A method and system for identifying ring defects

Abstract

The present application relates to the technical field of semiconductor, in particular to a kind of annular defect identification method and system, including the following steps: using defect classification model to carry out defect classification to defect pattern and obtain at least one pattern set;Actual production information is obtained, including the maintenance level of machine table and the pattern quantity of pattern set;Recommended pattern processing mode is selected according to actual production information, and pattern processing mode includes: overlay, fitting;Wherein, when actual production information meets first and second processing conditions, then recommended pattern processing mode is overlay and fitting respectively;Recommended pattern processing mode is selected to carry out pattern processing to pattern set, to obtain new defect pattern;The fault type that causes annular defect is judged according to new defect pattern.The present application proposes a kind of scheme of pattern processing based on machine table, product data fusion guidance, to improve defect positioning efficiency while ensuring the reliability of positioning result.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of semiconductor technology, in particular to a ring defect identification method and system. BACKGROUND

[0002] In the manufacturing process of semiconductor devices, each process flow may cause some unintended structures on the wafer, among which the ones that cause the circuit on the chip to malfunction are called wafer defects. In order to ensure wafer yield and production capacity, the maintenance of wafer processing equipment is crucial. The existing wafer equipment maintenance mainly has two difficulties:

[0003] 1. The maintenance of equipment failure has hysteresis

[0004] The existing equipment maintenance scheme is usually as follows: a defect detection model is used to detect defects on the wafer, and then the defect detection result is used to deduce whether the machine exists a fault. However, this scheme of deducing whether the machine exists a fault by using defects has two problems: first, there is serious hysteresis in fault discovery, and second, it often needs to build a judgment mechanism for different types of machines or processes to monitor whether the corresponding machine is in a normal state, which is very difficult and costly to monitor.

[0005] For example, CN202311213205.3 discloses an intelligent monitoring and early warning system for wafer production. The early warning system needs to use edge detection algorithm, double-branch defect detection model, equipment failure prediction model and other multiple processes to find the problem of possible increase in dust. Then deduce the fault of the machine from the phenomenon of dust. However, single problem analysis is very limited in the types of faults and machines that can be monitored, and it is difficult to find traces on the wafer at the beginning of the problem.

[0006] For example, CN202310893076.0 discloses a FDC traceability analysis method based on distributed parallel computing and storage medium. However, this method often only analyzes the failure cause by using testing, diagnosis and other operations after the fault problem is highlighted.

[0007] In fact, different types of machines have large and complex data in the processing process, which contains parameters of different properties, so the data acquisition, analysis and supervision are very difficult.

[0008] 2. Difficulty in automatic identification or traceability of defects

[0009] In order to find the hidden problems of the machine at an earlier stage, the early maintenance of the machine is also crucial. In fact, in the actual production process, the step of directly locating the specific machine equipment by using defects often needs manual analysis by on-site engineers.

[0010] Currently, there are also attempts to compare defect data with equipment parameters to conduct automatic positioning. For example, CN201310119893.7 discloses a method and system for automatically detecting mechanical scratches. The method matches the defect data and the generating equipment to locate the equipment that produces the defect. For another example, CN202111547662.7 discloses a method for detecting scratches of a mechanical arm. However, due to the complexity of the wafer processing flow and the concealment of early defects (such as it may be difficult to clearly appear on the wafer photo), the automatic detection method is also difficult to locate the accurate machine at the first time.

[0011] Therefore, there is an urgent need for a method that can monitor and warn in a timely manner the wafer quality and machine maintenance at the earliest stage of failure. SUMMARY

[0012] The purpose of the present application is to provide a ring defect recognition method and system, which partially solves or alleviates the above-mentioned deficiencies in the prior art, and can monitor and warn in a timely manner the wafer quality and machine maintenance at the earliest stage of failure.

[0013] In order to solve the above-mentioned technical problems, the present application specifically adopts the following technical solutions:

[0014] In a first aspect, the present application provides a ring defect recognition method, comprising the steps of:

[0015] S201 acquiring a plurality of first defect patterns of a plurality of sample wafers, and using a defect classification model to classify the plurality of first defect patterns to obtain at least one first pattern set, and the first pattern set includes at least one defect pattern with a ring defect;

[0016] S202 acquiring actual production information, and the actual production information includes: the maintenance level of at least one machine experienced by the sample wafer in the processing process acquired by the machine maintenance monitoring system, and the number of patterns of the first pattern set;

[0017] S203 selecting a recommended pattern processing mode according to the actual production information, the pattern processing mode including: superimposition and fitting; wherein, when the actual production information meets a first processing condition, the recommended pattern processing mode is superimposition; and / or, when the actual production information meets a second processing condition, the recommended pattern processing mode is fitting; wherein,

[0018] The first processing condition refers to that the maintenance level is less than or equal to a set maintenance level, and the number of patterns is greater than or equal to a first set number; the second processing condition refers to that the maintenance level is greater than the set maintenance level, and the number of patterns is less than the first set number;

[0019] S204 performing graphic processing on the first set of patterns according to the recommended graphic processing mode to obtain a second set of defect patterns;

[0020] S205 determining a fault type causing the ring-shaped defect according to the second set of defect patterns.

[0021] In some embodiments, the fault type includes fault machine information; and S205 includes the following steps:

[0022] The shape data of the defect is obtained from the second set of defect patterns, and the shape data includes one or more of the following: defect length, defect curvature, defect radius, and distance between the defect and the wafer center;

[0023] A preset machine feature library associated with the defect pattern is obtained; the machine feature library includes a first set of quantified features, and the first set of quantified features includes one or more of the following defect feature data: mechanical arm rotation radius, mechanical arm rotation speed, distance between the mechanical arm and the wafer center, and adsorption foot position of the adsorption device, and the defect feature data is respectively associated with corresponding wafer machine information;

[0024] The shape data is matched with the machine feature library, and when the shape data matches the corresponding defect feature data, the machine corresponding to the current shape data is identified as the corresponding fault machine.

[0025] In some embodiments, the machine feature library further includes a pattern feature set, and the pattern feature set includes a defect feature pattern corresponding to a defect caused by at least one machine or a device component in the machine.

[0026] In some embodiments, S201 further includes the following step: marking defect points in the defect patterns in the first set of patterns, wherein the defect points are points identified as ring-shaped defects by the defect classification model.

[0027] In some embodiments, before S203, the following steps are further included:

[0028] A plurality of defect sample groups are selected from the marked defect points, and each defect sample group includes one or more defect points.

[0029] The distance between the plurality of defect sample groups and the wafer center is calculated; and when the difference between the distance between the defect sample group and the wafer center and a standard distance is greater than a first difference threshold, the corresponding defect sample group is identified as an error group.

[0030] When the number of error groups is greater than an error threshold, the corresponding first defect pattern is excluded from the first set of patterns to obtain a new first set of patterns.

[0031] In some embodiments, S203 comprises:

[0032] When the actual production information does not meet the first processing condition or the second processing condition, a first prompt signal is sent to the user;

[0033] and a first selection signal sent by the user is received, the first selection signal comprising: graphic processing mode information, and a recommended graphic processing mode is determined according to the first selection signal.

