A wafer manufacturing defect prediction method, device, equipment and readable medium

Abstract

The application provides a wafer manufacturing defect prediction method, device, equipment and readable medium. The method comprises the following steps: obtaining wafer map data and machine parameter data of a current manufacturing machine manufacturing wafer; inputting the wafer map data and the machine parameter data into a pre-trained time series model to obtain a probability value of the current manufacturing machine existing a manufacturing defect at a target time; and if the probability value meets a preset prompt condition, generating a wafer manufacturing defect warning information. According to the technical scheme, the trend of whether a certain type of manufacturing defect exists on the wafer surface is predicted and warned by using the time series model, the prediction accuracy of generating the wafer manufacturing defect is improved, the wafer manufacturing defect is avoided, and the performance and manufacturing efficiency of the wafer are ensured.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of semiconductor manufacturing, and in particular to a wafer manufacturing defect prediction method, device, equipment and readable medium. BACKGROUND

[0002] In recent years, with the rapid development of science and technology, the semiconductor manufacturing industry is also developing rapidly. As a basic material for making silicon semiconductor chips, the manufacturing technology of wafers has become one of the foundations and core competencies that restrict the development of semiconductors.

[0003] The manufacturing process of wafers is very complex, especially in the polishing, slicing and other manufacturing processes. The wafers will be in frequent contact with various processing equipment, such as rotating discs, wafer feeders, cleaning machines, developing machines and other equipment, which may cause mechanical damage such as friction, shear and pressure to the wafer surface. In severe cases, there may be serious manufacturing defects such as macroscopic scratches. When there are serious manufacturing defects on the wafer, it will have a significant impact on the quality and performance of the wafer, which may cause the downstream products of the wafer to malfunction or performance degradation.

[0004] At present, the occurrence of wafer manufacturing defects is usually prevented by strictly controlling the condition and operation specification of the processing equipment and ensuring the cleanliness of the production environment during wafer production. However, using manual mechanisms to judge and warn of manufacturing defects will waste a lot of manpower and time, and the judgment result is easily affected by subjective factors and prone to errors. Therefore, how to provide a technical solution that can timely and accurately warn of wafer manufacturing defects is a technical problem that needs to be solved by those skilled in the art. SUMMARY

[0005] An object of the present application is to provide a wafer manufacturing defect prediction method, device, equipment and readable medium, at least to solve the problems of low efficiency of manual warning and low accuracy of warning results. The purpose of the present application is to provide a new wafer manufacturing defect prediction method. The pre-trained time series model used in the method can find the characteristics, trends and development rules of wafer manufacturing defects from the time series of wafer map data and machine parameter data, thereby predicting the probability of wafer manufacturing defects at a target time, improving the prediction accuracy of wafer manufacturing defects, avoiding wafer manufacturing defects, and ensuring the performance and manufacturing efficiency of wafers.

[0006] To achieve the above object, some embodiments of the present application provide the following aspects:

[0007] In a first aspect, some embodiments of the present application further provide a wafer manufacturing defect prediction method, which comprises:

[0008] obtaining wafer map data and machine parameter data of a wafer manufactured by a current manufacturing machine;

[0009] inputting the wafer map data and the machine parameter data into a pre-trained time series model to obtain a probability value of the current manufacturing machine existing a manufacturing defect at a target time;

[0010] if the probability value meets a preset prompt condition, generating early warning information of the wafer manufacturing defect.

[0011] In a second aspect, some embodiments of the present application further provide a wafer manufacturing defect prediction device, the device comprising:

[0012] a data acquisition module configured to acquire wafer map data and machine parameter data of a wafer manufactured by a current manufacturing machine;

[0013] a defect probability determination module configured to input the wafer map data and the machine parameter data into a pre-trained time series model to obtain a probability value of the current manufacturing machine existing a manufacturing defect at a target time;

[0014] a defect early warning module configured to, if the probability value meets a preset prompt condition, generate early warning information of the wafer manufacturing defect.

[0015] In a third aspect, some embodiments of the present application further provide a computer device, the device comprising:

[0016] one or more processors; and

[0017] a memory storing computer program instructions which, when executed, cause the processor to perform the wafer manufacturing defect prediction method as described above.

[0018] In a fourth aspect, some embodiments of the present application further provide a computer readable medium having stored thereon computer program instructions executable by a processor to implement the wafer manufacturing defect prediction method as described above.

