Self-supervision enhanced semi-supervised wafer failure prediction method and system

Abstract

The invention discloses a self-supervision enhanced semi-supervised wafer failure prediction method and system, and belongs to the field of wafer failure detection. Obtaining wafer acceptability test / probe test data to construct a training set; carrying out self-supervision pre-training on the wafer feature encoder; a pre-training wafer feature encoder is used to generate original wafer coding features of each sample, strong / weak noise is added, data enhancement constraint is introduced when the strong noise is added, and loss is calculated; using a wafer failure classifier to predict whether the wafer fails according to the wafer coding features before and after the strong / weak noise is added, and calculating consistency loss; in addition, the supervision loss of a small amount of labeled data needs to be calculated; and semi-supervised training is performed on the pre-trained wafer feature encoder and the wafer failure classifier by combining the loss, and the semi-supervised trained model is used for actual wafer failure detection. According to the method, a large amount of wafer label-free data and a small amount of wafer label data are fully utilized, and the prediction accuracy of wafer failure is improved.

Description

Technical Field

[0001] The invention belongs to the field of wafer failure detection, and in particular relates to a self-supervised enhanced semi-supervised wafer failure prediction method and system. Background Art

[0002] The manufacture of integrated circuits (ICs) involves complex procedures, including thin film deposition, ion implantation, etching, and meticulous polishing. As the industry moves toward miniaturization, circuit elements shrink and designs become increasingly layered, subjecting each chip to an extensive series of processing steps. In this fine orchestration, even the slightest deviation can cause defects in the wafer manufacturing process and lead to wafer failure. During and after manufacturing, a series of tests are performed to ensure that the chips are functionally intact, such as wafer acceptability testing and wafer probe testing. However, there is a time difference (1-2 weeks) between wafer acceptability testing and wafer probe testing, and during this period, the fab does not stop. Only after the wafer probe test results come out can we know whether there is a wafer failure in the manufacturing process. If there is a problem with the previous batch of manufacturing, it will cause a large number of wafer failures, resulting in a lot of waste of time and money, which may be as high as millions, so timely prediction of wafer failure problems is very critical.

[0003] The cornerstone of wafer failure prediction has always been model construction centered on machine learning. For example, some studies have adopted methods such as logistic regression, XGBoost, and RF. However, although machine learning models have good effects in most industrial scenarios, they require a large number of data sets as support. For wafer failure prediction, a sample is matched with a wafer. The manufacturing of a wafer requires hundreds or thousands of steps. The manufacturing cost is very high, and collecting a large number of data sets means high costs. At the same time, there are many wafers waiting to be tested in wafer manufacturing, which meet the standards of unlabeled data. Machine learning technology cannot use these unlabeled data for learning. In recent years, inspired by the success of self-supervised and semi-supervised learning in image fields and language tasks, some recent studies have explored the application of self-supervised and semi-supervised learning in tasks in the field of tabular data, but there is no related research in the scenario of wafer failure prediction. For example, the self-supervised and semi-supervised framework VIME introduced in 2020 proposed to use unlabeled data to improve the performance of the model in the field of tabular data. However, despite the breakthrough, the VIME model still faces limitations in industrial environments, mainly because its simple network structure is difficult to adapt to the complexity of industrial data. Specifically, VIME uses a simple fully connected layer as an encoder, which makes it difficult to extract the relationship between complex data sets, resulting in a decrease in network performance. In addition, the performance of the VIME model is limited under small sample conditions, which is related to the fact that the VIME model uses many perturbations for consistency during the semi-supervised training process. The small sample uses masked data for perturbation, which makes the model's learning of the relationship between data more ambiguous, resulting in a decrease in model performance. In addition, many studies have proposed their own views on the self-supervised training and fine-tuning methods in the field of tabular data, mainly migrating the self-supervised training and pre-training methods in the field of images to the field of tabular data, such as SAINT, SCARF, Subtab and other models. However, these models are not fully applied to unlabeled data, resulting in limited model performance and poor performance on industrial data sets. In general, the above methods are all aimed at prediction tasks such as house prices, income, and stock market data. At present, there are no excellent self-supervised and semi-supervised models proposed for industrial data sets that can efficiently extract complex correlations between industrial data sets.