[0034] Another aspect of the present application provides a ring defect identification system, comprising:

[0035] A ring defect classification module is configured to obtain a plurality of first defect patterns of a plurality of sample wafers, and perform defect classification on the plurality of first defect patterns using a defect classification model to obtain at least one first pattern set, and the first pattern set comprises at least one defect pattern with a ring defect;

[0036] An actual production information acquisition module is configured to obtain actual production information, and the actual production information comprises:

[0037] The maintenance level of at least one machine experienced by the sample wafer obtained through the machine maintenance monitoring system during the processing process, and the number of patterns of the first pattern set;

[0038] A processing mode selection module is configured to select a recommended graphic processing mode according to the actual production information, the graphic processing mode comprising: superposition, fitting; wherein, when the actual production information meets the first processing condition, the recommended graphic processing mode is superposition; and / or, when the actual production information meets the second processing condition, the recommended graphic processing mode is fitting; wherein,

[0039] The first processing condition refers to that the maintenance level is less than or equal to a set maintenance level, and the number of patterns is greater than or equal to a first set number; and the second processing condition refers to that the maintenance level is greater than the set maintenance level, and the number of patterns is less than the first set number;

[0040] A graphic processing module is configured to perform graphic processing on the first pattern set using the recommended graphic processing mode to obtain a second defect pattern;

[0041] A fault judgment module is configured to determine the fault type causing the ring defect according to the second defect pattern.

[0042] In some embodiments, the fault type comprises: fault machine information; and the fault judgment module comprises:

[0043] a shape data acquisition unit configured to acquire shape data of the defects through the second defect pattern, the shape data including one or more of defect length, defect curvature, defect radius, and spacing between the defect and the wafer center;

[0044] a feature library acquisition unit configured to acquire a preset machine feature library associated with the defect pattern, the machine feature library including a quantitative feature set including one or more of defect feature data of mechanical arm rotation radius, mechanical arm rotation speed, spacing between the mechanical arm and the wafer center, and adsorption foot position of an adsorption device, and the defect feature data being associated with corresponding wafer machine information respectively;

[0045] a first matching unit configured to match the defect feature data set with the machine feature library, and when the defect feature data matches the corresponding defect feature data, the machine corresponding to the shape data is identified as the corresponding failure machine.

[0046] In some embodiments, the machine feature library further includes a pattern feature set including a defect feature pattern corresponding to defects caused by at least one machine or a device component in the machine.

[0047] In some embodiments, the ring defect classification module is further configured to mark defect points in the defect pattern in the first pattern set, wherein the defect points are points identified as ring defects by the defect classification model.

[0048] In this embodiment, the operation of the machine (such as the risk level of the machine) and the actual defect condition (such as the number of defect patterns classified) are introduced to automatically select the processing mode of the defect pattern. Then, in the face of different actual production scenes, the identification or analysis mode of the defect can be flexibly adjusted on the basis of ensuring the reliability of defect positioning. BRIEF DESCRIPTION OF DRAWINGS

[0049] In order to more clearly illustrate the technical solutions of the embodiments of the present application or the prior art, the following will briefly introduce the drawings needed to be used in the embodiment or prior art description. In all the drawings, similar elements or parts are generally identified by similar reference signs. In the drawings, each element or part is not necessarily drawn according to the actual scale. Obviously, the drawings described below are some embodiments of the present application, and those skilled in the art can obtain other drawings according to these drawings without creative labor.

[0050] Figure 1 a flowchart of the scheduling method in an exemplary embodiment of the present application;

[0051] Figure 2System block diagram of the scheduling method in an exemplary embodiment of the present application;

[0052] Figure 3 Flowchart of the ring defect recognition method in an exemplary embodiment of the present application;

[0053] Figure 4 Flowchart of the straight line defect recognition method in an exemplary embodiment of the present application;

[0054] Figure 5 Sample diagram of a ring defect;

[0055] Figure 6 Sample diagram of another ring defect;

[0056] Figure 7 Sample diagram of a ray defect;

[0057] Figure 8 Sample diagram of a scratch defect;

[0058] Figure 9 Sample diagram of another scratch defect. DETAILED DESCRIPTION

[0059] In order to make the objects, technical solutions and advantages of the embodiments of the present application clearer, the technical solutions in the embodiments of the present application will be described clearly and completely below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are some but not all of the embodiments of the present application. Based on the embodiments in the present application, all other embodiments obtained by a person of ordinary skill in the art without creative work fall within the scope of protection of the present application.

[0060] In this document, the suffixes such as "module", "part" or "unit" used for an element are merely intended for facilitating the description of the present application, and have no particular meaning by themselves. Therefore, "module", "part" or "unit" can be mixedly used.

[0061] In this document, the terms "upper", "lower", "inner", "outer", "front", "back", "one end", "the other end" and the like indicate the orientation or positional relationship based on the orientation or positional relationship shown in the drawings, and are merely for the convenience of describing the present application and simplifying the description, and do not indicate or imply that the device or element referred to must have a particular orientation, be constructed and operated in a particular orientation, and therefore cannot be understood as a limitation on the present application. In addition, the terms "first", "second" are only for the purpose of description, and cannot be understood as indicating or implying relative importance.

[0062] In this document, unless otherwise indicated and / or unless the context clearly dictates otherwise, the terms "mounting", "provided with", "connected" and the like are to be construed broadly, for example, "connected" can be fixed connection, can be detachable connection, or integrally connected; can be mechanical connection, can be direct connection, can be indirect connection through intermediate medium, can be internal connection of two elements. For those skilled in the art, the specific meaning of the above terms in the present application can be understood according to the specific circumstances.

[0063] In this document, "and / or" includes any and all combinations of one or more of the associated items.

[0064] In this document, "a plurality of" means two or more, that is, it includes two, three, four, five, etc.

[0065] It should be noted that in this document, the terms "including", "containing" or any other variants thereof are intended to cover non-exclusive inclusion, so that the process, method, article or device including a series of elements not only includes those elements, but also includes other elements not explicitly listed, or includes elements inherent to such process, method, article or device. Without more limitations, the element defined by the sentence "including a…" does not exclude the presence of other identical elements in the process, method, article or device including the element.

[0066] As used in this specification, the term "about", typically means + / - 5% of the stated value, more typically + / - 4% of the stated value, more typically + / - 3% of the stated value, more typically + / - 2% of the stated value, even more typically + / - 1% of the stated value, even more typically + / - 0.5% of the stated value.

[0067] In this specification, certain embodiments can be disclosed in a format that is a range. It is to be understood that such a "range" format is used only for convenience and brevity, and should be considered equivalently to a list of individual values that are explicitly written out. For example, a range of "from 1 to 5" should be considered to have been explicitly disclosed as the individual values 1, 2, 3, 4, and 5, and as the sub-ranges 1-2, 2-3, 3-4, 4-5, and 1-5. A range of "from 1 to 10" should be considered to have been explicitly disclosed as the individual values 1, 2, 3, 4, 5, 6, 7, 8, 9, and 10, as well as the sub-ranges 1-2, 2-3, 3-4, 4-5, 5-6, 6-7, 7-8, 8-9, 9-10, and 1-10. No matter how broad the range is, the above rules apply.

[0068] In this document, the machine (or processing machine) is also referred to as equipment (or processing equipment), which can refer to any one or more processing modules or processing devices on the production processing line of a wafer.

[0069] In this article, "failure" can also be referred to as machine failure or equipment failure, which refers to the equipment or process problem that may affect the processing, detection and other processes of the wafer, and then affect the yield or production capacity of the wafer. For example, the failure can be a mechanical failure, such as a scratch defect on the wafer caused by loosening of the mechanical arm of the machine; for another example, the failure can also be a process failure, such as insufficient wafer production capacity caused by problems in the processing formula setting of the machine.

[0070] In this article, one failure can refer to a specific failure problem (such as loosening of a screw of a certain machine), or can also refer to the same type of failure (such as problems in the process formula of the associated multiple machines).

[0071] In this article, a defect point refers to a point on the wafer that has an abnormality and may cause a circuit failure problem (such as a pixel abnormal point on the wafer pattern).

[0072] In this article, a ring-shaped defect refers to a defect formed by the aggregation of multiple defect points in a ring (for example, a circular ring) or approximately ring shape on the wafer, such as Figure 5 、 Figure 6 as shown.