[0019] Compared with the prior art, in the scheme provided by the embodiments of the present application, wafer map data and machine parameter data of a wafer manufactured by a current manufacturing machine are acquired; the wafer map data and the machine parameter data are input into a pre-trained time series model to obtain a probability value of the current manufacturing machine existing a manufacturing defect at a target time; and if the probability value meets a preset prompt condition, early warning information of the wafer manufacturing defect is generated. By using the time series model to predict and warn the trend of whether a certain type of manufacturing defect exists on the wafer surface, the prediction accuracy of the wafer manufacturing defect is improved, the wafer manufacturing defect is avoided, and the performance and manufacturing efficiency of the wafer are ensured. BRIEF DESCRIPTION OF DRAWINGS

[0020] Figure 1is a flowchart of a wafer manufacturing defect prediction method provided by Embodiment One of the present application;

[0021] Figure 2 is a flowchart of a time series model training process provided by Embodiment Two of the present application;

[0022] Figure 3 is a sliding time window diagram provided by Embodiment Two of the present application;

[0023] Figure 4 is a structural diagram of a wafer manufacturing defect prediction device provided by Embodiment Three of the present application;

[0024] Figure 5 is a structural diagram of a computer device provided by Embodiment Five of the present application. DETAILED DESCRIPTION

[0025] To make the objectives, technical solutions, and superiorities of the embodiments of the present application clearer, the technical solutions of the embodiments of the present application will be described below in detail with the accompanying drawings of the embodiments of the present application. Obviously, the described embodiments are only some, but not all of the embodiments of the present application. Based on the embodiments of the present application, all other embodiments obtained by those of ordinary skill in the art without creative work fall within the scope of the present application.

[0026] The technical solutions provided by the embodiments of the present application will be described in detail below with the accompanying drawings and specific embodiments and application scenarios.

[0027] Embodiment One

[0028] Figure 1 is a flowchart of a wafer manufacturing defect prediction method provided by Embodiment One of the present application. As shown in Figure 1 , the method specifically includes the following steps: the flowchart includes:

[0029] Step S101, obtaining wafer map data and machine parameter data of a wafer manufactured by a current manufacturing machine.

[0030] The current manufacturing machine can be any processing or detection device used in the current manufacturing process of the wafer, such as a polishing machine, a photolithography machine, an ion implantation machine, a wire bonding machine, a wafer dicing machine, etc.

[0031] Wafer refers to a silicon wafer used for making silicon semiconductor circuits, and its raw material is silicon. High-purity polysilicon is dissolved and incorporated into a silicon crystal seed, then slowly pulled out to form a cylindrical single crystal silicon. After grinding, polishing and slicing, the silicon crystal rod forms a silicon wafer, that is, a wafer. The wafer production line on the market is mainly 8 inches and 12 inches. The main processing method of wafer is sheet processing and batch processing, that is, processing one or more wafers at the same time.

[0032] In the manufacturing process of the wafer, the machine parameter data of the manufacturing machine needs to be regulated and controlled to ensure the quality of each manufacturing process. Good machine parameter data is beneficial to control the manufacturing machine to produce excellent wafer products. Among them, the machine parameter data can be the temperature, pressure, current, oxygen concentration, manufacturing time and the like set by the current manufacturing machine.

[0033] In the present scheme, the wafer map data can be an image generated based on the electrical function test results of the wafer combined with the shape of the wafer. The wafer map data is in chip units, and the test results are marked on the position of each chip with different colors, shapes or codes. Through the spatial distribution of the wafer map data, the defect situation existing on the wafer surface at the current time can be determined.

[0034] Among them, the shape of the wafer can be obtained based on the image collector, which often takes pictures directly above the wafer.

[0035] Step S102, inputting the wafer map data and the machine parameter data into a pre-trained time series model to obtain a probability value of the current manufacturing machine existing manufacturing defects at a target time.

[0036] Among them, the manufacturing defects can be divided into point defects, line defects and area defects according to the generation mode of the wafer defects.

[0037] Point defects refer to local defect points existing on the surface of wafer chips, which are usually caused by raw material pollution or improper manufacturing process. Point defects are usually divided into the following categories:

[0038] 1. Cavernous defect: usually caused by bubbles or gas remaining in the wafer chip.

[0039] 2. Bubble defect: usually caused by uneven expansion of metal material or gas remaining in the wafer chip.

[0040] 3. Metal particles: metal particles are caused by sputtering or adhesion of metal materials during the manufacturing process.

