of surface SF1, and the indicated graphic size is... Figure 1A The units U shown are all the same size, denoted here as (X,Y), where X represents the horizontal length and Y represents the vertical length. The contact profile (TouchProfile) between the suction head and unit U is located within the feature profile map WF and includes two sub-profile maps, A1 and A2, corresponding to the contact profiles between the suction head and unit U, respectively. Figure 1A The contact areas of cores C1 and C2 are shown, where the positioning relationship of A1 and A2 within the WF is exactly the same as the relative position of C1 and C2 within the entire unit U. The blank areas within sub-contours A1 and A2 represent the contact areas between the tool TooL and cores C1 and C2 of unit U. In this example, the blank parts within contact contours A1 and A2 indicate the parts of cores C1 and C2 exposed to the suction head, i.e., the parts where C1 and C2 directly contact the suction head.

[0018] The touch profile of the suction head can be obtained or generated using any method known in the art. As an example, the touch profile can be generated based on the design parameters of the suction head specifically designed for unit U, such as CAD parameters and the inherent design parameters of unit U, thereby generating the touch profile when the suction head operates on unit U. As mentioned earlier, the touch profile represents the direct contact area or action area when the suction head interacts with unit U. How the design parameters of the suction head, such as CAD parameters, and the design parameters of unit U are used to determine the touch profile can be determined using any method known in the art. In another example, the touch profile of the suction head can also come from a third party, such as the suction head supplier, who can generate the corresponding touch profile because the suction head is designed specifically for unit U. According to one example of the invention, the contour feature map WF data containing the touch profile is stored in a database DB. Therefore, according to an embodiment of the present invention, contour feature data WF of all operating tools used for processing by each unit can be stored in the database DB in advance for subsequent use, wherein each contour feature data contains the touch profile TouchProfile data when the corresponding operating tool processes the unit U. Figure 2B An exemplary feature outline diagram of a series of operating tools used to operate unit U on a production line is schematically shown. It should be noted here that... Figure 2B The illustration only shows the feature outline of a portion of the operating tools, but the operating tools used to operate the unit on the production line may also include other operating tools, which will not be described in detail here.

[0019] Each unit U goes through multiple nodes on the production line, and at each node, a certain type of operation is performed on unit U. That is, at each node, a specific operation tool is used to perform a specific operation on unit U. For example... Figure 3As shown, this schematically illustrates four nodes, Node_1 to Node_4, on the production line. At node_1, the first type of operation on surface SF1 of unit U is performed by tool CT1. Similarly, at node_2, the second type of operation on surface SF1 of unit U is performed by tool CT2, at node_3 by tool CT3, and at node_4 by tool CT4. Typically, each unit U in a batch of unit {UNIT1} on the production line undergoes the same process processing by tools CT1-CT4. It should be noted that a tool, such as CT2, at a node on the production line represents a type of tool, not a limited number of tools. That is, there may actually be multiple CT2 tools at node_2. 2-1 CT 2-2 CT 2-3 Each node processes different cells in a batch of cells at node Node_2; the same applies to other nodes.

[0020] According to the present invention, when each unit U completes the process processing at all nodes on the production line, the unit surface SF1 processed by the operating tools CT1-CT4 on the production line is usually photographed at the last node Node_4 to obtain an image of the surface SF1. This yields images of the surface SF1 of each unit U in, for example, 100 units in a batch {UNIT1} on the production line, hereinafter referred to as:

[0021] U1_Img_SF1,

[0022] U2_Img_SF1,

[0023] U3_Img_SF1,

[0024] ...

[0025] U100_Img_SF1.

[0026] Furthermore, according to embodiments of the present invention, the information of the operating tool CT that processes each unit U at each node is stored as historical record information HisInfor and associated with the characteristic information of the unit U, such as the identity identifier U_ID, so as to trace the processing history of the unit U. For example, in the above example, for the current unit U with the identifier U40_ID, it can be determined that it has experienced the operating tool CT on the production line. 1-2 CT 2-2 CT 3-2 CT 4-2 The processing.

[0027] Figure 4 The diagram illustrates a fault source identification process according to an embodiment of the present invention. As shown, in step 401, the image of unit U obtained at the last node Node_4 on the production line is read, for example, the image U1_Img_SF1 of unit U1 is read.