[0073] In this article, a straight line defect (or straight line-shaped defect) refers to a defect formed by the aggregation of multiple defect points in a straight line on the wafer. The straight line defect includes one or more of the following types: linear scratch, straight line-shaped particle defect, and radial particle defect. Among them, the linear scratch can be a scratch (such as Figure 8 、 Figure 9 as shown) or multiple scratches; the straight line-shaped particle defect is usually a particle defect aggregated in a straight line shape; and the radial particle defect is usually a particle defect formed by divergence in multiple directions (such as Figure 7 as shown, which is a plurality of particle-shaped defects diverging in a radial shape on the wafer).

[0074] In a large wafer processing plant, it often involves the simultaneous operation of multiple wafer processing lines. And due to the large number of types and quantities of machines, it is very difficult to monitor the running state of the machines, so it is very likely to be missed in the early stage of failure (for example, the lack of experience of the on-site engineer misjudges the failure as normal operation), which leads to the expansion of the impact of the failure and seriously affects the yield of the wafer.

[0075] Currently, the machine failure is usually deduced in reverse through the defect pattern, but in the earliest stage when the failure may occur or has occurred, the actual defect on the wafer may not be obvious, such as Figure 8 、 Figure 9As shown, the defect is actually incomplete. At this time, if the work experience of the staff is also relatively less, it is very likely that the defect will be missed, so that the adverse effects of the failure gradually expand. Alternatively, for the incomplete defect pattern, it is also difficult for the automated classification model to perform positioning analysis.

[0076] In order to be able to reduce the impact of the failure on the operation of the production line as much as possible, the present application proposes a scheme that can maintain or investigate possible hidden dangers of the failure before the wafer defects are generated, or at an early stage when the number of defects is relatively small.

[0077] Embodiment one

[0078] As Figure 1 As shown, the present application provides a scheduling method and system for early maintenance of the machine. The present application first provides a scheduling method that can maintain the machine as early as possible based on the combination of dynamic and static production line data, which includes:

[0079] S101 determines the static maintenance level of at least one machine according to the static database; wherein the static database includes: the set maintenance time of a plurality of machines; and the static maintenance level of the machine can be divided into level one, level two and level three according to the interval between the current time and the set maintenance time;

[0080] S102 determines the first dynamic maintenance level of the machine according to the first dynamic database; the first dynamic database includes one or more of the following working data: voltage working data, current working data, and corresponding standard data; wherein the working data is used to represent the change relationship between the working value of the machine and the working time; correspondingly, S102 includes:

[0081] obtaining at least one working data of the machine and standard data corresponding to the working data;

[0082] When the difference between the working value (such as voltage value or current value) of a data point in the working data and the standard data is greater than the preset difference threshold (i.e. the second difference threshold), the data point is regarded as an abnormal point;

[0083] determining the first dynamic maintenance level according to the abnormal point; wherein when the number of abnormal points belongs to the first set threshold, the first dynamic maintenance level is level one; when the number of abnormal points belongs to the second set threshold, the first dynamic maintenance level is level two; when the number of abnormal points belongs to the third set threshold, the first dynamic maintenance level is level three;

[0084] S103 determining a comprehensive maintenance level of the corresponding machine according to the static maintenance level and the first dynamic maintenance level; wherein, when the static maintenance level is level one and the first dynamic maintenance level is level one, the comprehensive maintenance level is level one; when the static maintenance level is level two and the first dynamic maintenance level is level two, the comprehensive maintenance level is level two; when the static maintenance level is level three and the first dynamic maintenance level is level three, the comprehensive maintenance level is level three.

[0085] S104 sequencing the maintenance order of the machine according to the comprehensive maintenance level. At this time, the scheduling scheme sequenced according to the comprehensive maintenance level can be obtained.

[0086] In some embodiments, when the static maintenance level is level one and the first dynamic maintenance level is level two,

[0087] the comprehensive maintenance level is level two;

[0088] when the static maintenance level is level one and the first dynamic maintenance level is level three, the comprehensive maintenance level is level three;

[0089] when the static maintenance level is level two and the first dynamic maintenance level is level one, the comprehensive maintenance level is level one;

[0090] when the static maintenance level is level two and the first dynamic maintenance level is level three, the comprehensive maintenance level is level three;

[0091] when the static maintenance level is level three and the first dynamic maintenance level is level one, the comprehensive maintenance level is level two;

[0092] when the static maintenance level is level three and the first dynamic maintenance level is level two, the comprehensive maintenance level is level three.

[0093] For example, in some embodiments, the static database includes: the set maintenance time (or recommended maintenance time) of a plurality of machines set by the user in advance; and the time classification rule corresponding to each machine in the static maintenance level classification.

[0094] For example, for different types of machines, the set maintenance time of the next machine can be defined in combination with the service life and regular maintenance cycle of the machine. And according to the interval between the current time and the set maintenance time, the static maintenance level of the machine can be at least divided into level one, level two and level three (equivalent to dividing the static operation risk of the machine into low risk, medium risk and high risk).

[0095] For example, in some embodiments, when the distance between the current time and the set maintenance time is greater than or equal to 20 days, the static maintenance level of the machine tool is level one; when the distance between the current time and the set maintenance time is less than 20 days and greater than or equal to 10 days, the static maintenance level of the machine tool is level two; and when the distance between the current time and the set maintenance time is less than 10 days, the static maintenance level of the machine tool is level three.

[0096] In some embodiments, the set maintenance time can be the safe use period (or average service life) of the machine tool or the equipment component of the machine tool.

[0097] For example, in some embodiments, the safe use period of the mechanical arm A in the machine tool is one year, the static maintenance level of the mechanical arm A in the first 1-6 months can be set to level one, the static maintenance level of the mechanical arm A in the 7th-9th month can be set to level two, and the static maintenance level of the mechanical arm A in the 10th-12th month can be set to level three. It can be understood that the specific time grading rules can be freely adjusted by the user.

[0098] In some embodiments, the voltage working data used to determine the first dynamic maintenance level can be an SPC control chart of voltage (for example, the SPC Chart is used to monitor the voltage). When it is monitored that the difference between the voltage value (equivalent to the working value) of a data point in the SPC control chart and the control point (equivalent to the standard data) is large, it is considered that the data point is abnormal. Moreover, according to the number of abnormal data points, the first dynamic maintenance level can be divided into level one, level two and level three (equivalent to low risk, medium risk and high risk in turn).

[0099] Alternatively, in other embodiments, the SPC control chart of current can also be used to implement the grading processing of the first dynamic maintenance level.

[0100] In some embodiments, the standard data can be data selected or input by the user in combination with the actual production situation.

[0101] In some embodiments, before S103, the method further comprises the following step:

[0102] S104 determining a second dynamic maintenance level of the machine tool according to a second dynamic database, wherein the second dynamic database comprises product data of the wafer, and S104 comprises: selecting the product data associated with the machine tool from the second dynamic database as dynamic adjustment data, and dividing the second dynamic maintenance level into level one, level two and level three according to the quality of the wafer according to the dynamic adjustment data;

[0103] S105 verifying or adjusting the first dynamic maintenance level according to the second dynamic maintenance level; wherein,

[0104] When the second dynamic maintenance level is less than or equal to the first dynamic maintenance level, the original level of the second dynamic maintenance level is maintained; when the second dynamic maintenance level is greater than the first dynamic maintenance level, the first dynamic maintenance level is adjusted according to the second dynamic maintenance level; wherein when the second dynamic maintenance level is level two or level three, and the first dynamic maintenance level is level one, the first dynamic maintenance level is adjusted to level two; when the second dynamic maintenance level is level three, and the first dynamic maintenance level is level two, the first dynamic maintenance level is adjusted to level three.

