Wafer defect detection method and device based on neural network and storage medium

Abstract

The invention provides a wafer defect detection method and device based on a neural network and a storage medium, and the method comprises the steps: determining a to-be-detected wafer, and collecting a radio frequency signal passing through the to-be-detected wafer; preprocessing the radio frequency signal to obtain a wafer radio frequency signal; converting the wafer radio frequency signal into modal data of the wafer to be detected; and inputting the modal data into a preset classification-neural network model, and performing data processing and identification on the modal data by using the preset classification-neural network model to obtain a defect position and type corresponding to the wafer to be detected. The wafer defect is detected by using an automatic detection technology of the neural network to obtain the wafer defect position and the wafer defect type, so that the wafer defect detection efficiency is improved, accurate identification and classification can be realized in the face of complex and variable defect types and backgrounds, and the missing detection rate and the false detection rate are greatly reduced.

Description

Technical Field

[0001] The present invention relates to the technical field of intelligent defect detection, and in particular to a wafer defect detection method, device and storage medium based on neural network. Background Art

[0002] With the rapid development of the semiconductor industry, wafers, as the core material for integrated circuit manufacturing, have a quality that directly determines the performance and reliability of the final chip. However, wafers undergo multiple complex process flows during production, including photolithography, etching, ion implantation, deposition, etc., all of which may introduce various defects, such as scratches, protrusions, dislocations, metallic contaminants, etching rust, etc. These defects not only affect the yield and performance of the chip, but may also cause the entire production line to shut down and cause huge economic losses.

[0003] In the prior art, traditional wafer defect detection methods mainly rely on manual visual inspection, which results in low detection efficiency and is easily affected by subjective factors. Summary of the invention

[0004] The basic disclosure provides a wafer defect detection method, device and storage medium based on a neural network, which improves the detection efficiency through the automatic detection technology of the neural network, and can also achieve accurate identification and classification when facing complex and changeable defect types and backgrounds, greatly reducing the missed detection and false detection rates.

[0005] In a first aspect, the present disclosure provides a wafer defect detection method based on a neural network, comprising: determining a wafer to be tested, and collecting a radio frequency signal passing through the wafer to be tested;

[0006] Preprocessing the radio frequency signal to obtain a wafer radio frequency signal;

[0007] Converting the wafer radio frequency signal into modal data of the wafer to be tested;

[0008] The modal data is input into a preset classification-neural network model, and the preset classification-neural network model is used to process and identify the modal data to obtain the defect position and type corresponding to the wafer to be tested.

[0009] In some embodiments, converting the wafer radio frequency signal into the modal data of the wafer to be tested includes:

[0010] Obtaining the actual signal receiving time corresponding to each wafer radio frequency signal and the preset signal receiving time corresponding to each wafer radio frequency signal, wherein the actual signal receiving time and the preset signal receiving time have a one-to-one correspondence;

[0011] Calculate the time difference between the actual signal reception time corresponding to each wafer radio frequency signal and the preset signal reception time to obtain each time difference;

[0012] Converting each of the time differences into defect data of the wafer to be tested;

[0013] The modal data of the standard wafer is acquired, and the modal data of the standard wafer is updated using the defect data to obtain the modal data of the wafer to be tested.

[0014] In some embodiments, before inputting the modal data into a preset classification-neural network model, and using the preset classification-neural network model to process and identify the modal data to obtain the defect position and type corresponding to the wafer to be tested, the method further includes:

[0015] Determine a classification-neural network model, wherein the classification-neural network model is a convolutional neural network model;

[0016] Obtain at least two modal data corresponding to the standard wafer, as well as the target defect position and target defect type corresponding to each modal data,

[0017] Input each modal data into the initial-classification neural network model for calculation, and obtain the predicted defect position and predicted defect type corresponding to the modal data;

[0018] Respectively calculating a position deviation value between the predicted defect position and the target defect position, and a type deviation value between the predicted defect type and the target defect type;

[0019] Determine whether the position deviation value is within a preset position deviation value range, and whether the type deviation value is within a preset type deviation value range;

[0020] When the position deviation value is within the preset position deviation range and the type deviation value is within the preset type deviation value range, the classification-neural network is trained in the next round until the classification-neural network is in a convergence state to obtain a preset classification-neural network model.

