WF wafer detection equipment

Abstract

The invention discloses WF wafer detection equipment, and belongs to the field of wafer detection equipment, the WF wafer detection equipment comprises a workbench, the edge of the top end surface of the workbench is fixedly connected with a protection cover, and the two sides of the top end surface of the workbench are fixedly connected with basket assemblies; a WF detection platform assembly is arranged between the two basket assemblies, one side of the WF detection platform assembly is fixedly connected with a chuck manipulator assembly, the other side of the WF detection platform assembly is fixedly connected with a dotting assembly and an upper detection camera assembly, and the dotting assembly is located on one side of the upper detection camera assembly. The WF detection platform assembly specifically comprises a Y-axis linear module fixed to the top end face of the workbench, the top end of the Y-axis linear module is fixedly connected with an X-axis linear module, and the top end of the X-axis linear module is fixedly connected with a WF detection platform. According to the automatic feeding and discharging device, frequent feeding and discharging of workers are not needed, feeding and discharging can be automatically completed only by placing a plurality of wafer discs at designated positions, the overall automation degree is high, and use is convenient.

Description

Technical Field

[0001] The present invention relates to the field of wafer inspection equipment, and specifically to a WF wafer inspection equipment. Background Art

[0002] A wafer is called so because of its circular shape. Wafers can be used to fabricate silicon semiconductor integrated circuits. Multiple chips can be cut from one wafer, and the chips are formed into the chips seen in daily life after packaging. During the wafer production process, there may be defects such as damage, dirt, scratches, and cracks on the front and back sides of the wafer. These defects will affect the performance of the chips. Therefore, it is necessary to perform appearance inspection on the front and back sides of the wafer to determine whether the front and back sides of each chip are qualified.

[0003] The existing detection mode is to automatically detect through equipment, but it requires staff to frequently load and unload materials. Multiple wafers cannot be placed at designated positions to automatically complete the loading and unloading. The overall automation degree is low and it is not convenient to use. Therefore, those skilled in the art have provided a WF wafer inspection equipment to solve the problems raised in the above background art. Summary of the Invention

[0004] The purpose of the present invention is to provide a WF wafer inspection equipment that does not require staff to frequently load and unload materials. Only by placing multiple wafer trays at designated positions can the loading and unloading be automatically completed. The overall automation degree is high and it is convenient to use, so as to solve the problems raised in the above background art.

[0005] To achieve the above purpose, the present invention provides the following technical solutions:

[0006] A WF wafer inspection equipment includes a workbench. The edge of the top surface of the workbench is fixedly connected with a protective cover, and basket components are fixedly connected to both sides of the top surface of the workbench.

[0007] A WF inspection platform component is arranged between the two basket components. A chuck manipulator component is fixedly connected to one side of the WF inspection platform component, and a dotting component and an upper inspection camera component are fixedly connected to the other side of the WF inspection platform component, and the dotting component is located on one side of the upper inspection camera component.

[0008] As a further solution of the present invention: The WF inspection platform component specifically includes: a Y-axis linear module fixed on the top surface of the workbench, an X-axis linear module is fixedly connected to the top of the Y-axis linear module, and a WF inspection platform is fixedly connected to the top of the X-axis linear module.

[0009] As a further solution of the present invention: the upper detection camera assembly specifically includes: a Z-axis linear module fixed on one side of the WF detection platform assembly, the outer side surface of the Z-axis linear module is fixedly connected to the detection camera, and the bottom of the detection camera is fixedly connected to the detection lens, and the bottom of the detection lens is fixedly connected to the detection light source.

[0010] As a further solution of the present invention: the upper detection camera component captures images of the wafer after it arrives at the bottom, and then sends the captured images to the external background control terminal, which imports the images into a pre-trained OCR model, and the OCR model uses image processing and machine learning algorithms to perform defect detection on the wafer; wherein, the training process of the OCR model uses a feature extractor to extract information from the original data, and then combines the discriminator, classifier and class activation mapping to perform model training; in the process of defect detection on the wafer, the OCR model is combined with a text encoder, an image encoder, a prediction model, a text enhancement module, a visual enhancement module, an image decoder, a modulation module, and a large language model to distinguish defects on the wafer.