[0028] In step 403, for image U1_Img_SF1, it is determined whether a defect exists on surface SF1 of identification unit U1. Here, known image processing and computer vision algorithms can be used for defect identification, such as OpenCV, a cross-platform computer vision and machine learning software library running on Linux, Windows, Android, and Mac OS operating systems. OpenCV can be trained to automatically identify anomalies, i.e., defects, on the chip surface and provide the defect features. According to one example of the invention, the defect feature can be a location indicator of one or more sampling points of the defect on the unit surface, where p(x,y) represents each sampling point or defect feature, where x represents the pixel abscissa within the plane defined by the unit's planar dimensions X and Y, and y represents its pixel ordinate. It should be noted that there may be more than one defect in a unit U, although these defects may be accidental or systematic. Accidental defects refer to defects appearing on one or a few units due to accidental factors, while systematic defects refer to defects appearing on the vast majority of units processed by the tool CT due to the tool's characteristics.

[0029] In step 405, based on the read identification U1_ID of unit U1 and the pre-stored production history traceability information HisInfor, the tool CT that operated on it is determined, and the feature contour map WF of the operating tool CT is read from the database. For example, in this case, based on the identification U1_ID of unit U1 and the historical traceability information HisInfor, it can be determined that U1 has undergone operation with the tool CT on the production line. 1-1 CT 2-1 CT 3-1 CT 4-1 The processing; therefore, further, based on the CT operating tool 1-1 CT 2-1 CT 3-1 CT 4-1 CT logo 1-1 _Id~CT 4-1 The _Id is used to read its respective feature contour maps WF1 to WF4 from the database DB. Each feature contour map WF1 to WF4 contains the operation tool CT. 1-1 CT 2-1 CT3-1 CT 4-1 Each has its own specific contact profile, TouchProfile_1 to TouchProfile_4.

[0030] In step 407, the defect features p(x,y) of unit U1 are determined and transferred to the operating tool CT. 1-1 CT 2-1 CT 3-1 CT 4-1 The projections of the contact profiles TouchProfile_1 to TouchProfile_4 are as follows. As mentioned earlier, each defect feature p(x,y) is its position coordinate in the coordinate system of the plane defined by the planar dimensions (X,Y). The profile feature WF has the same planar dimensions X,Y as the element. Therefore, the projection here involves determining the position of each defect feature p(x,y) within the profile feature WF.

[0031] In step 409, if it is determined that the projections of each defect feature p(x,y) of unit U1, or at least a certain proportion (e.g., more than 80%) of the defect features fall within a certain contact profile, for example, within the range of contact profile TouchProfile_1, then it indicates that the defects on unit U1 are highly likely to be caused by the operating tool CT corresponding to contact profile TouchProfile_1. 1-1 This is caused by [the CT scanner], therefore an indicator signal can be output to instruct the operating tool [to respond]. 1-1 This may have caused a defect in unit U1, and verification is recommended. In another example, consider the case where each defect feature p(x,y) of unit U1 is projected onto different contact profiles at different proportions. For instance, 60% of the defect features fall within the range of contact profile TouchProfile_2, 30% fall within the range of contact profile TouchProfile_1, and 10% fall within the range of contact profile TouchProfile_3. Based on the proportion of defect features falling within the contact profiles, the suspected operating tools are ranked. For example, in this case, tool CT... 2-1 The most suspicious, and the tool is CT. 3-1 The least suspicious item is the tool used; therefore, an indicator signal can be output to identify the suspected operating tools that may cause unit defects for the operator's reference. It should be noted that defect features falling within different contact profiles may have overlapping portions or may not overlap, but the above judgment method applies to both.

[0032] In another example, for the currently selected unit U1, all the operating tools CT1-CT4 associated with unit U1 can be graphically displayed, while the operating tool that causes the defect on U1, such as CT1, is displayed by highlighting, for example, or marking.

[0033] In another example of the invention, since there may be more than one defect on unit U1, the various defect features on the current unit U1 and all the operating tools CT1-CT4 associated with unit U1 can be displayed in contrast on the graphical user interface (GUI). The user can select a portion of the defect features on the current unit U1 on the graphical interface. For the selected defect feature, step 409 is performed, that is, determining which of the contact contours TouchProfile_1 to TouchProfile_4 the projection of the selected defect feature p(x,y) falls within. For example, if it falls within the range of contact contour TouchProfile_2, then the operating tool causing the selected defect on U1 is determined to be CT2, and CT2 is highlighted on the graphical interface.