[0021] In some embodiments, the preprocessing of the radio frequency signal to obtain a wafer radio frequency signal includes:

[0022] Amplifying the radio frequency signal using a signal amplifier to obtain an amplified radio frequency signal;

[0023] The amplified radio frequency signal is subjected to interference signal filtering processing to obtain a wafer radio frequency signal.

[0024] In some embodiments, the modality data is one of image data, spectral data or radio frequency data.

[0025] In some embodiments, the RF transmitter and the receiver are placed on the same side or on opposite sides of the wafer to be tested.

[0026] In some embodiments, when the RF transmitter and the receiver are placed on opposite sides of the wafer to be tested, the distance between the RF transmitter and the RF receiver is D, where D satisfies the following conditions:

[0027] D>0, ensuring that the RF signal can propagate inside the wafer;

[0028] Where H is the maximum thickness of the wafer to prevent the signal from propagating outside the wafer;

[0029] The selection of D is based on the wavelength λ of the RF signal, so that the distance between the transmitter and the receiver is an integer multiple of the wavelength to improve the coherence of the signal reception.

[0030] In a second aspect, the present disclosure provides a wafer defect detection device based on a neural network, comprising:

[0031] A determination module, used to determine a wafer to be tested and collect a radio frequency signal passing through the wafer to be tested;

[0032] A preprocessing module, used for preprocessing the radio frequency signal to obtain a wafer radio frequency signal;

[0033] A first conversion module, used for converting the wafer radio frequency signal into modal data of the wafer to be tested;

[0034] The detection module is used to input the modal data into a preset classification-neural network model, and use the preset classification-neural network model to process and identify the modal data to obtain the defect position and type corresponding to the wafer to be tested.

[0035] In a third aspect, the present disclosure provides a computer device comprising a memory and a processor, wherein the memory stores a computer program, and wherein the processor implements the steps of the neural network-based wafer defect detection method described in any of the above embodiments when executing the computer program.

[0036] In a fourth aspect, the present disclosure provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of the neural network-based wafer defect detection method described in any of the above embodiments.

[0037] The present invention provides a neural network-based wafer defect detection method, device and storage medium, which detect wafer defects by utilizing the automatic detection technology of the neural network to obtain the wafer defect position and wafer defect type, thereby improving the efficiency of wafer defect detection, and can also achieve accurate identification and classification when facing complex and changeable defect types and backgrounds, greatly reducing the missed detection and false detection rates. BRIEF DESCRIPTION OF THE DRAWINGS

[0038] Hereinafter, the present disclosure will be described in more detail based on embodiments and with reference to the accompanying drawings;

[0039] Figure 1 1 is a schematic diagram of a process of a wafer defect detection method based on a neural network in one embodiment;

[0040] Figure 2 is a schematic structural diagram of a wafer defect detection device based on a neural network in one embodiment;

[0041] Figure 3 Schematic diagram of the internal structure of a computer device in one embodiment.

[0042] In the drawings, the same reference numerals are used for the same components, and the drawings are not drawn to scale. DETAILED DESCRIPTION

[0043] In order to enable those skilled in the art to better understand the technical solution of the present disclosure, and to fully understand and implement how the present disclosure applies technical means to solve technical problems and achieve the corresponding technical effects, the technical solution in the embodiments of the present disclosure will be clearly and completely described below in conjunction with the drawings in the embodiments of the present disclosure. Obviously, the described embodiments are only embodiments of a part of the present disclosure, not all of the embodiments. The embodiments of the present disclosure and the various features in the embodiments can be combined with each other without conflict, and the technical solutions formed are all within the scope of protection of the present disclosure. Based on the embodiments in the present disclosure, all other embodiments obtained by ordinary technicians in this field without making creative work should fall within the scope of protection of the present disclosure.

[0044] It should be noted that the terms "first", "second", etc. in the specification and claims of the present disclosure and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence. It should be understood that the data used in this way can be interchangeable where appropriate, so that the embodiments of the present disclosure described herein can be implemented in an order other than those illustrated or described herein. In addition, the terms "including" and "having" and any variations thereof are intended to cover non-exclusive inclusions, for example, a process, method, system, product, or device that includes a series of steps or units is not necessarily limited to those steps or units that are clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products, or devices.