[0011] As a further solution of the present invention: the dot-marking assembly specifically includes: a vertical plate fixed on one side of the upper detection camera assembly, a rotating motor is fixedly connected to the top of one side of the vertical plate, and a vertical synchronous wheel is rotatably connected to the other side of the vertical plate corresponding to the position of the rotating motor, the output shaft of the rotating motor passes through the vertical plate and is fixedly connected to the vertical synchronous wheel, and a synchronous wheel idler is rotatably connected below the vertical synchronous wheel, a vertical synchronous belt is transmission-connected between the vertical synchronous wheel and the synchronous wheel idler, and a dot-marking pen is fixedly connected to the outer side of the vertical synchronous belt.

[0012] As a further solution of the present invention: the chuck manipulator assembly specifically includes: a support frame fixed on one side of the WF detection platform assembly, the support frame is internally rotatably connected to two parallel transverse synchronous wheels, a transverse synchronous belt is transmission-connected between the two transverse synchronous wheels, a drive motor is fixedly connected to the bottom of one of the transverse synchronous wheels, and the top output shaft of the drive motor is fixedly connected to the corresponding transverse synchronous wheel, the outer side surface of the transverse synchronous belt is fixedly connected to the chuck manipulator, and one end of the chuck manipulator is fixedly connected to a cylinder, the output shaft of the cylinder is connected to a clamp, and one side of the cylinder is fixedly connected to a sensor.

[0013] As a further solution of the present invention: the chuck robot assembly uses a 3D vision system to locate the wafer disc in the left basket assembly, and guides the chuck robot to grab it, and the chuck robot assembly uses vision to guide the chuck robot to place the wafer disc that has completed the inspection to a preset position on the right basket assembly; wherein, the 3D vision system imports the first-perspective point cloud map and the second-perspective point cloud map of the wafer disc into the point cloud encoder, and imports the 2D image corresponding to the point cloud map into the image encoder, and then combines exponential mapping, intra-modal hyperbolic contrast learning, and cross-modal hyperbolic contrast learning to accurately locate the wafer disc; in the process of using vision to guide the chuck robot to place the wafer disc that has completed the inspection to the preset position on the right basket assembly, the historical trajectory of the chuck robot is imported into the historical trajectory projection module, and the planned path of the chuck robot is imported into the planned path encoder, and then the robot encoder is combined with pooling, learnable type embedding, position embedding, scene encoder, and decoder to obtain the trajectory prediction result of the chuck robot, thereby realizing the material unloading guidance of the chuck robot.

[0014] As a further solution of the present invention: the basket lifting assembly specifically includes: a fixed plate fixed on the top surface of the workbench, a lifting motor is fixedly connected to the bottom end of one side of the fixed plate, and a transmission screw is fixedly connected to the top output shaft of the lifting motor, the external thread of the transmission screw is connected to a lifting seat, and one side of the lifting seat is fixedly connected to two parallel limit sliders, one side of the fixed plate is fixedly connected to a limit slide rail corresponding to the position of the limit slider, and the limit slider is movably connected to the limit slide rail, one side of the lifting seat is fixedly connected to a lifting frame, and the top surface of the lifting frame is fixedly connected to the basket.

[0015] Compared with the prior art, the present invention has the following beneficial effects:

[0016] 1. This application provides a basket assembly, a WF inspection platform assembly, a chuck manipulator assembly, a dot assembly, and an upper inspection camera assembly. Workers do not need to frequently load and unload materials. They only need to place multiple wafer trays in designated positions to automatically complete the loading and unloading. The overall degree of automation is high and easy to use.

[0017] 2. This application can quickly and accurately detect defects on wafer discs through the set OCR model, image processing and machine learning algorithms.