[0041] Line defects refer to local defect lines existing on the surface of wafer chips, which are usually caused by manufacturing process or equipment failure. Line defects are usually divided into the following categories:

[0042] 1. Scratch: usually caused by wear and tear of equipment during manufacturing or impact during the process.

[0043] 2. Streaks: usually caused by improper mixing of ingredients during manufacturing or uneven distribution of materials within the wafer chip.

[0044] 3. Bumps or dents: usually caused by damage to equipment or tools during manufacturing.

[0045] Regional defects refer to localized defects present on the wafer chip, which are typically caused by manufacturing processes or equipment failures. Regional defects are generally classified into the following categories:

[0046] 1. Grain defects: if a certain grain on the wafer chip is damaged or shrinks, it can cause the performance of the wafer chip to decrease or fail.

[0047] 2. Inhomogeneous Vickers hardness: usually caused by uneven pressure, uneven materials, or equipment failure during wafer chip manufacturing.

[0048] 3. Thermal residual stress: temperature changes during wafer chip manufacturing can cause residual stress in the wafer chip. If these residual stresses are too large, they can cause the performance of the wafer chip to decrease or fail.

[0049] The above are common defect classifications and their related characteristics for wafers. During manufacturing and testing, these defects need to be effectively predicted to ensure the quality and performance of the wafers.

[0050] It can be understood that before the formation of macroscopic scratches and other manufacturing defects on the wafer surface, there will usually be intermittent points or line segments formed first. After the formation of multiple points or line segments, it is possible to form macroscopic scratches and other manufacturing defects. Therefore, in order to avoid the formation of macroscopic scratches and other manufacturing defects, it is necessary to observe and analyze the formation trend of wafer defects to determine whether there is a possibility of macroscopic scratches and other manufacturing defects at the target time.

[0051] In this scheme, a pre-trained time series model is used to analyze the trend of defects to obtain the probability value of the existence of manufacturing defects at the target time for the current manufacturing machine.

[0052] The time series model can be obtained by arranging the collected historical machine parameter data and wafer map data of the current manufacturing machine in time sequence, obtaining at least one time series of the machine parameter data and wafer map data, and training each time series. After inputting the wafer map data and machine parameter data of the wafer manufactured by the current manufacturing machine, the model can obtain the probability value of the current manufacturing machine having a manufacturing defect at the target time. Taking the input of typhoon data to predict the typhoon path as an example, after the model is trained, the probability of the typhoon at each position at the next moment can be output. For example, by inputting the basic data and historical path information of a typhoon, the probability of the position of the typhoon at the next moment is (33.0N / 137.3E) is 0.8, and the probability of the position of the typhoon at the next moment is (33.7N / 135.7E) is 0.2.

[0053] It should be noted that the probability value of the current manufacturing machine having a manufacturing defect at the target time can also be zero, that is, there is no manufacturing defect.

[0054] In step S103, if the probability value meets the preset prompt condition, the wafer manufacturing defect warning information is generated.

[0055] The preset prompt condition can be set according to the wafer type and size, or can be set according to the machine parameter data of the current manufacturing machine, or can be set according to historical experience, or can be set according to actual needs, and the embodiments of the present application are not limited thereto.

[0056] In this scheme, the preset prompt condition can be a preset probability value. For example, the preset prompt condition is set as the preset probability value is 0.8, and if the probability value of the current manufacturing machine having a manufacturing defect at the target time is greater than or equal to 0.8, the wafer manufacturing defect warning information is generated.

[0057] The warning information can be used to prompt the relevant staff that the wafer manufactured by the current manufacturing machine has a high risk of manufacturing defect at the target time, so that the relevant staff can timely process it. The warning information can include the name, number, position, and predicted target time of the manufacturing defect of the current manufacturing machine. The form of the warning information can be one or more of light, bell sound, voice, video, pop-up window, etc.

[0058] It can be understood that in the present scheme, a plurality of preset prompt conditions and corresponding warning information can also be set. For example, a first preset prompt condition with a preset probability value of 0.6 and a second preset prompt condition with a preset probability value of 0.8 can be set respectively; when the probability value meets the first preset prompt condition, the warning information can be generated and the warning yellow light can be flashed to prompt the relevant staff to adjust the machine parameter data of the current manufacturing machine to avoid generating more serious manufacturing defects; when the probability value meets the second preset prompt condition, the warning information can be generated and the warning red light can be flashed to prompt the relevant staff to immediately stop the operation of the current manufacturing machine.