[0034] Therefore, according to the solution of the present invention, the defect source causing the defect on the unit can be quickly given a prompt or suggestion, that is, which operating tool may have caused the defect. Moreover, the solution can improve the efficiency of manual inspection and judgment of defects, and is therefore suitable for large-scale defect diagnosis.

[0035] Figure 5 This paper illustrates a defect source identification process according to another embodiment of the present invention. According to this embodiment, centralized defect diagnosis can be performed on batches of units on a production line. Here, we will still use... Figure 3 The following example illustrates the process using four nodes on the production line.

[0036] As shown in the figure, in step 501, image data of all units U obtained at the last node Node_4 on the production line, for example, 200 units in a batch {UNIT1}, are read. In this example, the image data can be defect image data that has been processed for defect identification, which includes the defect features p of the identified defects.

[0037] In step 503, the defects p of all units U1-U200 are formed into a set {P} and visually presented in a single unit to form a defect map of the batch of units. Figure 6 Figure (A) in the figure exemplarily illustrates the image outline UF of a single unit, which is still referred to here as Figure 1B The example shown illustrates the defects caused by the operating nozzle to cores C1 and C2. It includes a defect profile, which contains two contour maps. One contour map area contains the defect features generated on the surface SF1 of core C1 of all units U1-U200 during processing on the production line, while the other contour map area contains the defect features formed on the surface SF1 of core C2 of all units U1-U200.

[0038] In step 505, based on the read unit U's identity identifier U_ID and the pre-stored production history traceability information HisInfor, the tool CT for operating on it is determined. Figure 3 The example batch unit {UNIT1} can be identified as having used operating tools CT1-CT4 on the production line. Furthermore, based on the identifiers of tools CT1-CT4, their feature profiles WF1-WF4 are read from the database DB. Since all units in the same batch {UNIT1} undergo the same process on the production line, although different tools may be used at the same node for each unit U within the same batch {UNIT1}, they will have the same feature profile because the operating tools at the same node are of the same type. Figure 6 Figure (B) in B schematically illustrates the feature profile WF1 of one type of operating tool, such as CT1. As previously described, the feature profile WF1 has the unit planar features processed by tool CT1. These unit planar features include not only the planar view dimensions (X, Y) of unit U, but also the contact profile TouchProfile_1 of the contact area or interaction area when tool CT operates unit U. The profile TouchProfile_1 includes not only the contact profile features, but also their orientation information within the entire feature profile WF1.

[0039] In step 507, based on the determined defect feature Defect_Profile and contour map TouchProfile, a projection map of the defect on the feature contour map TouchProfile is generated. Specifically, the feature contour map WF is overlaid with the unit contour image UF so that a tool-based projection map is generated on the feature contour map WF by determining whether each defect feature p(x,y) in the defect map Defect_Profile can be projected into the contour map TouchProfile. When a defect feature p1 can be projected into, for example, the contour map TouchProfile_2, a tool-based projection map is generated on the feature contour map WF2, which includes the mirror point defect feature p1′ of defect feature p1. Thus, a dataset of mirror points of each defect feature p in the defect feature Defect_Profile within a corresponding contact contour of TouchProfile_1 to TouchProfile_4 can be established, denoted here as {Defect_Mirror_1}, {Defect_Mirror_2}, {Defect_Mirror_3}, and {Defect_Mirror_4}. It should be noted that defect features in the Defect_Profile that cannot be projected onto any TouchProfile contour map are discarded. For example... Figure 6 As shown in (B), a projection of some defect features in Defect_Profile onto TouchProfile_1 is shown.

[0040] In step 509, the cell image contour (UF) containing the defect feature `Defect_Profile` is simultaneously presented graphically, along with the feature contours (WF) of each operation tool containing the mirror point datasets {Defect_Mirror_1}, {Defect_Mirror_2}, {Defect_Mirror_3}, and {Defect_Mirror_4}. This allows for a visual overview of the defect feature distribution on a single cell. For example, using... Figure 6 For example, in the Defect_Profile, a specific densely distributed region SHA clearly exhibits a particular defect, which is highly likely a systemic defect caused by the operating tool. The user can select this region SHA, and as shown in Figure (B), a similar region appears within the contact contour of CT1. This suggests that the operating tool causing the defect in region SHA is tool CT1 at node Node_1, and a similar distribution of SHA is synchronously highlighted on the feature contour WF1 of CT1. Accordingly, the operator can use this information to inspect tool CT1 located at node Node_1.1。