[0018] 3. This application uses a 3D vision system and visual guidance to quickly and accurately transfer the wafer to a specified position, thereby improving the accuracy of loading and unloading the wafer. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] Figure 1 It is a structural schematic diagram of a WF wafer inspection equipment;

[0020] Figure 2 It is a schematic structural diagram of a workbench in a WF wafer detection device;

[0021] Figure 3 It is a schematic structural diagram of a WF detection platform assembly in a WF wafer detection device;

[0022] Figure 4 It is a schematic structural diagram of an upper detection camera assembly in a WF wafer detection device;

[0023] Figure 5 It is a schematic structural diagram of a dotting assembly in a WF wafer detection device;

[0024] Figure 6 It is a schematic structural diagram of a chuck manipulator assembly in a WF wafer detection device;

[0025] Figure 7 It is a schematic structural diagram of a basket assembly in a WF wafer detection device;

[0026] Figure 8 It is a schematic diagram of positioning a wafer disk using a 3D vision system in this application;

[0027] Figure 9 It is a schematic diagram of training an OCR model in this application;

[0028] Figure 10 It is a schematic diagram of defect detection for a wafer disk in this application;

[0029] Figure 11 It is a schematic diagram of visual - guided chuck manipulator unloading in this application.

[0030] In the figure: 1. Workbench; 2. Protective cover; 3. Basket assembly; 301. Fixed plate; 302. Lifting motor; 303. Transmission lead screw; 304. Lifting seat; 305. Limit slide rail; 306. Limit slider; 307. Lifting frame; 308. Basket; 4. WF detection platform assembly; 401. Y - axis linear module; 402. X - axis linear module; 403. WF detection platform; 5. Chuck manipulator assembly; 501. Support frame; 502. Driving motor; 503. Horizontal synchronous pulley; 504. Horizontal synchronous belt; 505. Chuck manipulator; 506. Cylinder; 507. Clip; 508. Inductor; 6. Dotting assembly; 601. Vertical plate; 602. Rotating motor; 603. Vertical synchronous pulley; 604. Synchronous pulley idler; 605. Vertical synchronous belt; 606. Dotting pen; 7. Upper detection camera assembly; 701. Z - axis linear module; 702. Detection camera; 703. Detection lens; 704. Detection light source; 8. Wafer disk. Detailed implementation manners

[0031] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.

[0032] As mentioned in the background art of the present application, through research, it is found that the existing detection mode is to automatically detect through equipment, but it requires staff to frequently load and unload materials. It is impossible to automatically complete the loading and unloading of multiple wafers at designated positions, with low overall automation, inconvenient to use, and there are certain defects.

[0033] To solve the above defects, the present application discloses a WF wafer detection device, which does not require staff to frequently load and unload materials. Only by placing multiple wafer trays 8 at designated positions can the loading and unloading be automatically completed, with high overall automation and convenient to use.

[0034] The following will introduce in detail how the solution of the present application solves the above technical problems in conjunction with the accompanying drawings.

[0035] Please refer to Figure 1 and Figure 2 In the embodiments of the present invention, a WF wafer detection device includes a workbench 1. A protective cover 2 is fixedly connected to the edge of the top surface of the workbench 1, and basket components 3 are fixedly connected to both sides of the top surface of the workbench 1; a WF detection platform component 4 is arranged between the two basket components 3, and a chuck manipulator component 5 is fixedly connected to one side of the WF detection platform component 4. A dotting component 6 and an upper detection camera component 7 are fixedly connected to the other side of the WF detection platform component 4, and the dotting component 6 is located on one side of the upper detection camera component 7. By setting the basket component 3, WF detection platform component 4, chuck manipulator component 5, dotting component 6, and upper detection camera component 7 in the present application, it is not necessary for staff to frequently load and unload materials. Only by placing multiple wafer trays 8 at designated positions can the loading and unloading be automatically completed, with high overall automation and convenient to use.