[0045] It should be noted that the steps shown in the flowcharts of the accompanying drawings can be executed in a computer system such as a set of computer executable instructions, and that, although a logical order is shown in the flowcharts, in some cases, the steps shown or described can be executed in an order different from that shown here.

[0046] Embodiment 1

[0047] In this embodiment, Figure 1 As shown, a wafer defect detection method based on a neural network is provided, which includes:

[0048] Step 111 , determining a wafer to be tested, and collecting a radio frequency signal passing through the wafer to be tested.

[0049] In this embodiment, the wafer is identified by the wafer identification system, and when the wafer enters the identification area, the wafer entering the identification area is determined to be a wafer to be tested. In one embodiment, the RF transmitter and the receiver are placed on the same side or opposite sides of the wafer to be tested. When the RF transmitter and the receiver are placed on opposite sides of the wafer to be tested, the wafer to be tested is located between the transmitter and the receiver. In one embodiment, when the RF transmitter and the receiver are placed on opposite sides of the wafer to be tested, the distance between the RF transmitter and the RF receiver is D, where D satisfies the following conditions: D>0, ensuring that the RF signal can propagate inside the wafer; Where H is the maximum thickness of the wafer to prevent the signal from propagating outside the wafer; D is selected based on the wavelength λ of the RF signal, so that the distance between the transmitter and the receiver is an integer multiple of the wavelength to improve the coherence of the signal reception.

[0050] Based on the above foundation, a signal channel is established between the RF transmitter and the video receiver. The transmitter transmits the RF signal to the signal channel, and the receiver receives the RF signal on the signal channel. The RF transmitter is turned on to transmit the RF signal, and the RF signal is received by the receiver after passing through the wafer to be tested. The RF signal received by the receiver is the RF signal passing through the wafer to be tested. In this embodiment, the RF signal received by the receiver is accompanied by the receiving time of the RF signal.

[0051] In this embodiment, the principle of collecting the radio frequency signal passing through the wafer to be tested is: the principle of collecting the radio frequency signal passing through the wafer to be tested.

[0052] Step 112: pre-process the radio frequency signal to obtain a wafer radio frequency signal.

[0053] After the RF signal passing through the wafer to be tested is collected, the RF signal is preprocessed to obtain the wafer RF signal. In this embodiment, the preprocessing of the RF signal may include: signal enhancement, bandwidth limitation, frequency conversion, analog filtering, digital filtering, signal quantization or gain control and other technical processing. When the RF signal received by the receiver is weak, the RF signal is enhanced (amplified); when the RF signal received by the receiver has an interference signal, the RF signal is filtered to obtain a RF signal without interference.

[0054] In this embodiment, the RF signal is preprocessed to improve the quality, stability and anti-interference ability of the RF signal and provide reliable data support for subsequent RF signal processing and analysis. The signal obtained after the RF signal is preprocessed is used as the wafer RF signal.

[0055] Step 113: convert the wafer radio frequency signal into modal data of the wafer to be tested.

[0056] After obtaining the wafer RF signal, determine the actual receiving time of each wafer RF signal and the preset receiving time of the signal corresponding to each wafer RF signal, and there is a one-to-one correspondence between the actual receiving time of the signal and the preset receiving time of the signal. Then calculate the time difference of the wafer to be tested that affects the propagation of the RF signal based on the actual receiving time and the preset receiving time; finally, further determine the defect data of the wafer to be tested based on the time difference. In this embodiment, the principle corresponding to the defect data of the wafer to be tested determined by the time difference is: different media use different times to transmit RF signals. When the medium changes, the receiving time of the RF signal will also change, because the receiving time of the RF signal can be used to determine whether the current wafer to be tested is the same as the standard wafer; when defects or foreign matter appear on the surface of the wafer to be tested, the receiving time of the RF signal will change, so the modal data of the wafer to be tested is obtained by converting each of the time differences into defect data of the wafer to be tested, and using the defect data to update the modal data of the standard wafer.

[0057] In one embodiment, the modality data is one of image data, spectral data or radio frequency data.

[0058] In this embodiment, the preset signal receiving time can be obtained in a database, in the cloud, or on any storage medium; no specific limitation is made here.