[0105] In some embodiments, the second dynamic maintenance level can be classified according to the number of defects of the wafer. For example, inputting the wafer into a defect classification model (for example, an ADC classification model) can preliminarily identify the number or type of defects that may exist on the wafer. Then, according to the number of defects that may exist, the second dynamic maintenance level is classified into level one, level two and level three in order from small to large.

[0106] Alternatively, in other embodiments, the yield information of the wafer can be obtained from the product data, and the risk of the second dynamic maintenance level is directly obtained according to the yield data.

[0107] In some embodiments, the product data includes one or more of the following: ring defects, scratch defects, radial particle defects, and linear particle defects; and the machines in the static database are labeled with corresponding machine labels; the machine labels include: a first machine label, and the machine corresponding to the first machine label will perform a rotation operation in the wafer processing process; a second machine label, and the machine corresponding to the second machine label will perform a clamping operation in the wafer processing process; a third machine label, and the first machine label will perform a blowing operation in the wafer processing process; wherein for machines labeled with different labels, different defect data (or product data) have different preset recommended priorities;

[0108] Correspondingly, in S104, the dynamic adjustment data corresponding to the machine label is selected according to the recommended priority of the product data.

[0109] For example, in some embodiments, for the first type of machine labeled with the first machine label, the ring defect is associated as the first recommended priority, and the scratch defect, the radial particle defect, and the linear particle defect are associated as the second recommended priority. That is, at this time, for the first type of machine, the product data of the ring defect is preferred as the dynamic adjustment data.

[0110] For example, in some embodiments, for the second type of machine marked with the second machine label, the scratch defect is associated as the first recommended priority, and the ring defect, the radial particle defect, the linear particle defect are associated as the second recommended priority. That is, at this time, for the second type of machine, the product data of the scratch defect is preferably selected as the dynamic adjustment data.

[0111] For example, in some embodiments, for the third type of machine marked with the third machine label, the radial particle defect, the linear particle defect are associated as the first recommended priority, and the scratch defect, the ring defect are associated as the second recommended priority. That is, at this time, for the third type of machine, the product data of the radial particle defect, the linear particle defect is preferably selected as the dynamic adjustment data.

[0112] For example, in some embodiments, when two or more labels are associated with a machine, multiple product data can also be selected as dynamic adjustment data at the same time.

[0113] For example, in some embodiments, the machine corresponding to the first machine label has one or more of the following: a rotating machine, a polishing adhesive film machine, etc.

[0114] For example, in some embodiments, the machine corresponding to the second machine label has one or more of the following: a wafer dicing machine.

[0115] For example, in some embodiments, the machine corresponding to the third machine label has one or more of the following: an oven machine, a blowing machine, etc.

[0116] In the present application, factors related to the operation of the machine are selected from the static data and dynamic data, and a multi-level evaluation scheme is used to integrate the limited data. Further, the multi-level integration of static and dynamic data is used to evaluate the operation of the machine at multiple levels, so that timely maintenance and repair of the machine can be performed at the early stage of failure, or even before the failure is truly reflected on the wafer pattern. Moreover, through this rapid evaluation mechanism, the present application can improve the maintenance efficiency by advancing the maintenance time without the need for additional personnel configuration, and the limited personnel configuration can play a greater maintenance effect.

[0117] In order to improve the linkage between static data and dynamic data, the present application also synchronously defines the machine type and the defect type, and mainly selects the linear defect (which can include scratch defect or particle defect) and the ring defect as the key identification object to assist in improving the reliability and accuracy of the dynamic maintenance level evaluation of the machine.

[0118] For example, in some embodiments, taking the annular defect as an example, when it is monitored that the wafer passing through a certain machine 1 (which is associated with a first machine tag) forms an annular defect on the defect, and the annular defect is identified as being generated by the machine 1, then the second dynamic maintenance level of the machine is identified as level three. When the annular defect on the wafer does not match the machine 1, then the second dynamic maintenance level of the machine 1 can be identified as level two; otherwise, the second dynamic maintenance level of the machine can be identified as level one.

[0119] Correspondingly, the present application also provides a scheduling system (or a machine maintenance monitoring system) for early maintenance of a machine, as shown in Figure 2 The scheduling system comprises:

[0120] A static rating module 10 configured to determine a static maintenance level of at least one machine according to a static database; wherein the static database comprises: set maintenance time of a plurality of machines; and the static maintenance level of the machine is divided into level one, level two and level three according to the interval between the current time and the set maintenance time;

[0121] A first dynamic rating module 11 configured to determine a first dynamic maintenance level of the machine according to a first dynamic database; the first dynamic database comprises one or more of the following working data: voltage working data, current working data, and corresponding standard data; wherein the working data is used to represent the change relationship of the working value of the machine with the working time; correspondingly, the first dynamic rating module comprises:

[0122] A data acquisition unit 111 configured to acquire at least one working data of the machine and standard data corresponding to the working data;

[0123] An abnormality monitoring unit 112 configured to regard a data point as an abnormal point when it is monitored that the difference between the working value of the data point and the standard data is greater than a preset difference threshold value;

[0124] A dynamic rating unit 113 configured to determine the first dynamic maintenance level according to the abnormal point; wherein when the number of abnormal points belongs to a first set threshold value, the first dynamic maintenance level is level one; when the number of abnormal points belongs to a second set threshold value, the first dynamic maintenance level is level two; when the number of abnormal points belongs to a third set threshold value, the first dynamic maintenance level is level three;

[0125] The comprehensive rating module 13 is configured to determine a comprehensive maintenance rating of the machine according to the static maintenance rating and the first dynamic maintenance rating; when the static maintenance rating is level one and the first dynamic maintenance rating is level one, the comprehensive maintenance rating is level one; when the static maintenance rating is level two and the first dynamic maintenance rating is level two, the comprehensive maintenance rating is level two; when the static maintenance rating is level three and the first dynamic maintenance rating is level three, the comprehensive maintenance rating is level three.

[0126] The maintenance sequencing module 14 is configured to sequence the maintenance of the machine according to the comprehensive maintenance rating.

[0127] In some embodiments, when the static maintenance rating is level one and the first dynamic maintenance rating is level two, the comprehensive maintenance rating is level two; when the static maintenance rating is level one and the first dynamic maintenance rating is level three, the comprehensive maintenance rating is level three; when the static maintenance rating is level two and the first dynamic maintenance rating is level one, the comprehensive maintenance rating is level one; when the static maintenance rating is level two and the first dynamic maintenance rating is level three, the comprehensive maintenance rating is level three; when the static maintenance rating is level three and the first dynamic maintenance rating is level one, the comprehensive maintenance rating is level two; when the static maintenance rating is level three and the first dynamic maintenance rating is level two, the comprehensive maintenance rating is level three.

[0128] In some embodiments, the system further comprises:

[0129] The second dynamic rating module 15 is configured to determine a second dynamic maintenance rating of the machine according to a second dynamic database; the second dynamic database comprises product data of wafers, and the second dynamic rating module is further configured to select the product data associated with the machine from the second dynamic database as dynamic adjustment data, and divide the second dynamic maintenance rating into level one, level two and level three according to the quality of the wafers according to the dynamic adjustment data;

[0130] The first dynamic rating correction module 16 is configured to verify or adjust the first dynamic maintenance rating according to the second dynamic maintenance rating; when the second dynamic maintenance rating is less than or equal to the first dynamic maintenance rating, the original level of the second dynamic maintenance rating is maintained; when the second dynamic maintenance rating is greater than the first dynamic maintenance rating, the first dynamic maintenance rating is adjusted according to the second dynamic maintenance rating; when the second dynamic maintenance rating is level two or level three, and the first dynamic maintenance rating is level one, the first dynamic maintenance rating is adjusted to level two; when the second dynamic maintenance rating is level three, and the first dynamic maintenance rating is level two, the first dynamic maintenance rating is adjusted to level three.