[0036] In this embodiment, as Figure 3 shown, the WF detection platform component 4 specifically includes: a Y-axis linear module 401 fixed on the top surface of the workbench 1, an X-axis linear module 402 is fixedly connected to the top of the Y-axis linear module 401, and a WF detection platform 403 is fixedly connected to the top of the X-axis linear module 402. The WF detection platform component 4 can carry the wafer tray 8 and move it to a designated position for detection.

[0037] In this embodiment, as Figure 4As shown in the figure, the upper detection camera assembly 7 specifically includes: a Z-axis linear module 701 fixed to one side of the WF detection platform assembly 4. The outer side of the Z-axis linear module 701 is fixedly connected with a detection camera 702, and a detection lens 703 is fixedly connected below the detection camera 702, and a detection light source 704 is fixedly connected below the detection lens 703. The upper detection camera assembly 7 can quickly and accurately detect the crystal disk 8.

[0038] In this embodiment, the upper detection camera assembly 7 collects images of the crystal disk 8 after it reaches the lower part, and then sends the collected images to an external background control terminal. The external background control terminal imports the images into a pre-trained OCR model. The OCR model uses image processing and machine learning algorithms to detect defects on the crystal disk 8. Among them, the training process of the OCR model is as Figure 9 shown. A feature extractor is used to extract meaningful information from the original data, and then combined with a discriminator, a classifier, and class activation mapping for model training. It should be noted that the discriminator is usually used in generative adversarial networks (GANs). It is a binary classifier used to distinguish whether the input data comes from the real data distribution or the fake data generated by the generator. The main role of the discriminator is to improve the quality of the generated data. Through continuous adversarial training with the generator, the fake data generated by the generator is getting closer and closer to the real data. The classifier is a machine learning model used to assign input data to predefined categories. The main task of the classifier is to learn a mapping function according to the characteristics of the input data and map the input data to the category label. Class Activation Mapping (CAM) is a technique commonly used in convolutional neural networks (CNNs), aiming to identify and explain which regions play a key role in the final classification decision when the model performs image or time series classification tasks. The basic principle of CAM is to use the spatial information of the convolutional layer to understand the regions that the model focuses on when making classification decisions. It uses the Global Average Pooling (GAP) technique to convert the last layer feature map in the model into a heat map, thus revealing the regions that the model pays attention to when identifying images. The specific process of detecting defects on the crystal disk 8 is as Figure 10As shown, in the process, the OCR model is combined with a text encoder, an image encoder, a prediction model, a text enhancement module, a visual enhancement module, an image decoder, a modulation module, and a large language model to quickly identify defects on the wafer 8. Among them, the text encoder is one of the key technologies in natural language processing (NLP), which is responsible for converting text information into a vector representation that can be understood and processed by a computer. The image encoder is a machine learning model used to convert an image into a vector, thereby achieving tasks such as image classification, clustering, and retrieval. The prediction model is a model established by mathematical and statistical methods based on historical data and other relevant information, which is used to predict the probability or value of an event or phenomenon in the future. The text enhancement module is a module used to generate more training set data and adversarial samples, and to improve the effect of NLP tasks by improving the robustness and performance of the model. The visual enhancement module is a module that improves the visual quality of a video or image through algorithmic processing. The image decoder is a software or hardware device that converts the encoded image data back to its original form. In a communication system, the modulation module is used to convert the original signal into a signal suitable for transmission in a channel. A large language model is a language model with a large number of parameters and training data, which can capture the complexity and diversity of language.