[0059] The advantage of the present scheme is that the probability value of the current manufacturing machine existing manufacturing defects at the target time predicted based on the time series model and the preset prompt condition are used to generate the risk of wafer manufacturing defects to realize the automatic warning of manufacturing defects, improve the efficiency and accuracy of prevention, avoid wafer manufacturing defects, and ensure the performance and manufacturing efficiency of the wafer.

[0060] Embodiment Two

[0061] Figure 2 is a flowchart of the training process of the time series model provided by the present embodiment two. As shown in Figure 2 , the specific steps include the following steps:

[0062] Step S201, collecting wafer pattern samples of a current manufacturing machine at at least two time nodes, obtaining defect information from the wafer pattern samples, collecting real-time machine parameters of the current manufacturing machine, and taking the defect information, the real-time machine parameters and the collected time nodes as sample data.

[0063] The time nodes can be determined according to the type and size of the wafer. The defect information can include the number, distribution density, shape feature and the like of the defects. The real-time machine parameters can include the temperature, pressure, current, oxygen concentration, manufacturing time and the like set by the current manufacturing machine.

[0064] In some embodiments of the present application, the defect information includes at least one of the number of defects, the shape of defects and the distribution of defects in the wafer pattern data.

[0065] The number of defects can be the total number of all defects in the wafer pattern data, or the number of defects of a specified shape. The defect shape can be a point defect, a line defect and a region defect, etc. The defect distribution can be the distribution area of the defects in the wafer pattern data, or the distribution of the defects in each region, etc.

[0066] In the scheme, the defect information can be a klarf file, that is, the defect information of the wafer pattern sample can be recorded as a defect information file, and the klarf file can be a document for accessing defect information data such as xrel, yrel, xindex, yindex coordinate positions and the like.

[0067] Specifically, each time node and the defect information and real-time machine parameters corresponding to each time node can be integrated into one piece of sample data.

[0068] In step S202, data preprocessing and feature engineering processing are performed on the sample data to obtain standard sample data.

[0069] The data preprocessing is used to improve the quality of the sample data, so that the sample data is more suitable for analysis and model training. Common data preprocessing methods include data cleaning, data conversion, data integration, data reduction, etc. For example, repeated data, abnormal data and missing data in the sample data can be processed to clean the sample data.

[0070] The feature engineering processing is used to extract features from the sample data through data mining technology. Common feature engineering processing methods include standardization processing, interval scaling processing, normalization processing, quantitative feature binary processing, qualitative feature dummy coding processing, missing value processing and data transformation processing, etc.

[0071] The advantages of the scheme are that the standardization of the sample data is improved, the convergence speed of the model training is accelerated, the training efficiency is improved, and the training effect of the model is more rapid and accurate.

[0072] In step S203, the standard sample data is input into an initial model, the initial model is trained, and a time series model is obtained.

[0073] The training method can be supervised training or unsupervised training.

[0074] The time series model is used to predict future data based on historical data. The existing historical data can show the development trend of the data. The time series model can be an LSTM model, an AR model, an MA model, an ARMA model, an ARIMA model, etc.

[0075] Specifically, the standard sample data can be arranged in the order of time nodes and input into the initial model. The parameters of the initial model are adjusted in each training round until the training is completed to obtain the time series model.

[0076] In the training process, the training end condition can be that the prediction accuracy of the time series model is less than a preset threshold, for example, the preset threshold is 10%. The training end condition can also be that the number of training reaches a preset number. For example, the preset number is 20,000 times. A person skilled in the art can set the preset threshold or the preset number according to the actual situation, so that the time series model learns the trend of the current manufacturing machine standard sample data while avoiding overfitting.

[0077] In some embodiments of the present application, after obtaining the standard sample data, the method further comprises:

[0078] In the standard sample data, the defect information is taken as a continuous feature, and the real-time machine parameter is taken as a discrete feature.

[0079] Correspondingly, the standard sample data is input into an initial model, the initial model is trained, and a time series model is obtained, comprising:

[0080] The continuous feature and the discrete feature in the standard sample data are input into the initial model for feature recognition, and the training of the time series model is completed.

[0081] In the present solution, the number, distribution density, shape feature, etc. of the defects can be taken as the continuous feature continue_cols, and the real-time machine parameter can be taken as the discrete feature discrete_cols. Further, the continuous feature and the discrete feature in the standard sample data can be input into the initial model for feature recognition, and the training of the time series model is completed.