[0131] In some embodiments, the product data includes one or more of: a ring defect, a straight line defect, and a ray defect; and the machine in the static database is marked with a corresponding machine label; the machine label includes: a first machine label, and the machine corresponding to the first machine label will perform a rotation operation in the wafer processing process; a second machine label, and the machine corresponding to the second machine label will perform a clamping operation in the wafer processing process; a third machine label, and the first machine label will perform a blowing operation in the wafer processing process.

[0132] Correspondingly, the second dynamic rating module is also configured to select dynamic adjustment data corresponding to the machine label according to the recommended priority of the product data.

[0133] In some embodiments, for the first type of machine, the first dynamic database further includes: rotation speed working data, and standard data corresponding to the rotation speed working data.

[0134] The exemplary identification schemes of the straight line defect and the ring defect will be described below:

[0135] Embodiment two

[0136] In the early stage of development of machine failure, the morphology of the failure problem on the wafer (or defect pattern) may not be obvious, at which time if the engineer has less experience, the wafer with failure may be ignored, resulting in hidden dangers being excluded. Moreover, due to the incompleteness of the defect pattern, it is also difficult to obtain accurate results when using a defect identification or comparison model for analysis. Correspondingly, the present application provides a rapid defect identification method to assist in evaluating the dynamic maintenance rating of the machine, as shown in Figure 3 The following takes the ring defect as an example, including the following steps:

[0137] S201, a plurality of first defect patterns of a plurality of sample wafers are obtained, and a defect classification model is used to classify defects of the plurality of first defect patterns to obtain at least one first pattern set, and the first pattern set includes at least one defect pattern in which a ring-shaped defect exists; at this time, the first pattern set is preliminarily identified as existing the same ring-shaped defect;

[0138] The defect classification model can be obtained by training a plurality of defect pattern sets of a plurality of different types of ring-shaped defects by a neural network modeling method. For example, the defect classification model (preferably a ring-shaped defect classification model) can be an ADC classification model.

[0139] S202, actual production information is obtained, and the actual production information includes: a maintenance level (equivalent to a risk level) of at least one machine experienced by the sample wafer in the processing process obtained by a machine maintenance monitoring system, and / or a number of patterns in the first pattern set;

[0140] In some embodiments, the machine maintenance monitoring system can be the above-mentioned scheduling system, which can automatically classify the maintenance level of the historical processing machine of the wafer.

[0141] For example, in some embodiments, the "maintenance level" can refer to the above-mentioned comprehensive maintenance level determined according to the static maintenance level and the first dynamic maintenance level, or can refer to the above-mentioned first dynamic maintenance level.

[0142] Alternatively, in other embodiments, the machine maintenance monitoring system can also be automatically input by the user. For example, when the user thinks that the machine has a fault, the first input signal can be directly input to the machine maintenance monitoring system; wherein the first input signal includes: machine information, and the maintenance level of the machine (for example, when the user thinks that the machine has a fault, the maintenance level can be directly adjusted to level three); the machine maintenance monitoring system determines the maintenance level of the machine in response to the input first input signal.

[0143] For another example, when the user temporarily does not find that the machine has a fault, the second input signal can be directly input to the machine, and the maintenance level in the second input signal is level two or level one.

[0144] S203, a recommended pattern processing mode is selected according to the actual production information, and the pattern processing mode includes: overlay, fitting; wherein when the actual production information meets the first processing condition, the recommended pattern processing mode is overlay; and / or when the actual production information meets the second processing condition, the recommended pattern processing mode is fitting; wherein,

[0145] The first processing condition refers to that the maintenance level is less than or equal to a set maintenance level, and the number of patterns is greater than or equal to a first set number; and the second processing condition refers to that the maintenance level is greater than the set maintenance level, and the number of patterns is less than the first set number.

[0146] S204 selects a recommended pattern processing mode to perform pattern processing on the first pattern set to obtain a second defect pattern;

[0147] S205 determines a fault type causing the ring defect according to the second defect pattern.

[0148] The fault type can refer to a machine type causing the fault.

[0149] In this document, "overlay" can also be referred to as "pattern fusion", which refers to superimposing at least two patterns (or markers on the patterns) to form a new pattern. The overlay processing can use one or more of the following methods: pixel-level superposition, region-level superposition, mixed mode superposition, and mixed channel superposition method.

[0150] In this document, "fitting" refers to using data enhancement methods (e.g., pattern inversion, rotation, magnification, cropping, etc.) to fit and calculate complete defect geometry (e.g., complete ring defect) using partial defect geometry (e.g., incomplete defect point).

[0151] In some embodiments, the fault type includes fault machine information; and S205 includes the following steps:

[0152] The shape data of the defect is obtained from the second defect pattern, and the shape data includes one or more of the following: defect length, defect curvature, defect radius, and distance between the defect and the wafer center;

[0153] A first machine feature library associated with the defect pattern is obtained; the first machine feature library includes a first set of quantified features, and the first set of quantified features includes one or more of the following defect feature data: mechanical arm rotation radius, mechanical arm rotation speed, distance between the mechanical arm and the wafer center, and adsorption foot position of the adsorption device, and the defect feature data is associated with corresponding wafer machine information;

[0154] The shape data is matched with the machine feature library, and when the shape data matches the corresponding defect feature data, the machine corresponding to the shape data is identified as the corresponding fault machine.

[0155] In some embodiments, the machine feature library further includes a first pattern feature set, and the pattern feature set includes a defect feature pattern corresponding to a defect caused by at least one machine or a device component in the machine.

[0156] In some embodiments, when no machine is matched, a second prompt signal can also be sent to the user for manual intervention.

[0157] It can be understood that an engineer can construct a machine feature library in advance in combination with engineering experience, wherein the quantitative features in the machine feature library can be obtained from the specification parameters of the machine, and the defect feature patterns can be obtained from historical production data, and typical defect feature patterns that the machine is likely to generate are bound to the machine to facilitate positioning of machine failure.

[0158] In some embodiments, S201 further includes the step of marking the defect points in the defect patterns in the first pattern set, wherein the defect points are points identified as ring-shaped defects by the defect classification model. In this embodiment, the marked defect points are equivalent to the marking elements.

[0159] In some embodiments, the first processing condition refers to the number of patterns being greater than or equal to a first set number, and the second processing condition refers to the number of patterns being less than the first set number.

[0160] Alternatively, in other embodiments, the first processing condition refers to the maintenance level (equivalent to the risk level) being less than or equal to a set maintenance level (for example, when the maintenance level is level one); and the second processing condition refers to the maintenance level being greater than the set maintenance level (for example, when the maintenance level is level two or level three).

[0161] In this embodiment, the operation of the machine (such as the risk level of the machine) and the actual defect situation (such as the number of classified defect patterns) are introduced to automatically select the processing method of the defect pattern. In this way, the recognition or analysis method of the defect can be flexibly adjusted in the face of different actual production scenarios.

[0162] In addition, in the early stage of the existence of hidden dangers, the number of defects is relatively limited, and the defects are relatively large. The probability of misjudgment of both manual and automatic matching models is high. The present application proposes a pattern processing scheme based on machine and product data fusion guidance, which can improve the efficiency of defect positioning to a certain extent, and also ensures the reliability of the defect positioning result.

[0163] In some embodiments, when the classification accuracy of the defect classification model is relatively low, before S203, the method further includes the step of:

[0164] a plurality of defect sample groups are selected from the marked defect points, and each defect sample group includes one or more defect points;

[0165] calculate the distance between the plurality of defect sample groups and the wafer center; and when the difference between the distance between the defect sample group and the wafer center and the standard distance is greater than a first difference threshold, the corresponding defect sample group is identified as an error group;

[0166] The standard distance can refer to the average distance value of the plurality of defect sample groups.