[0039] In this embodiment, if Figure 5 As shown, the dotting assembly 6 specifically includes: a vertical plate 601 fixed on one side of the upper detection camera assembly 7, a rotating motor 602 is fixedly connected to the top of one side of the vertical plate 601, and a vertical synchronous wheel 603 is rotatably connected to the other side of the vertical plate 601 corresponding to the position of the rotating motor 602, the output shaft of the rotating motor 602 passes through the vertical plate 601 and is fixedly connected to the vertical synchronous wheel 603, and a synchronous wheel idler wheel 604 is rotatably connected below the vertical synchronous wheel 603, a vertical synchronous belt 605 is transmission-connected between the vertical synchronous wheel 603 and the synchronous wheel idler wheel 604, and a dotting pen 606 is fixedly connected to the outer side of the vertical synchronous belt 605. The dotting assembly 6 can quickly dot the wafer disc 8 that has completed the inspection.

[0040] In this embodiment, if Figure 6As shown, the chuck manipulator assembly 5 specifically includes: a support frame 501 fixed on one side of the WF detection platform assembly 4, the support frame 501 is internally rotatably connected to two parallel transverse synchronous wheels 503, a transverse synchronous belt 504 is transmission-connected between the two transverse synchronous wheels 503, a drive motor 502 is fixedly connected to the bottom of one of the transverse synchronous wheels 503, and the top output shaft of the drive motor 502 is fixedly connected to the corresponding transverse synchronous wheel 503, a chuck manipulator 505 is fixedly connected to the outer side of the transverse synchronous belt 504, and one end of the chuck manipulator 505 is fixedly connected to a cylinder 506, the output shaft of the cylinder 506 is connected to a clamp 507, and one side of the cylinder 506 is fixedly connected to a sensor 508. The chuck manipulator assembly 5 can transfer the wafer disc 8, thereby realizing automatic loading and unloading of the wafer disc 8.

[0041] In this embodiment, if Figure 8 As shown, the chuck robot assembly 5 uses a 3D vision system to locate the wafer disk 8 in the left basket assembly 3 and guide the chuck robot 505 to grab it. Figure 11As shown, the chuck manipulator assembly 5 uses visual guidance to place the inspected crystal disk 8 at a preset position on the right basket assembly 3. Among them, the 3D vision system imports the first perspective point cloud map and the second perspective point cloud map of the crystal disk 8 into the point cloud encoder, and imports the corresponding 2D image of the point cloud map into the image encoder. Then, combined with exponential mapping, in-modal hyperbolic contrast learning, and cross-modal hyperbolic contrast learning, precise positioning of the crystal disk 8 is achieved. The point cloud encoder is a technology widely used in the field of 3D computer vision. It is mainly responsible for processing discrete point sets in three-dimensional space, that is, point cloud data. The image encoder is a device or algorithm used to convert image data into a format suitable for storage, transmission, or processing. Exponential mapping is an important tool for describing the relationship between Lie groups and Lie algebras. It can map elements in Lie algebras to elements in Lie groups, thereby realizing parametric representation of the elements of Lie groups. In-modal hyperbolic contrast learning is a self-supervised learning method that focuses on learning data representations within the same modality (such as images, text, or audio). Cross-modal hyperbolic contrast learning is a deep learning technology that involves processing and understanding data from different modalities (such as text, images, sounds, etc.). During the process of using visual guidance to place the inspected crystal disk 8 at a preset position on the right basket assembly 3, the historical trajectory of the manipulator is imported into the historical trajectory projection module, and the planned path of the manipulator is imported into the planned path encoder. Then, combined with the manipulator encoder, pooling, learnable type embedding, position embedding, scene encoder, and decoder, the manipulator trajectory prediction result is obtained, and further, the blanking guidance of the manipulator is realized. Here, the manipulator refers to the chuck manipulator 505. Among them, pooling is a commonly used operation in deep learning, usually used in Convolutional Neural Networks (CNNs). The purpose of pooling is to reduce the computational amount and the number of parameters of the network by downsampling the input feature map, thereby preventing overfitting. Learnable type embedding is a technology used in machine learning and deep learning that allows the model to automatically obtain and represent the type information of the data through learning. Position embedding is a technology used to encode the position information of elements in a sequence. The scene encoder refers to an encoder in the field of machine vision or computer vision that is used to convert the input image or video scene into a compact and effective representation form. The decoder is an electronic device or algorithm used to convert the encoded signal or data back to its original form.