[0059] Step 114, input the modal data into a preset classification-neural network model, use the preset classification-neural network model to process and identify the modal data, and obtain the defect position and type corresponding to the wafer to be tested.

[0060] In this embodiment, the modal data of the wafer to be tested is input into the trained model, and the preset classification-neural network model will output the type and location of the defect. For image data, the model may directly output an image with defect location markers (such as bounding boxes) and type labels. For other types of modal data, the output of the preset classification-neural network model needs to be further processed to obtain intuitive defect location and type information, that is, data type conversion is performed.

[0061] In this embodiment, the preset classification-neural network model can be a convolutional neural network (CNN) or a recurrent neural network (RNN); it can also be a combination of CNN and RNN, or combined with other types of networks (such as a fully connected network) to process complex modal data; no specific limitation is made here.

[0062] In one embodiment, the defect detection process of the preset classification-neural network model includes: extracting defect features and classifying the features; building a 13-layer CNN model to perform PCA dimensionality reduction on the classified defect features; and sorting the reduced features by similarity to infer the defect cause of the test wafer.

[0063] In this embodiment, the wafer to be tested is first determined, and the RF signal passing through the wafer to be tested is collected; the RF signal is then preprocessed to obtain the wafer RF signal; the wafer RF signal is then converted into the modal data of the wafer to be tested; finally, the modal data is input into a preset classification-neural network model, and the modal data is processed and identified using the preset classification-neural network model to obtain the defect position and type corresponding to the wafer to be tested. By using the automatic detection technology of the neural network to detect wafer defects, the wafer defect position and wafer defect type are obtained, so as to improve the efficiency of wafer defect detection, and when facing complex and changeable defect types and backgrounds, accurate identification and classification can also be achieved, greatly reducing the missed detection and false detection rates.

[0064] It should be understood that although Figure 1 The steps in the flowchart are shown in sequence as indicated by the arrows, but these steps are not necessarily executed in the order indicated by the arrows. Unless otherwise specified in this document, there is no strict order restriction for the execution of these steps, and these steps can be executed in other orders. Moreover, Figure 1 At least part of the steps may include multiple sub-steps or multiple stages. These sub-steps or stages are not necessarily executed at the same time, but can be executed at different times. The execution order of these sub-steps or stages is not necessarily sequential, but can be executed in turn or alternately with other steps or at least part of the sub-steps or stages of other steps.

[0065] In one embodiment, converting the wafer radio frequency signal into the modal data of the wafer to be tested includes:

[0066] 1-1) Obtaining the actual signal receiving time corresponding to each wafer RF signal and the preset signal receiving time corresponding to each wafer RF signal, wherein the actual signal receiving time and the preset signal receiving time have a one-to-one correspondence.

[0067] In this embodiment, before converting the wafer RF signal into the modal data of the wafer to be tested, it is necessary to obtain the actual signal receiving time corresponding to each wafer RF signal and the corresponding signal preset receiving time, and there is a one-to-one correspondence between the actual signal receiving time and the signal preset receiving time.

[0068] For example, the RF signal of the n-th wafer is N, and the actual receiving time and the preset receiving time of the RF signal N are recorded as: {N→T1, N→T2}.

[0069] 1-2) Calculate the time difference between the actual signal reception time corresponding to each wafer RF signal and the signal preset reception time to obtain each time difference.

[0070] After obtaining the actual receiving time and the preset receiving time of each signal corresponding to the wafer RF signal, the time difference between the actual receiving time and the preset receiving time is calculated to obtain each time difference. For example, N→T2 minus N→T1 is used to obtain the time difference P N When the time difference is within the time error range, it means that the actual signal reception time is the same as the signal preset reception time. When the time difference is not within the time error range, it means that the actual signal reception time is different from the signal preset reception time.

[0071] In this embodiment, the time error range can be preset according to actual conditions and is not specifically limited here.

[0072] 1-3) Converting each of the time differences into defect data of the wafer to be tested.