[0041] According to another embodiment of the present invention, for a defect feature set Defect_Profile of 200 units in a batch {UNIT1}, the number of defects in the generated mirror point datasets {Defect_Mirror_1}, {Defect_Mirror_2}, {Defect_Mirror_3}, and {Defect_Mirror_4} may be different. For example, in the Defect_Profile dataset, 70% of the defect features fall within the range of contact contour TouchProfile_2, 80% fall within the range of contact contour TouchProfile_1, 20% fall within the range of contact contour TouchProfile_3, and 10% fall within the range of contact contour TouchProfile_4. It should be noted that since the same defect features may fall within different contact contours, the sum of the proportions of data falling within each contact contour may be greater than 1, as shown in this example. According to one implementation of the present invention, suspected operating tools are ranked according to the proportion of defect features falling into the contact contour. For example, in this case, tool CT1 is the most suspected, while tool CT4 is the least suspected. Therefore, an indication signal can be output to indicate the suspected operating tools that may cause unit defects for the operator's reference. According to another aspect of the present invention, data falling into different contact contours can be further analyzed. For example, based on the production history traceability information HisInfor, it can be determined that the unit defect feature data in {Defect_Mirror_2} within the range of contact contour TouchProfile_2 all come from the same tool CT. 2-1 That is, 70% of the defect features in the Defect_Profile dataset come from the first of multiple CT2 tools at node Node_2. 2-1 This indicates that these defects were all caused by the same CT tool. 2-1 Caused by, therefore, the tool CT 2-1 It is highly likely to be a source of defects. Furthermore, based on the production history traceability information HisInfor, the data within {Defect_Mirror_1} falling within the range of the contact profile TouchProfile_1 originates from multiple tools, such as CT. 1-1 CT 1-2 CT 1-3 CT 1-4 That is, 80% of the defect features in the Defect_Profile dataset come from multiple tool CTs at node_1. 1-1 CT 1-2 CT1-3 CT 1-4 Since the likelihood of multiple tools malfunctioning simultaneously is extremely low in practice, the implementation of this invention eliminates the possibility of defects caused by the tool at node_1, and only outputs prompts indicating that defects may be caused by the tool at node_2. It even suggests that the operator should focus on checking the first tool CT at node_2. 2-1 Since only 20% of the defect features fall within the contact profile TouchProfile_3 and 10% fall within the contact profile TouchProfile_4, these defects can be ignored. Of course, these defects can also be further analyzed using the analysis method described above to provide further suggestions.

[0042] According to another embodiment of the invention, in addition to visually representing defects in batch units and their potential sources of operational tool defects in graphical form, the invention can also automatically identify and provide suspected sources of defective tools to the user. Figure 7 As shown, in step 701, the defect clustering pattern DCP formed by each defect feature p in the defect map Defect_Profile is identified. Here, the defect pattern DCP can be determined based on conditions such as cluster distance and density, thereby forming one or more clustering patterns DCP. In this example, it is assumed that there are two clustering patterns DCP1 and DCP2 in the defect map Defect_Profile. Various clustering algorithms known in the prior art can be used to generate the defect clustering pattern DCP.

[0043] In step 703, a clustering algorithm is used to identify the projected clustering patterns in the mirror point datasets {Defect_Mirror_1}, {Defect_Mirror_2}, {Defect_Mirror_3}, and {Defect_Mirror_4}, denoted as PCP. In this example, it is assumed that there is one clustering pattern PCP1 in the mirror point dataset {Defect_Mirror_1}; two clustering patterns PCP2 and PCP3 in the dataset {Defect_Mirror_2}; two clustering patterns PCP4 and PCP5 in {Defect_Mirror_3}; and one clustering pattern PCP6 in {Defect_Mirror_4}.

[0044] In step 705, the matching degree between the defect clustering patterns (DCP1, DCP2) and the projected clustering patterns (PCP1 to PCP6) is determined. Various pattern recognition algorithms known in the prior art can be used for matching. Then, the matched pattern pairs are sorted, for example, based on the density of mirror data points within the pattern and the matching degree.