[0042] In this embodiment, as Figure 7As shown in the figure, the basket assembly 3 specifically includes: a fixing plate 301 fixed on the top surface of the workbench 1. At the bottom end of one side surface of the fixing plate 301, a lifting motor 302 is fixedly connected. And the top output shaft of the lifting motor 302 is fixedly connected with a transmission lead screw 303. The outer part of the transmission lead screw 303 is threadedly connected with a lifting seat 304. And on one side surface of the lifting seat 304, two juxtaposed limiting sliders 306 are fixedly connected. At the position corresponding to the limiting sliders 306 on one side surface of the fixing plate 301, a limiting slide rail 305 is fixedly connected. And the limiting sliders 306 are movably connected with the limiting slide rail 305. On one side surface of the lifting seat 304, a lifting frame 307 is fixedly connected. And at the top surface of the lifting frame 307, a basket 308 is fixedly connected. The basket assembly 3 on the left can cooperate with the chuck manipulator assembly 5 to complete the automatic loading of the wafer disk 8. The basket assembly 3 on the right can cooperate with the chuck manipulator assembly 5 to complete the automatic unloading of the wafer disk 8.

[0043] The working principle of the present invention is as follows: When in use, first, the staff places the wafer disk 8 to be detected into the basket assembly 3 on the left. Subsequently, the chuck manipulator assembly 5 operates to pick up a wafer disk 8 from the basket assembly 3 on the left and transfer the wafer disk 8 to the WF detection platform assembly 4. Immediately afterwards, the WF detection platform assembly 4 transfers the wafer disk 8 under the upper detection camera assembly 7 for barcode scanning recognition and defect detection. After the detection is completed, the WF detection platform assembly 4 transfers the wafer disk 8 under the dotting assembly 6 for dotting. After the dotting is completed, the chuck manipulator assembly 5 operates again to transfer the wafer disk 8 to the basket assembly 3 on the right to complete the unloading.

[0044] The operation process of the chuck manipulator assembly 5 is specifically as follows: The driving motor 502 operates to drive the upper horizontal synchronous pulley 503 to rotate. Another horizontal synchronous pulley 503 rotates synchronously with the horizontal synchronous belt 504. During the process, the chuck manipulator 505 on the horizontal synchronous belt 504 moves accordingly. When the inductor 508 senses a preset signal, the air cylinder 506 controls the clamp 507 to perform the opening and clamping actions on the wafer disk 8, so as to realize functions such as picking up, placing and transferring the wafer disk 8.

[0045] The operation process of the WF detection platform assembly 4 is specifically as follows: After the chuck manipulator assembly 5 transfers the wafer disk 8 to the WF detection platform 403, the Y-axis linear module 401 and the X-axis linear module 402 can be used in cooperation to adjust the position of the wafer disk 8 in the horizontal plane. Among them, both the Y-axis linear module 401 and the X-axis linear module 402 are composed of a servo motor, a lead screw and a guide rail.

[0046] The Z-axis linear module 701 of the upper detection camera assembly 7 is composed of a stepper motor and a linear module. During the detection process, the operation of the stepper motor can control the lifting action of the detection camera 702. When the WF detection platform 403 moves to the position of the detection camera 702, the detection camera 702 and the detection lens 703 are used to detect the appearance of the product, and the detection light source 704 is used for fill light adjustment.

[0047] The operation process of the dotting assembly 6 is as follows: When the WF detection platform 403 moves to the position of the dotting pen 606, the rotation motor 602 runs to drive the vertical synchronous pulley 603 to rotate, and the synchronous pulley idler 604 rotates synchronously with the vertical synchronous belt 605. The dotting pen 606 on the vertical synchronous belt 605 descends to dot on the surface of the crystal disk 8 on the WF detection platform 403.