[0073] When the actual signal receiving time is different from the preset signal receiving time, it indicates that there is a foreign object or defect on the wafer to be tested. The time difference between the actual signal receiving time and the preset signal receiving time is calculated, and the defect data corresponding to the time difference is queried. In this embodiment, the time difference and the defect data may have a one-to-one correspondence or a one-to-many relationship. When the time difference and the defect data are in a one-to-many relationship, it is necessary to calculate the probability of each defect data corresponding to the time difference, and then further determine the defect data based on the probability.

[0074] 1-4) Acquire the modal data of the standard wafer, and use the defect data to update the modal data of the standard wafer to obtain the modal data of the wafer to be tested.

[0075] After determining the defect data of the wafer to be tested, the defect data of the wafer to be tested is converted into data in a data format corresponding to the modal data of the standard wafer, so that the modal data of the wafer to be tested can be directly formed by replacement. Specifically, the signal channel corresponding to the defect data is determined, and the partial modal data of the standard wafer corresponding to the signal channel is found, and the defect data is used to replace the partial modal data to update the modal data of the standard wafer, and finally the modal data of the wafer to be tested is obtained.

[0076] In this embodiment, by converting the wafer RF data (whether directly from the RF signal or not) into modal data, it can be more conveniently analyzed and processed automatically. This helps to quickly identify abnormal areas or performance defects on the wafer, thereby improving overall test efficiency.

[0077] In one embodiment, before inputting the modal data into a preset classification-neural network model, and using the preset classification-neural network model to process and identify the modal data to obtain the defect position and type corresponding to the wafer to be tested, the method further includes:

[0078] 2-1) Determine a classification-neural network model, wherein the classification-neural network model is a convolutional neural network model.

[0079] 2-2) Obtain at least two modal data corresponding to the standard wafer, and the target defect position and target defect type corresponding to each modal data.

[0080] In this embodiment, the various states of the standard wafer can be captured by the feature capture device, and each state corresponds to a modal data; in practical applications, at least the modal data of the normal state of the standard wafer, the modal data of various types of defect states of the standard wafer, and the modal data of the state of the standard wafer with preset foreign matter are captured. And the target defect position and target defect type corresponding to each modal data are used to provide an accurate data basis for training and determining the classification-neural network model.

[0081] 2-3) Input each modal data into the initial-classification neural network model for calculation to obtain the predicted defect position and predicted defect type corresponding to the modal data.

[0082] 2-4) Calculate the position deviation value between the predicted defect position and the target defect position, and the type deviation value between the predicted defect type and the target defect type respectively.

[0083] 2-5) Determine whether the position deviation value is within a preset position deviation value range, and whether the type deviation value is within a preset type deviation value range.

[0084] 2-6) When the position deviation value is within the preset position deviation range, and the type deviation value is within the preset type deviation value range, the classification-neural network is subjected to the next round of training until the classification-neural network is in a convergence state to obtain a preset classification-neural network model.

[0085] In this embodiment, when the position deviation value is within the preset position deviation range, and the type deviation value is within the preset type deviation value range, it means that the various weight parameters of the classification-neural network model at this time can satisfy the trained modal data, and the next round of training is required for the untrained modal data until the classification-neural network is in a convergence state to obtain the preset classification-neural network model.

[0086] In this embodiment, multimodal data of the standard wafer (such as images, spectra, electrical characteristics, etc.) can provide more comprehensive information, which may have limitations when used alone, but can complement each other when used in combination, thereby improving the accuracy of classification.

[0087] In one embodiment, after determining the preset classification-neural network model, it also includes: extracting any defect position, size and type information output by the neural network model; actually marking the extracted wafer according to the extracted information; using a high-precision scanning electron microscope to analyze the defect position and compare it with the output results of the neural network model; based on the comparison results, optimizing the preset classification-neural network model, that is, adjusting the weight parameters corresponding to the defect position and the weight parameters corresponding to the defect type in the preset classification-neural network model.

[0088] In one embodiment, the preprocessing of the radio frequency signal to obtain a wafer radio frequency signal includes:

[0089] 3-1) Amplifying the radio frequency signal using a signal amplifier to obtain an amplified radio frequency signal.

[0090] In this embodiment, there is usually an output matching network near the output port of the amplifier to ensure that the impedance of the output signal matches the impedance of the load (such as an antenna or the next stage amplifier). This helps to maximize energy transmission and reduce reflections. The signal amplified by the amplifier and passed through the output matching network is transmitted to the output port of the amplifier to form an amplified RF signal.