[0045] In step 707, the tool associated with the highest-ranking projected clustering pattern (PCP1 to PCP6) is identified as the fault source tool causing the defect. For example, assuming that clustering pattern PCP3 has the highest matching degree with defect pattern DCP2 and contains the largest number of data points, it can be determined that the operating tool CT2 caused the defect indicated by defect clustering pattern DCP2. Therefore, fault indication information can be automatically output to the user to prompt the user to re-inspect tool CT2 at node Node_2. The fault indication information here may include CT2, which is suspected to be the fault source, and the defect location indicated by the matched defect clustering pattern DCP2.

[0046] The fault source identification method of the present invention can be executed by on-site monitoring equipment on the production line, or by back-end monitoring equipment. Figure 8A An example of such a monitoring device is shown. As shown, the monitoring device includes an image acquisition module 801, a controller 802, and a database 803.

[0047] The image acquisition module 801 acquires image data of the unit U processed by the operating tools at each node on the production line. This image data can be data that has already undergone defect identification. As an example, the acquired defect data can contain only one type of defect for ease of processing; in another example, the acquired defect data can contain multiple types of defects. As mentioned earlier, the defect data includes defect feature p information. This image data information can be stored in any format, such as a txt or csv file. In another example, the acquired image data can also be the raw image without defect identification, for example, directly connected to a camera on the production line via wired or wireless means to acquire the surface image U_Img_SF of the unit U that has completed all corresponding operation processing at the final node online. In this case, the image acquisition module 801 can use known image processing and computer vision algorithms, such as OpenCV, to process the surface image U_Img_SF, automatically identifying whether a defect appears on the surface SF1 of the corresponding unit U, and providing the defect feature p of the corresponding defect.

[0048] The controller 802 is configured to process the defect features of unit U to identify suspected defect source operating tools. First, based on the unit U's identifier U_ID read by the image acquisition module 801 and the pre-stored production history traceability information HisInfor, the controller 802 determines all tools CTs operating on the current surface SF1 of the current unit U, for example, operating tools CT1-CT4 in this example. Then, based on the operating tool's identifier, it reads the feature profile data WF1-WF4 of each operating tool CT from the database 803. Each feature profile WF1-WF4 includes the contact profiles TouchProfile1-TouchProfile4 between each operating tool and unit U when operating unit U; that is, the area contour on the surface SF1 of unit U where the operating tool contacts the operating tool when operating unit U. Therefore, the controller 802 can identify the corresponding operating tool causing the defect from multiple operating tools based on the defect feature p and the feature profile data WF1-WF4. Specifically, according to one example, controller 802 determines the projection of each defect feature p(x,y) of unit U onto contact profiles TouchProfile_1 to TouchProfile_4. As mentioned earlier, each defect feature p(x,y) can be the defect position coordinates in a coordinate system of a plane defined by planar dimensions (X,Y), and profile features WF have the same planar dimensions X,Y as the unit. Therefore, the projection here includes determining the position of each defect feature p(x,y) within the profile features WF. If it is determined that the projection of at least a portion of the defect features p of unit U falls within a certain contact profile, for example, within the range of contact profile TouchProfile_1, it indicates that the defect on unit U was caused by the operating tool CT1 corresponding to contact profile TouchProfile_1, and therefore an indication signal can be output indicating that the operating tool CT1 caused the defect on unit U1. Alternatively, in another example, for the currently selected unit U1, all operating tools CT1-CT4 associated with unit U1 can be graphically displayed, and the operating tool that caused the defect on U1, such as CT1, is displayed by highlighting, for example, or annotating CT1.

[0049] In another example of the present invention, as shown in FIG8, the controller 802 may include a defect selection module 8021, a projection module 8022, and an identification module 8023. The defect selection module 8021 is used to receive a user's selection of one or more defect features of a unit U from a plurality of defect features on a graphical user interface, referred to here as p. m×n This represents the m×n defect features selected by the user. The projection module 8022 projects the selected m×n defect features p m×nThe contact profiles TouchProfile_1 to TouchProfile_4 of the operating tools CT1-CT4 are projected respectively. If the identification module 8023 determines the defect feature p... m×n (Or a portion thereof, such as more than 80%) can be projected onto one of the contact profiles TouchProfile_1 to TouchProfile_4. For example, if it can be projected onto the range of TouchProfile_3, then the operating tool CT3 is determined to be the defective feature p. m×n The defect source operation tools are as follows. Similar to the previous discussion, when the defect features of the unit are projected onto different contact contours at different proportions, the suspected operation tools are sorted according to the proportion of defect features falling into the contact contour, and the sorted suspected operation tools are output for the operator's reference.