[0082] The advantage of the present technical solution is that the combination of discrete features and continuous features increases the richness and generalization ability of the features.

[0083] In some embodiments of the present application, the continuous feature and the discrete feature in the standard sample data are input into the initial model for feature recognition, and the training of the time series model is completed, comprising:

[0084] According to the time node of collecting the wafer pattern sample of the current manufacturing machine, a sliding time window is constructed.

[0085] Based on the standard sample data corresponding to each time node, the continuous feature and the discrete feature of each time node in the sliding time window are determined.

[0086] According to the continuous feature and the discrete feature of each time node in the sliding window, the initial model is used to predict defects of the target time node.

[0087] The sliding time window can be used to segment the time series feature data composed of the standard sample data corresponding to each time node, realize dimension reduction processing of the time series feature data, and reduce the calculation complexity.

[0088] Specifically, the time length, sliding length and prediction label length of the sliding time window can be set according to the time nodes of the wafer map sample of the current manufacturing machine collected to construct the sliding time window, wherein the prediction label is a target time to be predicted and a standard feature corresponding to the target time.

[0089] For example, in hours (H), the time length of the sliding time window is set to 10H, the sliding length is 1H, and the prediction label length is 1H, that is, the sliding time window can be 1H to 10H, and the prediction window can be 11H.

[0090] It can be understood that each time node in each sliding time window window may not have corresponding standard sample data, and the standard sample may be missing, but it does not affect the prediction of the prediction label. For example, the standard sample data corresponding to the time node 4H in the sliding time window window is missing, but the feature of the time node 11H can still be predicted according to the sliding time window window.

[0091] In addition, the prediction label length can also be determined according to actual needs. For example, the sliding time window is 1H to 10H, and the prediction label length can be set to 3H, that is, the prediction window can be 13H.

[0092] Further, according to the time length, sliding length and prediction label length of the sliding time window, and the standard sample data, the continuous features and discrete features corresponding to each time node in the sliding time window are determined.

[0093] The technical scheme has the advantages that by using the sliding time window, the time complexity can be reduced, the repeated calculation in the model training process can be reduced, the formation trend of the wafer manufacturing defect can be determined to predict the defect of the target time, and the training precision and accuracy of the model are improved.

[0094] In some embodiments of the present application, after the initial model is used to predict the defect of the target time node according to the continuous features and discrete features of each time node in the sliding window, the method further comprises:

[0095] Collecting a target wafer map sample of the target time node, and extracting the actual defect of the target wafer map sample;

[0096] Comparing the predicted defect result based on the actual defect to generate feedback information;

[0097] Optimizing the initial model based on the feedback information.

[0098] In the present solution, the target wafer pattern sample of the wafer at the target time node can be collected, and the defects in the collected target wafer pattern sample can be identified and judged to determine whether there is an actual defect in the target wafer pattern sample and the form of the actual defect if there is an actual defect.

[0099] Further, the actual defect is compared with the defect result obtained by the initial model for defect prediction at the target time node to obtain the defect similarity of the actual defect and the predicted defect, so as to determine the prediction accuracy of the initial model for defect prediction at the target time node. For example, if the defect similarity reaches a preset threshold, it is determined that the defect prediction result of the initial model meets the accuracy requirement, and the parameters of the initial model can not be changed; if the defect similarity does not reach the preset threshold, it is determined that the defect prediction result of the initial model does not meet the accuracy requirement, and part or all of the parameters of the initial model can be changed to improve the prediction accuracy of the model. For another example, part or all of the parameters of the initial model to be changed can be determined according to the defect similarity degree, and the higher the defect similarity degree, the smaller the change of the initial model parameters to be changed.

[0100] In some embodiments of the present application, after the initial model is optimized based on the feedback information, the method further comprises:

[0101] The sliding time window is slid to the right to cover the target time node, and defect prediction is performed on the next target time node based on the continuous features and discrete features of each collection time node in the sliding time window after sliding.

[0102] For example, in hours (H), if the difference between the starting time node and the ending time node collected is 300H, the time length of the sliding time window can be set to 10H, the sliding length is 1H, and the prediction label length is 1H, that is, the first sliding time window is 1H to 10H, the prediction label corresponding to the first sliding time window is 11H, the second sliding time window is 2H to 11H, the prediction label corresponding to the second sliding time window is 12H, and so on.