[0091] In this embodiment, the amplifier may be a low noise amplifier (LNA), a power amplifier (PA), or a driver amplifier; no specific limitation is made here.

[0092] 3-2) Filter the interference signal of the amplified RF signal to obtain a wafer RF signal.

[0093] In this embodiment, the filter is placed after the amplifier, and it is ensured that the filter can effectively filter the interference signal while retaining the target signal. The amplified RF signal is input to the filter, and the amplified RF signal is filtered by the filter to remove the interference signal. The filtered RF signal is the wafer RF signal.

[0094] In this embodiment, the reason for amplifying the RF signal first and then filtering the interference signal is that if the noise is filtered first and then the signal is amplified, the useful components in the signal may be excessively weakened or lost due to the attenuation characteristics of the filter. Amplifying the signal first can ensure that sufficient useful information is retained during the filtering process. Furthermore, the amplified signal has a higher signal-to-noise ratio, which makes it easier to distinguish between the signal and the noise during the filtering process, so that a more appropriate filtering method and parameters can be used to remove the noise.

[0095] In one embodiment, the RF transmitter and the receiver are placed on the same side or on two opposite sides of the wafer to be tested.

[0096] In this embodiment, when the RF transmitter and the RF receiver are located on the same side of the wafer to be tested, the wafer to be tested transmits the RF signal to the RF receiver by way of RF signal reflection, so that the RF receiver can receive the RF signal.

[0097] In this embodiment, placing on the same side simplifies the test structure and reduces complexity; placing on opposite sides may increase the complexity of system design. When the RF transmitter and receiver are placed on opposite sides of the wafer, the signal path between them is longer, which helps to separate the transmit and receive signals in space and reduce mutual interference.

[0098] Embodiment 2

[0099] In this embodiment, Figure 2 As shown, a wafer defect detection device based on a neural network is provided, including: a determination module 211, a preprocessing module 212, a first conversion module 213 and a detection module 214.

[0100] The determination module 211 is used to determine a wafer to be tested and collect a radio frequency signal passing through the wafer to be tested.

[0101] The preprocessing module 212 is used to preprocess the RF signal to obtain a wafer RF signal.

[0102] The first conversion module 213 is used to convert the wafer RF signal into modal data of the wafer to be tested.

[0103] The detection module 214 is used to input the modal data into a preset classification-neural network model, and use the preset classification-neural network model to process and identify the modal data to obtain the defect position and type corresponding to the wafer to be tested.

[0104] In one embodiment, the first conversion module 213 may include: an acquisition unit, a calculation unit, a conversion unit, and an update unit.

[0105] The acquisition unit is used to acquire the actual signal receiving time corresponding to each wafer RF signal and the preset signal receiving time corresponding to each wafer RF signal, and there is a one-to-one correspondence between the actual signal receiving time and the preset signal receiving time.

[0106] The calculation unit is used to calculate the time difference between the actual signal receiving time corresponding to each wafer radio frequency signal and the signal preset receiving time to obtain each time difference.

[0107] The conversion unit is used to convert each of the time differences into defect data of the wafer to be tested.

[0108] An updating unit is used to obtain the modal data of the standard wafer, and use the defect data to update the modal data of the standard wafer to obtain the modal data of the wafer to be tested.

[0109] In one embodiment, the wafer defect detection device based on neural network may further include: a model determination module, a data acquisition module, a first calculation module, a second calculation module, a judgment module and a training module.

[0110] The model determination module is used to determine the classification-neural network model, and the classification-neural network model is a convolutional neural network model.

[0111] The data acquisition module is used to acquire at least two modal data corresponding to the standard wafer, and the target defect position and target defect type corresponding to each modal data.

[0112] The first calculation module is used to input each modal data into the initial-classification neural network model for calculation to obtain the predicted defect position and predicted defect type corresponding to the modal data.

[0113] The second calculation module is used to respectively calculate the position deviation value between the predicted defect position and the target defect position, and the type deviation value between the predicted defect type and the target defect type.

[0114] The judging module is used to judge whether the position deviation value is within a preset position deviation value range, and whether the type deviation value is within a preset type deviation value range.