[0167] Alternatively, each defect sample group can calculate a plurality of distances, and when the number of defect sample groups belonging to the same distance is the largest, the corresponding distance is taken as the standard distance.

[0168] When the number of error groups is greater than an error threshold, the corresponding first defect pattern is excluded from the first pattern set to obtain a new first pattern set.

[0169] In some embodiments, S203 comprises:

[0170] When the actual production information does not meet the first processing condition or the second processing condition, a first prompt signal is sent to the user;

[0171] and receiving a first selection signal sent by the user, the first selection signal comprising: pattern processing mode information, and determining the recommended pattern processing mode according to the first selection signal.

[0172] In this embodiment, a semi-automatic identification mode of human-computer cooperation is provided, which can not only perform automatic pattern processing selection and pattern processing through a fast evaluation mode (such as the setting of the first and second processing conditions), but also can sharply capture key problems and introduce manual intervention. Thus, the defect pattern can be quickly and accurately processed by using the human-computer cooperation mode.

[0173] In this embodiment, the defect types are preliminarily classified first, and the matching between the defects and the machine is quickly performed by using the semi-automatic mode to assist the user to quickly determine the risk degree of the existence of faults or hidden dangers of the machine. Furthermore, the actual wafer product data can also be quickly fed back to the machine maintenance stage to quickly maintain and evaluate the running state of the machine.

[0174] For example, in some embodiments, it further comprises the steps of:

[0175] Obtaining the number of defects generated by a machine;

[0176] Determining the second dynamic maintenance level of the machine according to the number of defects (which can be used to preliminarily reflect the quality of the wafer);

[0177] When the number of defects of one machine belongs to the first defect threshold range, the second dynamic maintenance level of the machine is level one; when the number of defects of one machine belongs to the second defect threshold range, the second dynamic maintenance level of the machine is level two; and when the number of defects of one machine belongs to the third defect threshold range, the second dynamic maintenance level of the machine is level three.

[0178] In the embodiment, according to the rapid identification and positioning of the ring-shaped defects, in addition to assisting the user (such as an engineer) in preliminarily classifying the defects, a reference rating factor (such as the second dynamic maintenance level) can be provided to a scheduling system of machine maintenance (or a machine maintenance monitoring system), so as to realize comprehensive evaluation of the maintenance sequence of the machine from multiple maintenance factors such as the static machine maintenance period, the actual running state of the machine, and the final wafer product quality, so as to quickly troubleshoot the fault or hidden danger of the machine in the case of relatively limited selected rating factors.

[0179] It can be understood that one or more method steps in the embodiment can also be applicable to the identification of linear defects.

[0180] The embodiment also correspondingly provides an identification system of ring-shaped defects, comprising:

[0181] The ring-shaped defect classification module 21 is configured to obtain a plurality of first defect patterns of a plurality of sample wafers, and perform defect classification on the plurality of first defect patterns by using a defect classification model to obtain at least one first pattern set, and the first pattern set comprises at least one defect pattern in which a ring-shaped defect exists;

[0182] The actual production information acquisition module 22 is configured to obtain actual production information, and the actual production information comprises a maintenance level of at least one machine experienced by the sample wafer in the processing process obtained by a machine maintenance monitoring system, and a number of patterns of the first pattern set;

[0183] The processing mode selection module 23 is configured to select a recommended pattern processing mode according to the actual production information, and the pattern processing mode comprises superimposition and fitting; when the actual production information meets a first processing condition, the recommended pattern processing mode is superimposition; and / or when the actual production information meets a second processing condition, the recommended pattern processing mode is fitting; wherein,

[0184] The first processing condition refers to that the maintenance level is less than or equal to a set maintenance level, and the number of patterns is greater than or equal to a first set number; and the second processing condition refers to that the maintenance level is greater than the set maintenance level, and the number of patterns is less than the first set number;

[0185] The graphic processing module 24 is configured to perform graphic processing on the first defect pattern by using a recommended graphic processing mode to obtain a second defect pattern.

[0186] The fault judgment module 25 is configured to judge the fault type causing the annular defect according to the second defect pattern.

[0187] In some embodiments, the fault type includes fault machine information; and the fault judgment module 25 includes:

[0188] The shape data acquisition unit 251 is configured to acquire shape data of the defect through the second defect pattern, the shape data including one or more of the following: defect length, defect curvature, defect radius, and distance between the defect and the wafer center;

[0189] The feature library acquisition unit 252 is configured to acquire a preset machine feature library associated with the defect pattern; the machine feature library includes a quantitative feature set, the quantitative feature set including one or more of the following defect feature data: mechanical arm rotation radius, mechanical arm rotation speed, distance between the mechanical arm and the wafer center, and adsorption foot position of the adsorption device, and the defect feature data is respectively associated with corresponding wafer machine information;

[0190] The first matching unit 253 is configured to match the defect feature data set with the machine feature library, and when the defect feature data matches the corresponding defect feature data, the machine corresponding to the defect feature data is identified as the corresponding fault machine.

[0191] In some embodiments, the machine feature library further includes a pattern feature set, the pattern feature set including a defect feature pattern corresponding to a defect caused by at least one machine or a device component in the machine.

[0192] In some embodiments, the annular defect classification module is further configured to mark defect points in the defect pattern in the first pattern set, wherein the defect points are points identified as annular defects by the defect classification model.

[0193] In some embodiments, the system further includes a filtering module configured to select a plurality of defect sample groups from the marked defect points, and each defect sample group includes one or more defect points; calculate the distance between the plurality of defect sample groups and the wafer center; and when the difference between the distance between the defect sample group and the wafer center and the standard distance is greater than a first difference threshold, the corresponding defect sample group is identified as an error group; when the number of error groups is greater than an error threshold, the corresponding first defect pattern is excluded from the first pattern set to obtain a new first pattern set.

[0194] In some embodiments, the system further comprises a user adjustment module configured to send a first prompt signal to a user when the actual production information does not meet the first processing condition or the second processing condition, and receive a first selection signal sent by the user, the first selection signal comprising graphic processing mode information, and determining the recommended graphic processing mode according to the first selection signal.

[0195] In some embodiments, the system further comprises a second dynamic rating module, and the second dynamic rating module is further configured to obtain a defect quantity of defects generated by a machine; determine a second dynamic maintenance rating of the machine according to the defect quantity (which can be used to preliminarily reflect wafer quality); wherein when the defect quantity of a machine belongs to a first defect threshold range, the second dynamic maintenance rating of the machine is level one; when the defect quantity of a machine belongs to a second defect threshold range, the second dynamic maintenance rating of the machine is level two, and when the defect quantity of a machine belongs to a third defect threshold range, the second dynamic maintenance rating of the machine is level three.

[0196] Embodiment three

[0197] The present application also provides a rapid identification method of defects, which is described below by taking a straight-line defect as an example. As shown in FIG. 1, the method comprises the following steps: Figure 4

[0198] S301: obtaining a plurality of first defect patterns of a plurality of sample wafers, and performing defect classification on the plurality of first defect patterns by using a straight-line defect classification model to obtain at least one second pattern set, and the second pattern set comprises at least one defect pattern with a straight-line defect; wherein the straight-line defect comprises one or more types of linear scratches, straight-line particle defects, and ray-like particle defects;

[0199] In some embodiments, the straight-line defect classification model can select a plurality of defect pattern sets of different types of straight-line defects as samples, and is trained by using a neural network modeling method.

[0200] In some embodiments, the straight-line defect classification model can also directly select an existing ADC classification model.

[0201] S302: obtaining a pre-stored second feature database associated with the first defect pattern, wherein the feature database comprises a second quantitative feature set comprising a workpiece size of a device component in at least one machine, and / or a second pattern feature set comprising a defect feature pattern of a wafer defect caused by the device component of the at least one machine or the machine;

[0202] For example, in some embodiments, the workpiece size can be the width of a clamp.