[0048] The operation process of the basket assembly 3 is as follows: When loading the crystal disk 8, every time the chuck manipulator assembly 5 takes away a crystal disk 8, the lifting motor 302 of the left basket assembly 3 runs to raise the left basket 308 by one unit height. And every time the chuck manipulator assembly 5 sends a crystal disk 8 into the right basket assembly 3 for receiving materials, the lifting motor 302 of the right basket assembly 3 runs to lower the right basket 308 by one unit height. Here, the unit height refers to the vertical height difference between two adjacent crystal disks 8. During the operation of the lifting motor 302, the transmission lead screw 303 rotates accordingly, the lifting seat 304 moves up and down along the transmission lead screw 303, the lifting frame 307 together with the basket 308 moves up and down, and the limit slider 306 and the limit slide rail 305 have relative displacement to improve stability.

[0049] In this application, by setting the basket assembly 3, the WF detection platform assembly 4, the chuck manipulator assembly 5, the dotting assembly 6, and the upper detection camera assembly 7, it is not necessary for workers to frequently load and unload materials. Just place multiple crystal disks 8 at the designated positions, and the loading and unloading can be automatically completed. The overall automation degree is high and it is convenient to use.

[0050] Obviously, those skilled in the art can make various changes and modifications to the present invention without departing from the spirit and scope of the present invention. Thus, if these modifications and variations of the present invention fall within the scope of the claims of the present invention and their equivalent technologies, the present invention also intends to include these changes and modifications.

[0051] The above is only the preferred specific implementation manner of the present invention, but the protection scope of the present invention is not limited thereto. Any person skilled in the art within the technical scope disclosed by the present invention, according to the technical solution and inventive concept of the present invention, makes equivalent substitutions or changes, and all should be covered within the protection scope of the present invention.

Claims

1. A WF wafer inspection device, characterized in that: It comprises a workbench (1), wherein a protective cover (2) is fixedly connected to the edge of the top surface of the workbench (1), and basket lifting assemblies (3) are fixedly connected to both sides of the top surface of the workbench (1); A WF detection platform component (4) is provided between the two basket components (3), and a chuck manipulator component (5) is fixedly connected to one side of the WF detection platform component (4), and a dot-marking component (6) and an upper detection camera component (7) are fixedly connected to the other side of the WF detection platform component (4), and the dot-marking component (6) is located on one side of the upper detection camera component (7).

2. A WF wafer inspection device according to claim 1, characterized in that: The WF detection platform assembly (4) specifically comprises: a Y-axis linear module (401) fixed on the top surface of the workbench (1), the top of the Y-axis linear module (401) is fixedly connected to an X-axis linear module (402), and the top of the X-axis linear module (402) is fixedly connected to the WF detection platform (403).

3. A WF wafer inspection device according to claim 1, characterized in that: The upper detection camera assembly (7) specifically comprises: a Z-axis linear module (701) fixed on one side of the WF detection platform assembly (4); the outer side surface of the Z-axis linear module (701) is fixedly connected to a detection camera (702); the bottom of the detection camera (702) is fixedly connected to a detection lens (703); and the bottom of the detection lens (703) is fixedly connected to a detection light source (704).

4. A WF wafer inspection device according to claim 3, characterized in that: The upper detection camera assembly (7) collects images of the wafer disc (8) after it reaches the bottom, and then sends the collected images to the external background control terminal. The external background control terminal imports the images into a pre-trained OCR model. The OCR model uses image processing and machine learning algorithms to perform defect detection on the wafer disc (8).

5. A WF wafer inspection device according to claim 4, characterized in that: The training process of the OCR model uses a feature extractor to extract information from raw data, and then combines a discriminator, a classifier, and a class activation map to perform model training.

6. A WF wafer inspection device according to claim 4, characterized in that: In the process of defect detection on the wafer disc (8), the OCR model is combined with a text encoder, an image encoder, a prediction model, a text enhancement module, a visual enhancement module, an image decoder, a modulation module, and a large language model to identify defects on the wafer disc (8).