[0115] A training module is used to perform the next round of training on the classification-neural network when the judgment module determines that the position deviation value is within the preset position deviation range and the type deviation value is within the preset type deviation value range, until the classification-neural network is in a convergence state, so as to obtain a preset classification-neural network model.

[0116] In one embodiment, the pre-processing module 212 may include: an amplification unit and a filtering unit.

[0117] The amplifying unit is used to amplify the radio frequency signal using a signal amplifier to obtain an amplified radio frequency signal.

[0118] The filtering unit is used to filter the interference signal of the amplified radio frequency signal to obtain a wafer radio frequency signal.

[0119] In one embodiment, the modality data is one of image data, spectral data or radio frequency data.

[0120] In one embodiment, the RF transmitter and the receiver are placed on the same side or on two opposite sides of the wafer to be tested.

[0121] In one embodiment, when the RF transmitter and the receiver are placed on opposite sides of the wafer to be tested, the distance between the RF transmitter and the RF receiver is D, where D satisfies the following conditions:

[0122] D>0, ensuring that the RF signal can propagate inside the wafer;

[0123] Where H is the maximum thickness of the wafer to prevent the signal from propagating outside the wafer;

[0124] The selection of D is based on the wavelength λ of the RF signal, so that the distance between the transmitter and the receiver is an integer multiple of the wavelength to improve the coherence of the signal reception.

[0125] For the specific limitations of the wafer defect detection device based on a neural network, please refer to the limitations of the wafer defect detection method based on a neural network above, which will not be repeated here. Each unit in the above-mentioned wafer defect detection device based on a neural network can be implemented in whole or in part through software, hardware and a combination thereof. The above-mentioned units can be embedded in or independent of the processor in the computer device in the form of hardware, or can be stored in the memory of the computer device in the form of software, so that the processor can call and execute the operations corresponding to the above-mentioned units.

[0126] Embodiment 3

[0127] In this embodiment, a computer device is provided. Its internal structure diagram can be shown as follows: Figure 3 As shown. The computer device includes a processor, a memory, a network interface, a display screen and an input device connected through a system bus. Among them, the processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and a computer program, and the non-volatile storage medium is deployed with a database, which is used to store all data involved in the wafer defect detection method based on a neural network. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The network interface of the computer device is used to communicate with other computer devices that deploy application software. When the computer program is executed by the processor, a wafer defect detection method based on a neural network is implemented. The display screen of the computer device can be a liquid crystal display screen or an electronic ink display screen, and the input device of the computer device can be a touch layer covered on the display screen, or a key, trackball or touchpad set on the computer device housing, or an external keyboard, touchpad or mouse.

[0128] Those skilled in the art will understand that Figure 3 The structure shown in the figure is only a block diagram of a part of the structure related to the solution of the present application, and does not constitute a limitation on the computer device to which the solution of the present application is applied. The specific computer device may include more or fewer components than those shown in the figure, or combine certain components, or have a different arrangement of components.

[0129] In one embodiment, a computer device is provided, including a memory and a processor, wherein the memory stores a computer program, and the processor implements the steps of the neural network-based wafer defect detection method described in any of the above embodiments when executing the computer program.

[0130] Embodiment 4

[0131] In this embodiment, a computer-readable storage medium is provided, on which a computer program is stored. When the computer program is executed by a processor, the steps of the neural network-based wafer defect detection method described in any of the above embodiments are implemented.

[0132] Those skilled in the art can understand that all or part of the processes in the above-mentioned embodiment methods can be completed by instructing the relevant hardware through a computer program, and the computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to memory, storage, database or other media used in the embodiments provided in this application can include non-volatile and / or volatile memory. Non-volatile memory can include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM) or flash memory. Volatile memory can include random access memory (RAM) or external cache memory. As an illustration and not limitation, RAM is available in many forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous link (Synchlink) DRAM (SLDRAM), memory bus (Rambus) direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM).

[0133] The technical features of the above embodiments may be combined arbitrarily. To make the description concise, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0134] The above-mentioned embodiments only express several implementation methods of the present application, and the descriptions thereof are relatively specific and detailed, but they cannot be understood as limiting the scope of the invention patent. It should be pointed out that, for a person of ordinary skill in the art, several variations and improvements can be made without departing from the concept of the present application, and these all belong to the protection scope of the present application. Therefore, the protection scope of the patent of the present application shall be subject to the attached claims.