[0050] In another example of the invention, the monitoring device can also perform defect diagnosis on batches of units. For this purpose, the controller 802 establishes a defect feature point dataset {P} based on image data of multiple units, wherein these multiple units originate from the same production line and have undergone processing by the same operating tools CT1-CT4. Furthermore, as... Figure 8B As shown, the controller 802 may include a pattern recognition module 8024 and an output module 8025. The pattern recognition module 8024 is configured to establish multiple mirror point datasets within the contact contours TouchProfile_1 to TouchProfile_4 of multiple operating tools by projecting a defect feature point dataset {P}. Subsequently, the pattern recognition module 8024 identifies at least one defect clustering pattern (DCP) in the defect feature point dataset {P} and identifies a projection clustering pattern (PCP) in each mirror point dataset, thereby obtaining multiple projection clustering patterns (PCP). Here, conditions known in the prior art, such as clustering distance and density, can be used to determine the clustering patterns DCP and PCP. Further, the pattern recognition module 8024 determines the matching degree between each clustering pattern DCP and the multiple clustering patterns (PCP) of the multiple mirror point datasets.

[0051] The output module 8025 is configured to sort the matched pattern pairs, for example, based on the matching degree or the data density of mirror points within the pattern, sorting multiple cluster patterns PCPs; and automatically identifying the operating tool CT associated with the cluster pattern PCP with the highest matching degree as the operating source causing the defect. For example, if the third pattern PCP3 among multiple cluster patterns PCPs has the highest matching degree and the largest data density with DCP, then the tool CT2 corresponding to the contact contour, such as TouchProfile_2, where the third pattern PCP3 is located is determined to be a suspected defect source. Therefore, the monitoring device can output fault indication information to the user to prompt the user to re-inspect the tool CT2. The fault indication information may include the suspected fault source CT2 and the defect location of unit U indicated by the defect cluster pattern DCP corresponding to the matched projection cluster pattern PCP3. Using the identification scheme of the present invention, not only are all units considered for defect source identification, but different defect categories can also be automatically identified with higher accuracy, thus improving identification efficiency and avoiding losses caused by manual omissions.

[0052] According to another embodiment of the present invention, the pattern recognition module 8024 determines the defect ratio in the generated mirror point dataset, that is, the proportion of data in each mirror point dataset to the defect feature point dataset {P}, and analyzes the mirror point dataset. As shown in the previous example, it is assumed that 70% of the defect features in the dataset {P} fall within the range of contact contour TouchProfile_2, 80% of the defect features fall within the range of contact contour TouchProfile_1, 20% of the defect features fall within the range of contact contour TouchProfile_3, and 10% of the defect features fall within the range of contact contour TouchProfile_4. The pattern recognition module 8024 sorts the suspected operating tools according to the proportion of defect features falling within the contact contours and outputs an indication signal indicating each suspected operating tool that may cause unit defects. According to another aspect of the present invention, the pattern recognition module 8024 determines, based on the production history traceability information HisInfor, that the data falling within the range of contact contour TouchProfile_2 basically comes from the same tool CT at the same node Node_2. 2-1 That is, 70% of the defect features in the dataset {P} come from the first of the multiple CT2 tools at node Node_2. 2-1 This indicates that these defects were all caused by the same CT tool. 2-1 Caused by, therefore, the tool CT 2-1It is highly likely to be a source of defect. Furthermore, based on the production history traceability information HisInfor, the data falling within the contact profile TouchProfile_1 was determined to originate from multiple tools, such as CT. 1-1 CT 1-2 CT 1-3 CT 1-4 That is, 80% of the defect features in the dataset {P} come from multiple tool CTs at node Node_1. 1-1 CT 1-2 CT 1-3 CT 1-4 Since the probability of multiple tools malfunctioning simultaneously is extremely low in practice, according to the implementation of this invention, the possibility of defects caused by the tool at node_1 is eliminated, and only the prompt information that the defects may be caused by the tool at node_2 is output. It even suggests that the operator should focus on checking the first tool CT at node_2. 2-1 .