[0103] Figure 3 is a sliding time window diagram provided by Embodiment Two of the present application, as shown in Figure 3 The first row includes the X features corresponding to each time node in 1H to 10H, and the Y target corresponding to the target time of 11H; the second row is the sliding time window obtained by sliding the sliding time window of the first row to the right by 1H, including the X features corresponding to each time node in 2H to 11H, and the Y target corresponding to the target time of 12H; the third row includes the X features corresponding to each time node in 3H to 12H, and the Y target corresponding to the target time of 13H; and so on. The X features include continuous features and discrete features, and the Y target is the prediction feature corresponding to the target time.

[0104] The technical scheme has the advantages that by sliding the sliding time window and prediction, the formation trend of the wafer manufacturing defect can be accurately grasped, and the prediction accuracy of the wafer manufacturing defect is further improved.

[0105] Embodiment Three

[0106] Figure 4 is a structural schematic diagram of a wafer manufacturing defect prediction device provided by the embodiment three of the present application. As shown in the figure, the device specifically comprises the following: Figure 4 As shown in the figure, the device specifically comprises the following:

[0107] The data acquisition module 410 is configured to acquire wafer map data and machine parameter data of a wafer manufactured by a current manufacturing machine;

[0108] The defect prediction module 420 is configured to input the wafer map data and the machine parameter data into a pre-trained time series model to obtain a probability value of existence of a manufacturing defect of the current manufacturing machine at a target time;

[0109] The defect warning module 430 is configured to generate a wafer manufacturing defect warning information if the probability value meets a preset prompt condition.

[0110] The wafer manufacturing defect prediction device in the embodiment of the present application can be a device, or a component, an integrated circuit or a chip in a terminal. The device can be a mobile electronic device, or a non-mobile electronic device. Exemplarily, the mobile electronic device can be a mobile phone, a tablet computer, a notebook computer, a palm computer, a vehicle-mounted electronic device, a wearable device, an ultra-mobile personal computer (UMPC), a netbook or a personal digital assistant (PDA), etc., and the non-mobile electronic device can be a server, a network attached storage (NAS), a personal computer (PC), a television (TV), a teller machine or a self-service machine, etc., and the embodiment of the present application is not limited in this regard.

[0111] The wafer manufacturing defect prediction device in the embodiment of the present application can be a device with an operating system. The operating system can be an Android operating system, an iOS operating system or other possible operating systems, and the embodiment of the present application is not limited in this regard.

[0112] The wafer manufacturing defect prediction device provided by the embodiment of the present application can realize each process realized by the method embodiment, and thus the details are not repeated here.

[0113] Embodiment five

[0114] Further, the embodiment of the present application also provides a computer device, Figure 5 is a structural schematic diagram of the computer device provided by the embodiment five of the present application. The structure of the device is as shown in the figure Figure 5 The device includes a memory 51 for storing computer readable instructions and a processor 52 for executing the computer readable instructions, wherein when the computer readable instructions are executed by the processor, the processor executes the method.

[0115] The method and / or embodiment in the embodiment of the present application can be implemented as a computer software program. For example, the embodiment of the present application includes a computer program product comprising a computer program carried on a computer readable medium, the computer program comprising program code for executing the method shown in the flow chart. When the computer program is executed by a processing unit, the above-mentioned functions defined in the method of the present application are executed.

[0116] It should be noted that the computer readable medium described in the present application can be a computer readable signal medium or a computer readable storage medium or any combination of the two. The computer readable medium may, for example, but is not limited to, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, device or component, or any combination of the above. More specific examples of computer readable storage media can include, but are not limited to, electrical connections having one or more conductors, portable computer disks, hard disks, random access memories (RAM), read-only memories (ROM), erasable programmable read-only memories (EPROM or flash memory), optical fibers, portable compact disk read-only memories (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination of the above. In the present application, the computer readable medium can be any tangible medium containing or storing a program that can be used by or in conjunction with an instruction execution system, device or component.

[0117] In the present application, the computer readable signal medium can include a data signal propagating in a baseband or as a carrier wave in a propagated data signal, which carries the computer readable program code. Such a propagated data signal can take many forms, including but not limited to electromagnetic signals, optical signals or any suitable combination thereof. The computer readable signal medium can also be any computer readable medium other than the computer readable storage medium, which can send, propagate or transmit the program for use by or in conjunction with an instruction execution system, device or component. The program code contained on the computer readable medium can be transmitted by any suitable medium, including but not limited to wireless, wire, optical cable, RF, etc., or any suitable combination thereof.