Claims

1. A wafer defect detection method based on neural network, characterized in that: include: Determining a wafer to be tested, and collecting a radio frequency signal passing through the wafer to be tested; Preprocessing the radio frequency signal to obtain a wafer radio frequency signal; Converting the wafer radio frequency signal into modal data of the wafer to be tested; The modal data is input into a preset classification-neural network model, and the preset classification-neural network model is used to process and identify the modal data to obtain the defect position and type corresponding to the wafer to be tested.

2. The wafer defect detection method according to claim 1, characterized in that: The step of converting the wafer radio frequency signal into the modal data of the wafer to be tested comprises: Obtaining the actual signal receiving time corresponding to each wafer radio frequency signal and the preset signal receiving time corresponding to each wafer radio frequency signal, wherein the actual signal receiving time and the preset signal receiving time have a one-to-one correspondence; Calculate the time difference between the actual signal reception time corresponding to each wafer radio frequency signal and the preset signal reception time to obtain each time difference; Converting each of the time differences into defect data of the wafer to be tested; The modal data of the standard wafer is acquired, and the modal data of the standard wafer is updated using the defect data to obtain the modal data of the wafer to be tested.

3. The wafer defect detection method according to claim 1, characterized in that: Before inputting the modal data into a preset classification-neural network model, and using the preset classification-neural network model to process and identify the modal data to obtain the defect position and type corresponding to the wafer to be tested, the method further includes: Determine a classification-neural network model, wherein the classification-neural network model is a convolutional neural network model; Acquire at least two modal data corresponding to the standard wafer, and the target defect position and target defect type corresponding to each modal data; Input each modal data into the initial-classification neural network model for calculation, and obtain the predicted defect position and predicted defect type corresponding to the modal data; Respectively calculating a position deviation value between the predicted defect position and the target defect position, and a type deviation value between the predicted defect type and the target defect type; Determine whether the position deviation value is within a preset position deviation value range, and whether the type deviation value is within a preset type deviation value range; When the position deviation value is within the preset position deviation range and the type deviation value is within the preset type deviation value range, the classification-neural network is trained in the next round until the classification-neural network is in a convergence state to obtain a preset classification-neural network model.

4. The wafer defect detection method according to claim 3, characterized in that: The preprocessing of the radio frequency signal to obtain a wafer radio frequency signal includes: Amplifying the radio frequency signal using a signal amplifier to obtain an amplified radio frequency signal; The amplified radio frequency signal is subjected to interference signal filtering processing to obtain a wafer radio frequency signal.

5. The wafer defect detection method according to claim 3, characterized in that: The modality data is one of image data, spectral data or radio frequency data.

6. The wafer defect detection method according to claim 1, characterized in that: The radio frequency transmitter and the receiver are placed on the same side or on two opposite sides of the wafer to be tested.

7. The wafer defect detection method according to claim 6, characterized in that: When the RF transmitter and the receiver are placed on opposite sides of the wafer to be tested, the distance between the RF transmitter and the RF receiver is D, where D satisfies the following conditions: D>0, ensuring that the RF signal can propagate inside the wafer; Where H is the maximum thickness of the wafer to prevent the signal from propagating outside the wafer; The selection of D is based on the wavelength λ of the RF signal, so that the distance between the transmitter and the receiver is an integer multiple of the wavelength to improve the coherence of the signal reception.

8. A wafer defect detection device based on neural network, characterized in that: include: A determination module, used to determine a wafer to be tested and collect a radio frequency signal passing through the wafer to be tested; A preprocessing module, used for preprocessing the radio frequency signal to obtain a wafer radio frequency signal; A first conversion module, used for converting the wafer radio frequency signal into modal data of the wafer to be tested; The detection module is used to input the modal data into a preset classification-neural network model, and use the preset classification-neural network model to process and identify the modal data to obtain the defect position and type corresponding to the wafer to be tested.

9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the processor executes the computer program, the steps of the method according to any one of claims 1 to 7 are implemented.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 7 are implemented.

Images