[0053] In another example of the invention, the controller 802 may also provide a selection interface (not shown in the figure) that allows the user to set defect identification conditions, such as specifying that defect source identification is performed only for tool CT4. Thus, the image acquisition module 801 can input only each unit U processed by all CT4 tools at node Node_4, and the controller directly extracts the feature contour map WF4 of tool CT4 from the database 803, and determines whether tool CT4 has caused a defect according to the aforementioned method flow. That is, by determining whether the defects on each unit can be projected into the contact contour of CT4, it can be intuitively determined whether CT4 has caused a defect. Furthermore, the user can also set the batch of units to be detected, or different surfaces of unit U (e.g., top surface SF or bottom surface BF), or possible defects in different parts of the unit (e.g., cores C1 and C2, or solder balls or substrate, etc.) through the selection interface. In addition, the above example illustrates the acquisition of an image of unit surface SF1 at the last node; however, in another implementation, the invention can also directly receive pre-processed image data that has already been marked with defects, thus avoiding the need for image processing for defect identification.

[0054] While exemplary defect diagnosis methods and monitoring devices of the present invention have been described above with reference to specific examples, those skilled in the art will also recognize that the various illustrative logical modules and method steps described in conjunction with the disclosure herein can be implemented as electronic hardware, computer software, or a combination of both. For example, the controller 802 according to the present invention can be implemented as a processor or main controller and a memory, wherein the memory stores modules in the form of computer programs, and the processor implements the modules and methods of the present invention by executing these modules. The memory storing the computer programs can be located either inside or outside the monitoring device and accessed by the monitoring device. Furthermore, another embodiment of the present invention provides a machine-readable medium storing machine-readable instructions that, when executed by a processor, cause the processor to perform any of the methods disclosed herein.

[0055] The present invention has been shown and described in detail above with reference to the accompanying drawings and preferred embodiments. However, the present invention is not limited to these disclosed embodiments. Based on the above embodiments, those skilled in the art will know that more embodiments of the present invention can be obtained by combining the code review methods in the different embodiments. These embodiments are also within the protection scope of the present invention.

Claims

1. A defect source identification method, comprising: Acquire image data of at least one object, wherein the image data includes defect features of at least one defect located on the surface of the object; Based on the object's identifier, multiple feature contour data of multiple operating tools for operating on the object's surface are obtained, wherein each feature contour data includes contact contour data, the contact contour representing the region contour on the object's surface that comes into contact with the operating tool when the operating tool operates on the object; Based on the defect features and the feature contour data, the corresponding operating tool that causes one or more defects in the at least one defect is identified from the plurality of operating tools.

2. The method of claim 1, wherein the contact includes direct contact or indirect interaction between the operating tool and the object surface.

3. The method of claim 1, wherein obtaining the plurality of feature contour data includes extracting, from a database, multiple feature contour data of the plurality of operating tools pre-stored based on the object's identifier and historical data, wherein the plurality of operating tools are used to perform different processes on the object on a production line and have different contact contours.

4. The method of claim 3, wherein the contact profile is determined in the following manner: Obtain the design parameters for each operating tool; Obtain the design parameters of the object; The contact contour data is determined based on the design parameters of the operating tool and the object.

5. The method of any one of claims 1-3, wherein, The image data and the feature contour data include the planar dimensions of the object on the plane where the surface is located, and the defect feature includes the position coordinates of the defect in the coordinate system of the plane defined by the planar dimensions; The determination of the corresponding operating tool causing the defect includes: identifying the corresponding operating tool by determining whether the defect feature can be projected into the contact contour of the plurality of tools.

6. The method of claim 5, wherein, The at least one defect includes multiple defects having multiple defect characteristics. The corresponding operating tools that cause the defect include: Select one or more of the aforementioned defect features, When it is determined that one or more defect features can be projected into the contact contour of at least the first operating tool among the plurality of operating tools, the first operating tool is determined to be the corresponding operating tool.

7. The method of claim 3, wherein the at least one object comprises a plurality of objects, the method further comprising: A defect feature point dataset is established based on the image data of the multiple objects; The defect feature point dataset is established by projection, which consists of multiple mirror point datasets within the contact contours of the multiple operating tools. Determine the ratio of the amount of data in each mirror point dataset to the amount of data in the defect feature point dataset; Based on the historical data, analyze whether the data in each mirror point dataset comes from multiple different operating tools of the same type; The operation tools corresponding to the mirror point datasets are sorted according to the stated proportion, wherein the sorted operation tools exclude mirror point datasets from multiple different operation tools of the same type.