[0118] Computer program code for carrying out operations of the present application can be written in any combination of one or more programming languages, including an object oriented programming language such as Java, Smalltalk, C++ or the like, and conventional procedural programming languages, such as the "C" programming language or similar programming languages. The program code can execute entirely on the user's computer, partly on the user's computer, as a stand-alone software package, partly on the user's computer and partly on a remote computer or entirely on the remote computer or server. In the latter scenario, the remote computer can be connected to the user's computer through any type of network, including a local area network (LAN) or a wide area network (WAN), or the connection can be made to an external computer (for example, through the Internet using an Internet Service Provider).

[0119] The flow diagrams and / or block diagrams in the drawings illustrate the architecture, functionality, and operation of possible implementations of devices, methods and computer program products according to various embodiments of the present application. In this regard, each block in the flow diagrams and / or block diagrams can represent a module, segment, or portion of code, which comprises one or more executable instructions for implementing the specified logical functions. It should also be noted that in some alternative implementations, the functions noted in the block can occur out of the order noted in the figures. For example, two blocks shown in succession may, in fact, be executed substantially concurrently, or the blocks can sometimes be executed in the reverse order, depending upon the functionality involved. It will also be noted that each block of the block diagrams and / or flow diagrams, and combinations thereof, can be implemented by special purpose hardware-based systems that perform the specified functions or operations, or combinations of special purpose hardware and

[0120] As another aspect, the embodiments of the present application also provide a computer readable medium, which can be contained in the apparatus described in the above embodiments, or can exist separately without being assembled into the apparatus. The above computer readable medium carries one or more computer readable instructions, which can be executed by a processor to implement the steps of the methods and / or the technical solutions of the above embodiments of the present application.

[0121] In a typical configuration of the present application, the apparatus of the terminal and the service network each includes one or more processors (CPU), input / output interface, network interface and memory.

[0122] Memory can include, without being limited to, non- transitory storage in computer-readable media, random access memory (RAM), and / or read-only memory (ROM), such as flash memory, etc. Memory is an example of computer-readable media.

[0123] Computer-readable media includes permanent and non- permanent, removable and non-removable media, implemented by any method or technology for storage of information such as computer-readable instructions, data structures, program modules or other data. Examples of storage media of a computer include, but are not limited to, phase change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other categories of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technologies, compact disc read-only memory (CD-ROM), digital versatile disc (DVD) or other optical storage, magnetic cassette, magnetic disk storage or other magnetic storage devices, or any other non-transmission medium that can be used to store information accessible to a computing device.

[0124] In addition, the embodiments of the present application further provide a computer program stored in a computer device, so that the computer device executes the method performed by the control code.

[0125] It should be noted that the present application can be implemented in software and / or a combination of software and hardware, for example, can be implemented by using an application specific integrated circuit (ASIC), a general purpose computer or any other similar hardware device. In some embodiments, the software program of the present application can be executed by a processor to implement the above steps or functions. Similarly, the software program of the present application (including related data structures) can be stored in a computer readable recording medium, for example, RAM memory, magnetic or optical drive or floppy disk and similar devices. In addition, some steps or functions of the present application can be implemented by hardware, for example, as a circuit cooperating with the processor to perform the respective steps or functions.

[0126] It will be obvious to a person skilled in the art that the application is not limited to the details of the above-described exemplary embodiments but can be implemented in other embodiments without departing from the scope of the application. The application is therefore not limited by the described examples but can vary within the scope of the claims and their equivalents. No reference signs in the claims should be considered as limiting the scope of the claims. Furthermore, it is obvious that the wording "comprising" does not exclude other parts and does not exclude other steps. Singularity is not excluded with respect to plurality and vice versa. Multiple units or devices recited in a device claim can also be implemented by one unit or device by means of software or hardware. The terms first, second and the like do not denote any ordering, but rather serve as names.