[0203] ​For example, in some embodiments, the defect feature pattern can be a typical defect photo selected by the user from historical production data.

[0204] S303 matches the corresponding defect pattern with the second feature database, and when at least one machine is matched, determines the fault type causing the linear defect according to the machine.

[0205] The matching process includes: when the linear defect is a linear scratch, obtaining the defect features of the corresponding defect pattern, including: the distance between the scratch and the center, and / or the length of the scratch, and / or the distance between at least two scratches; and matching the defect features with the second set of quantitative features.

[0206] Alternatively, the matching process includes: when the linear defect is a linear particle defect or a radial particle defect, using an image recognition model (or a similarity comparison model) to compare the similarity between the defect pattern and the second set of feature patterns, and when the similarity between the defect feature pattern and the defect pattern is greater than a preset similarity, it is considered that the two are matched.

[0207] In some embodiments, when the defect pattern is successfully matched with multiple machines, an engineer can also be provided with a machine list according to the specific results of the similarity comparison, arranged from high to low according to the similarity.

[0208] In some embodiments, when at least two machines are matched, i.e., at least including: a first machine and a second machine, the method further includes the steps of:

[0209] Obtaining the positioning point of the wafer when entering the first machine, the positioning point being used to represent the processing angle of the wafer entering the machine;

[0210] Determining the formation direction of the linear defect according to the positioning point, the formation direction being the direction from the starting point to the ending point of the linear defect;

[0211] Obtaining a plurality of sample defect groups from the linear defect, and calculating the defect point density of each sample defect group; wherein a sample defect group includes one or more defect points.

[0212] When the defect density of the sample defect group gradually increases along the formation direction, it is preliminarily considered that the linear defect is caused by the first machine.

[0213] If not, it can be preliminarily determined that the linear defect is caused by the second machine or other machines.

[0214] Alternatively, the positioning point of the second machine can also be used to determine whether the current linear defect is related to the second machine.

[0215] For example, such as Figure 8 As shown, in some embodiments, when the wafer enters the machine 2, positioning points (such as notches set at the edge of the wafer) are marked on it, and a coordinate system is constructed according to the direction of the positioning points so that users (such as engineers) can determine whether the wafer pattern has been offset. For example, a Y-axis can be established with the direction of the notch pointing, and an X-axis can be established along the direction parallel to the pointing direction. At this time, if it is observed that the defect density (such as the distribution density of particles) of the wafer gradually decreases along the Y-axis direction, it can be considered that the current defect may be caused by the machine 2.

[0216] In some embodiments, when no machine is matched in the second graphics set, the following steps are included:

[0217] S304 Obtain actual production information, and the actual production information includes: the maintenance level of at least one machine that the sample wafer has experienced during the processing, and the number of patterns in the second pattern set, obtained through the machine maintenance monitoring system.

[0218] S305 selects a recommended graphics processing method based on the actual production information. The graphics processing method includes: overlay and fitting; wherein, when the actual production information meets a first processing condition, the recommended graphics processing method is overlay; and / or, when the actual production information meets a second processing condition, the recommended graphics processing method is fitting; wherein,

[0219] The first processing condition refers to the maintenance level being less than or equal to a set maintenance level, and the number of graphics being greater than or equal to a first set number; the second processing condition refers to the maintenance level being greater than a set maintenance level, and the number of graphics being less than a first set number.

[0220] S306 selects the recommended graphic processing method to process the second graphic set to obtain the third defect graphic.

[0221] In some embodiments, S301 further includes the step of: marking defect points in the defect graphics of the second graphics set, wherein the defect points are points identified as straight-line defects by the defect classification model.

[0222] In this embodiment, it is preferable to use static data and dynamic data (such as machine operation data or preliminary defect classification results) for comprehensive judgment, so as to use actual production information to guide the rapid identification of defects.

[0223] In some embodiments, a first prompt signal is issued to the user when the actual production information does not meet the first processing condition or the second processing condition;

[0224] receiving a first selection signal issued by a user, the first selection signal including: graphic processing mode information, and determining a recommended graphic processing mode according to the first selection signal.

[0225] In this embodiment, for the key stage (graphic processing stage) of linear defect recognition, a semi-automatic mode is used to select the graphic processing mode for semi-automatic graphic processing. This can not only quickly analyze and recognize the defect graphics based on actual production data, but also timely introduce manual intervention in special cases to ensure the efficiency and accuracy of defect recognition.

[0226] In this embodiment, a linear defect recognition system is also provided, which includes:

[0227] The linear defect classification module 31 is configured to obtain a plurality of first defect graphics of a plurality of sample wafers, and use a linear defect classification model to classify the defects of the plurality of first defect graphics to obtain at least one second graphic set, and the second graphic set includes at least one defect graphic with a linear defect; wherein the linear defect includes one or more of the following types: linear scratch, linear particle defect, and radial particle defect.

[0228] The second feature data acquisition module 32 is configured to obtain a pre-stored second feature database associated with the first defect graphics, wherein the feature database includes: a second quantitative feature set including a workpiece size of a device component in at least one machine; and / or a second graphic feature set including a defect feature graphic of a wafer defect caused by a device component of at least one machine or machine;

[0229] The linear defect matching module 33 is configured to match the corresponding defect graphics with the second feature database, and when at least one machine is matched, determine the fault type causing the linear defect according to the machine; wherein,

[0230] The matching process includes: when the linear defect is a linear scratch, obtaining a defect feature of the corresponding defect graphic, the defect feature including: the distance between the scratch and the center, and / or the length of the scratch, and / or the distance between at least two scratches; and matching the defect feature with the second quantitative feature set;

[0231] Alternatively, the matching process includes: when the linear defect is a linear particle defect or a radial particle defect; using an image recognition model to compare the similarity of the defect graphic and the second graphic feature set, and when the similarity between the defect feature graphic and the defect graphic is greater than a preset similarity, it is considered that the two are matched.

[0232] In some embodiments, the straight-line defect matching module 33 comprises:

[0233] a first positioning unit 331 configured to obtain a positioning point of the wafer when entering a first machine when at least two machines are matched, the positioning point being used to represent a processing angle of the wafer entering the machine, the at least two machines including the first machine and a second machine;

[0234] a second positioning unit 332 configured to determine a formation direction of the straight-line defect according to the positioning point, the formation direction referring to a direction from a start position to an end position of the straight-line defect;

[0235] a second matching unit 333 configured to obtain a plurality of sample defect groups from the straight-line defect and calculate defect point density of each sample defect group, wherein one sample defect group includes one or more defect points, and when the defect density of the sample defect group gradually increases along the formation direction, it is preliminarily considered that the straight-line defect is caused by the first machine.

[0236] In some embodiments, the system further comprises:

[0237] a production information obtaining module configured to obtain actual production information when no machine is matched in the second pattern set, and the actual production information includes a maintenance level of at least one machine experienced by a sample wafer in a processing process obtained by a machine maintenance monitoring system, and a number of patterns of the second pattern set;

[0238] a pattern processing mode selection obtaining module configured to select a recommended pattern processing mode according to the actual production information, the pattern processing mode including superimposition and fitting, wherein when the actual production information meets a first processing condition, the recommended pattern processing mode is superimposition, and / or when the actual production information meets a second processing condition, the recommended pattern processing mode is fitting, wherein

[0239] the first processing condition refers to that the maintenance level is less than or equal to a set maintenance level, and the number of patterns is greater than or equal to a first set number; and the second processing condition refers to that the maintenance level is greater than the set maintenance level, and the number of patterns is less than the first set number;

[0240] a pattern processing module configured to perform pattern processing on the second pattern set by using the recommended pattern processing mode to obtain a third defect pattern.