8. The method of claim 5, wherein the at least one object comprises a plurality of objects, the method further comprising: A defect feature point dataset is established based on the image data of the multiple objects; The defect feature point dataset is established by projection, which consists of multiple mirror point datasets within the contact contours of the multiple operating tools. Identify at least one first clustering pattern in the defect feature point dataset; Identify at least one second clustering pattern for each mirror point dataset; Determine the matching degree between the first clustering pattern and multiple second clustering patterns of the multiple mirror point datasets; Based on the matching degree, the plurality of second clustering patterns are sorted; The operation tool associated with the second clustering pattern with the highest matching degree is automatically identified as the operation source causing the defect.

9. The method of any one of claims 1-4, wherein acquiring image data of multiple objects comprises: Obtain the original image of the surface of the object; The original image is processed to extract defect features of the at least one defect contained therein.

10. A device for defect source identification, comprising: An image acquisition module is configured to acquire image data of at least one object, wherein the image data includes defect features of at least one defect located on the surface of the object; The controller is configured as follows: Acquire multiple feature contour data of multiple operating tools for operating on the surface of the object, wherein each feature contour data includes contact contour data, the contact contour representing the contour of the area on the surface of the object that comes into contact with the operating tool when the operating tool operates on the object; Based on the defect features and the feature contour data, the corresponding operating tool that causes one or more defects in the at least one defect is identified from the plurality of operating tools.

11. The device of claim 10, wherein the contact includes direct contact or indirect interaction between the operating tool and the surface of the object.

12. The device of claim 10, wherein the controller is further configured to: Based on the object's identification and historical data, multiple feature contour data of the multiple operating tools that are pre-stored in the database are extracted, wherein the multiple operating tools are used to perform different processes on the object on the production line and have different contact contours.

13. The device as claimed in any one of claims 10-12, wherein, The image data and the feature contour data include the planar dimensions of the object on the plane where the surface is located, and the defect feature includes the position coordinates of the defect in the coordinate system of the plane defined by the planar dimensions; The controller is further configured to identify the corresponding operating tool by determining whether the defect feature can be projected into the contact contour of the plurality of tools.

14. The device of claim 13, wherein, The at least one defect includes multiple defects having multiple defect characteristics. The controller is further configured as follows: The defect selection module is configured to receive the selection of one or more defect features from the plurality of defect features; The projection module is configured to project one or more selected defect features onto the contact contours of the plurality of operating tools; The identification module determines that the first operating tool is the corresponding operating tool when it determines that one or more defect features can be projected into the contact contour of at least the first operating tool among the plurality of operating tools.

15. The device of claim 12, wherein the at least one object comprises a plurality of objects, and the controller establishes a defect feature point dataset based on image data of the plurality of objects; The controller further includes: The pattern recognition module is configured as follows: The defect feature point dataset is established by projection, which consists of multiple mirror point datasets within the contact contours of the multiple operating tools. Determine the ratio of the amount of data in each mirror point dataset to the amount of data in the defect feature point dataset; Based on the historical data, analyze whether the data in each mirror point dataset comes from multiple different operating tools of the same type; The output module is configured to sort the operation tools corresponding to the mirror point dataset according to the stated ratio, wherein the sorted operation tools exclude mirror point datasets from multiple different operation tools of the same type.

16. The device of claim 13, wherein the at least one object comprises a plurality of objects, and the controller establishes a defect feature point dataset based on image data of the plurality of objects; The controller further includes: The pattern recognition module is configured as follows: The defect feature point dataset is established by projection, which consists of multiple mirror point datasets within the contact contours of the multiple operating tools. Identify at least one first clustering pattern in the defect feature point dataset; Identify at least one second clustering pattern for each mirror point dataset; Determine the matching degree between the first clustering pattern and multiple second clustering patterns of the multiple mirror point datasets; The output module is configured to sort the plurality of second clustering patterns based on the matching degree; Furthermore, it automatically identifies the operational tool associated with the second clustering pattern with the highest matching degree as the operational source causing the defect.

17. The apparatus of any one of claims 10-12, wherein the image acquisition module is further configured to: acquire an original image of the surface of the object; and process the original image to extract defect features of the at least one defect contained therein.

18. A device for defect source identification, comprising: Memory that stores machine-readable programs; A processor configured to execute the program to implement the method of any one of claims 1-9.

19. A machine-readable storage medium having a machine-readable program stored thereon, wherein the program, when executed by a processor, causes the processor to perform the method of any one of claims 1-9.

Images