Claims

1. A method of predicting wafer manufacturing defects, characterized by, The method comprises: obtaining wafer map data and machine parameter data of a wafer manufactured by a current manufacturing machine; inputting the wafer map data and the machine parameter data into a pre-trained time series model, the time series model being used to analyze the formation trend of wafer manufacturing defects to obtain a probability value of the current manufacturing machine having manufacturing defects at a target time; if the probability value meets a preset prompt condition, generating a warning information of wafer manufacturing defects; wherein the training process of the time series model comprises: collecting wafer map samples of the current manufacturing machine at at least two time nodes, obtaining defect information from the wafer map samples, and collecting real-time machine parameters of the current manufacturing machine, taking the defect information, the real-time machine parameters and the collected time nodes as sample data; performing data preprocessing and feature engineering processing on the sample data to obtain standard sample data; inputting the standard sample data into an initial model, training the initial model to obtain a time series model; wherein the defect information comprises at least one of the defect quantity, the defect shape and the defect distribution of the wafer map data; wherein after obtaining the standard sample data, the method further comprises: in the standard sample data, taking the defect information as continuous features and taking the real-time machine parameters as discrete features; correspondingly, inputting the standard sample data into an initial model, training the initial model to obtain a time series model, comprises: inputting the continuous features and the discrete features in the standard sample data into the initial model for feature recognition to complete the training of the time series model; wherein inputting the continuous features and the discrete features in the standard sample data into the initial model for feature recognition to complete the training of the time series model comprises: according to the time nodes of collecting the wafer map samples of the current manufacturing machine, setting the time length, the sliding length and the prediction label length of the sliding time window, and constructing the sliding time window, wherein the prediction label length is used to define the time interval of the target time node to be predicted relative to the termination time of the sliding time window each time the sliding time window slides; based on the standard sample data corresponding to each time node, determining the continuous features and the discrete features of each time node in the sliding time window; based on the continuous features and the discrete features of each time node in the sliding window, using the initial model to predict the defect occurrence probability of the target time node.

2. The method of claim 1, wherein, after predicting the defect of the target time node based on the continuous features and the discrete features of each time node in the sliding window using the initial model, the method further comprises: collecting a target wafer map sample of the target time node, extracting the actual defect of the target wafer map sample; comparing the predicted defect result based on the actual defect to generate feedback information; optimizing the initial model based on the feedback information.

3. The method of claim 2, wherein, after optimizing the initial model based on the feedback information, the method further comprises: The sliding time window is slid right to cover the target time node, and a next target time node is predicted for defects based on continuous features and discrete features of each collection time node in the sliding time window after the sliding.

4. A wafer manufacturing defect prediction device characterized by comprising: The device comprises: a data acquisition module configured to acquire wafer map data and machine parameter data of a wafer manufactured by a current manufacturing machine; a defect prediction module configured to input the wafer map data and the machine parameter data into a pre-trained time series model, the time series model being configured to analyze a formation trend of wafer manufacturing defects to obtain a probability value of existence of a manufacturing defect of the current manufacturing machine at a target time; a defect warning module configured to generate a warning information of the wafer manufacturing defect if the probability value meets a preset prompt condition; The training process of the time series model comprises: collecting wafer map samples of the current manufacturing machine at at least two time nodes, acquiring defect information from the wafer map samples, and collecting real-time machine parameters of the current manufacturing machine, and taking the defect information, the real-time machine parameters, and the collected time nodes as sample data; performing data preprocessing and feature engineering processing on the sample data to obtain standard sample data; inputting the standard sample data into an initial model to train the initial model to obtain the time series model; The defect information comprises at least one of a defect number, a defect shape, and a defect distribution of the wafer map data. After obtaining the standard sample data, the method further comprises: in the standard sample data, taking the defect information as continuous features and taking the real-time machine parameters as discrete features; Correspondingly, inputting the standard sample data into an initial model to train the initial model to obtain the time series model comprises: inputting the continuous features and the discrete features in the standard sample data into the initial model for feature recognition to complete the training of the time series model; The inputting the continuous features and the discrete features in the standard sample data into the initial model for feature recognition to complete the training of the time series model comprises: setting a time length, a sliding length, and a prediction label length of a sliding time window according to the time nodes at which the wafer map samples of the current manufacturing machine are collected, and constructing the sliding time window, wherein the prediction label length is used to define a time interval of a target time node to be predicted for defects relative to a termination time of each sliding of the sliding time window; determining continuous features and discrete features of each time node in the sliding time window based on the standard sample data corresponding to each time node; performing defect prediction on a defect occurrence probability at a target time node using the initial model according to the continuous features and the discrete features of each time node in the sliding time window.

5. A computer device, comprising: The device comprises: one or more processors; and a memory storing computer program instructions that, when executed, cause the processor to perform the wafer manufacturing defect prediction method of any one of claims 1-3.

6. A computer readable medium characterized by A computer program product having stored thereon computer program instructions executable by a processor for implementing the method of claim 1-3.

Images