[0241] In some embodiments, the straight-line defect classification module labels defect points in the defect pattern in the second pattern set, wherein the defect point is a point recognized as a straight-line defect by the defect classification model.

[0242] The application also provides a computer device, comprising: a processor; and a memory for storing instructions executable by the processor; wherein the processor is configured to invoke the instructions stored by the memory to execute the method in any of the above embodiments.

[0243] The application also provides a computer program product configured to store computer-readable instructions which, when executed, cause a computer to perform the method in any of the above embodiments.

[0244] The application also provides a computer-readable storage medium having stored thereon computer program instructions which, when executed by a processor, implement the method in any of the above embodiments.

[0245] It should be noted that, in this document, the terms "comprise", "comprising", or any other variant thereof are intended to cover non-exclusive inclusions, such that processes, methods, articles, or apparatuses that comprise a list of elements are not limited to those elements, but can also include other elements not expressly listed, or inherent to such processes, methods, articles, or apparatuses. Without more limitations, an element defined by the phrase "comprising a" does not exclude the presence of additional identical elements in the process, method, article, or apparatus that includes the element.

[0246] From the above description of the embodiments, those skilled in the art can clearly understand that the above-mentioned embodiment methods can be realized by means of software and necessary general hardware platforms, and of course, they can also be realized by hardware, but in many cases, the former is a better embodiment. Based on this understanding, the technical solutions of the present application can be embodied in the form of a software product, which is stored in a storage medium (such as a ROM / RAM, a magnetic disk, or an optical disk) and includes a number of instructions for causing a computer terminal (which can be a mobile phone, a computer, a server, or a network device, etc.) to execute the methods described in the various embodiments of the present application.

[0247] The embodiments of the application are described above in conjunction with the accompanying drawings, but the application is not limited to the specific embodiments described above, which are merely illustrative and not restrictive. Those of ordinary skill in the art can make many modifications to the application under the guidance of the application without departing from the scope of the application and the protection scope of the claims, and all such modifications are within the scope of the application.

Claims

1. A method for identifying ring defects, characterized in that, Including the following steps: S201 acquires multiple first defect patterns from multiple sample wafers, and uses a defect classification model to classify the multiple first defect patterns to obtain at least one first pattern set, and the first pattern set includes: at least one defect pattern with a ring defect. S202 Obtain actual production information, and the actual production information includes: the maintenance level of at least one machine that the sample wafer has experienced during the processing, obtained through the machine maintenance monitoring system, and the number of patterns in the first pattern set; S203 selects a recommended graphic processing method based on the actual production information. The graphic processing method includes: overlay and fitting. Wherein, when the actual production information meets a first processing condition, the recommended graphic processing method is overlay; and / or, when the actual production information meets a second processing condition, the recommended graphic processing method is fitting. Wherein, the first processing condition refers to the maintenance level being less than or equal to a set maintenance level, and the number of graphics being greater than or equal to a first set number; the second processing condition refers to the maintenance level being greater than a set maintenance level, and the number of graphics being less than a first set number. S204 Selects the recommended graphic processing method to process the first graphic set to obtain the second defect graphic; S205 determines the fault type that caused the annular defect based on the second defect pattern.

2. The method for identifying ring defects according to claim 1, characterized in that, The fault type includes: faulty machine information; S205 includes the following steps: The shape data of the defect is obtained by the second defect pattern, and the shape data includes one or more of the following: defect length, defect curvature, defect radius, and distance between the defect and the center of the wafer; Obtain a preset machine tool feature library associated with the defect image; the machine tool feature library includes: a first quantitative feature set, the first quantitative feature set including one or more of the following defect feature data: robotic arm rotation radius, robotic arm rotation speed, distance between robotic arm and wafer center, and adsorption foot position of adsorption device, and the defect feature data are respectively associated with corresponding wafer machine tool information; The shape data is matched with the machine feature library, and when the shape data matches the corresponding defect feature data, the machine corresponding to the current shape data is identified as the corresponding faulty machine.

3. The method for identifying ring defects according to claim 2, characterized in that, The machine tool feature library also includes: a graphic feature set, which includes: defect feature graphics corresponding to defects caused by at least one machine tool or equipment components in the machine tool.

4. The method for identifying ring defects according to claim 1, characterized in that, S201 also includes the step of: marking defect points in the defect graphics in the first graphic set, wherein the defect points are points identified as ring defects by the defect classification model.

5. The method for identifying ring defects according to claim 4, characterized in that, Before S203, the following steps are also included: Multiple defect sample groups are selected from the marked defect points, and each defect sample group includes one or more defect points; the distance between the multiple defect sample groups and the center of the wafer is calculated; and when the difference between the distance between the defect sample group and the center and the standard distance is greater than a first difference threshold, the corresponding defect sample group is identified as an error group. When the number of error groups exceeds the error threshold, the corresponding first defect image is removed from the first image set to obtain a new first image set.

6. The method for identifying ring defects according to claim 1, characterized in that, S203 includes: When the actual production information does not meet the first processing condition or the second processing condition, a first prompt signal is issued to the user; and a first selection signal issued by the user is received, the first selection signal including: graphics processing method information, and a recommended graphics processing method is determined according to the first selection signal.

7. A system for identifying ring defects, characterized in that, include: The ring defect classification module is configured to acquire multiple first defect patterns from multiple sample wafers and classify the multiple first defect patterns using a defect classification model to obtain at least one first pattern set, and the first pattern set includes: at least one defect pattern with a ring defect. The actual production information acquisition module is configured to acquire actual production information, which includes: the maintenance level of at least one machine that the sample wafer has undergone during processing, as obtained through the machine maintenance monitoring system, and the number of patterns in the first pattern set. The processing method selection module is configured to select a recommended graphics processing method based on the actual production information. The graphics processing methods include: overlay and fitting; wherein, when the actual production information meets a first processing condition, the recommended graphics processing method is overlay; and / or, when the actual production information meets a second processing condition, the recommended graphics processing method is fitting; wherein... The first processing condition refers to the maintenance level being less than or equal to a set maintenance level, and the number of graphics being greater than or equal to a first set number; the second processing condition refers to the maintenance level being greater than a set maintenance level, and the number of graphics being less than a first set number. The graphics processing module is configured to perform graphics processing on the first graphics set using a recommended graphics processing method to obtain a second defect graphics. The fault determination module is configured to determine the fault type that caused the annular defect based on the second defect pattern.

8. The system according to claim 7, characterized in that, The fault types include: faulty machine information; the fault judgment module includes: The shape data acquisition unit is configured to acquire the shape data of a defect through a second defect pattern, the shape data including one or more of the following: defect length, defect curvature, defect radius, and the distance between the defect and the center of the wafer; The feature library acquisition unit is configured to acquire a preset machine feature library associated with the defect pattern; the machine feature library includes: a quantized feature set, which includes one or more of the following defect feature data: the rotation radius of the robotic arm, the rotation speed of the robotic arm, the distance between the robotic arm and the center of the wafer, and the position of the adsorption foot of the adsorption device, and the defect feature data are respectively associated with corresponding wafer machine information; The first matching unit is configured to match the defect feature dataset with the machine feature library, and when the defect feature data matches the corresponding defect feature data, the machine corresponding to the shape data is identified as the corresponding faulty machine.

9. The system according to claim 8, characterized in that, The machine tool feature library also includes: a graphic feature set, which includes: defect feature graphics corresponding to defects caused by at least one machine tool or equipment components in the machine tool.

10. The system according to any one of claims 7-9, characterized in that, The ring defect classification module is also configured to mark defect points in defect graphics in the first graphic set, wherein the defect points are points identified as ring defects by the defect classification model.

Images