7. A WF wafer inspection device according to claim 1, characterized in that: The dotting assembly (6) specifically comprises: a vertical plate (601) fixed on one side of the upper detection camera assembly (7); a rotating motor (602) is fixedly connected to the top of one side of the vertical plate (601); and a vertical synchronous wheel (603) is rotatably connected to the other side of the vertical plate (601) corresponding to the position of the rotating motor (602); an output shaft of the rotating motor (602) passes through the vertical plate (601) and is fixedly connected to the vertical synchronous wheel (603); and a synchronous wheel idler wheel (604) is rotatably connected below the vertical synchronous wheel (603); a vertical synchronous belt (605) is transmission-connected between the vertical synchronous wheel (603) and the synchronous wheel idler wheel (604); and a dotting pen (606) is fixedly connected to the outer side of the vertical synchronous belt (605).

8. A WF wafer inspection device according to claim 1, characterized in that: The chuck manipulator assembly (5) specifically comprises: a support frame (501) fixed on one side of the WF detection platform assembly (4); the support frame (501) is internally rotatably connected to two parallel transverse synchronous wheels (503); a transverse synchronous belt (504) is transmission-connected between the two transverse synchronous wheels (503); a driving motor (502) is fixedly connected to the bottom of one of the transverse synchronous wheels (503); and the top output shaft of the driving motor (502) is fixedly connected to the corresponding transverse synchronous wheel (503); a chuck manipulator (505) is fixedly connected to the outer side surface of the transverse synchronous belt (504); and one end of the chuck manipulator (505) is fixedly connected to a cylinder (506); the output shaft of the cylinder (506) is connected to a clamp (507); and one side of the cylinder (506) is fixedly connected to a sensor (508).

9. A WF wafer inspection device according to claim 8, characterized in that: The chuck robot assembly (5) uses a 3D vision system to locate the wafer disk (8) in the left basket assembly (3) and guides the chuck robot (505) to grasp it. The chuck robot assembly (5) uses vision to guide the chuck robot (505) to place the wafer disk (8) that has completed the inspection to a preset position on the right basket assembly (3).

10. A WF wafer inspection device according to claim 9, characterized in that: The 3D vision system imports the first-view point cloud image and the second-view point cloud image of the wafer disk (8) into a point cloud encoder, and imports the 2D image corresponding to the point cloud image into an image encoder, and then combines index mapping, intra-modal hyperbolic contrast learning, and cross-modal hyperbolic contrast learning to accurately locate the wafer disk (8).

11. The WF wafer inspection device according to claim 9, characterized in that: In the process of using vision to guide the chuck robot (505) to place the wafer disk (8) that has completed the inspection to the preset position on the right basket assembly (3), the historical trajectory of the chuck robot (505) is imported into the historical trajectory projection module, and the planned path of the chuck robot (505) is imported into the planned path encoder, and then the robot encoder is combined with pooling, learnable type embedding, position embedding, scene encoder, and decoder to obtain the trajectory prediction result of the chuck robot (505), thereby realizing the unloading guidance of the chuck robot (505).

12. The WF wafer inspection device according to claim 1, characterized in that: The basket lifting assembly (3) specifically comprises: a fixing plate (301) fixed on the top surface of the workbench (1); a lifting motor (302) is fixedly connected to the bottom end of one side of the fixing plate (301); a transmission screw (303) is fixedly connected to the top output shaft of the lifting motor (302); an external thread of the transmission screw (303) is connected to a lifting seat (304); and two parallel limiting sliders (306) are fixedly connected to one side of the lifting seat (304); a limiting slide rail (305) is fixedly connected to one side of the fixing plate (301) at a position corresponding to the limiting slider (306); and the limiting slider (306) is movably connected to the limiting slide rail (305); a lifting frame (307) is fixedly connected to one side of the lifting seat (304); and a basket lifting (308) is fixedly connected to the top surface of the lifting frame (307).

Images