[0004] Faced with these obstacles, there is an urgent need for a method that not only needs to extract and identify the interrelationships between manufacturing process data with higher accuracy, but also needs to provide robust wafer failure prediction results under limited data availability. In addition, in order to utilize failure test data to enhance the performance of the model and improve the accuracy of prediction, a new model for the field of wafer failure prediction must be developed. Summary of the invention

[0005] In order to realize wafer failure prediction using unlabeled data and a small amount of labeled data, the present invention proposes a self-supervised enhanced semi-supervised wafer failure prediction method and system, which alleviates the requirement of machine learning model on data volume, and utilizes a large amount of unlabeled data and a small amount of labeled data tested by probes in the wafer manufacturing process to improve the accuracy of wafer failure prediction.

[0006] In order to achieve the above object, the technical solution adopted by the present invention is as follows:

[0007] In a first aspect, the present invention proposes a self-supervised enhanced semi-supervised wafer failure prediction method, comprising:

[0008] Obtain wafer acceptability test data and probe test data, preprocess and match the data, and obtain labeled wafer samples and unlabeled wafer samples as training sets;

[0009] Using the training set to perform self-supervised mask pre-training on the wafer feature encoder to obtain a pre-trained wafer feature encoder;

[0010] The pre-trained wafer feature encoder is used to generate the original wafer coding features of each sample in the training set, and strong noise and weak noise are added to the original wafer coding features respectively; when adding strong noise to the original wafer coding features, a data enhancement constraint is introduced, and the data enhancement constraint loss is calculated;

[0011] The original wafer coding features, the strongly enhanced wafer coding features after adding strong noise, and the weakly enhanced wafer manufacturing acceptability data coding features after adding weak noise are respectively used by the wafer failure classifier to predict whether the wafer fails, and the consistency loss is calculated according to the failure prediction results; in addition, the supervision loss needs to be calculated for the labeled wafer samples in the training set according to the failure prediction results;

[0012] The data enhancement constraint loss, consistency loss and supervision loss are combined to perform semi-supervised training on the pre-trained wafer feature encoder and wafer failure classifier, and the pre-trained wafer feature encoder and wafer failure classifier after the semi-supervised training are used to predict whether the wafer fails.

[0013] Furthermore, the preprocessing and matching of data includes:

[0014] Handle outliers and missing values ​​on wafer acceptability test data;

[0015] Match the pre-processed wafer acceptability test data and probe test data, use a group of successfully matched data as labeled wafer samples, use unsuccessfully matched wafer acceptability test data as unlabeled wafer samples, and discard unsuccessfully matched probe test data;

[0016] The label of the labeled wafer sample is the result of the probe test data.

[0017] Furthermore, the self-supervised mask pre-training of the wafer feature encoder using the training set includes:

[0018] Randomly mask the samples in the training set;

[0019] The masked samples use the wafer feature encoder to generate wafer coding features, and the mask position and mask value are predicted based on the wafer coding features. The wafer feature encoder parameters are iteratively updated according to the prediction results to complete self-supervised pre-training.

[0020] Furthermore, the weak noise added to the original wafer coding feature is weak Gaussian noise.

[0021] Further, the method of adding strong noise to the original wafer coding feature includes:

[0022] Inputting the original wafer coding feature of the sample into the initialized strong noise decoder, and generating a strong enhanced noise sample after decoding;

[0023] The strongly enhanced noise samples are used to generate strongly enhanced wafer coding features using a pre-trained wafer feature encoder, thereby adding strong noise to the original wafer coding features;

[0024] The strong noise decoder participates in semi-supervised training together with the pre-trained wafer feature encoder and the wafer failure classifier.

[0025] Furthermore, the data enhancement constraint loss is as follows:

[0026]

[0027] in, represents the data augmentation constraint loss, represents the original sample i in the training set, represents a strongly enhanced noise sample, represents the square of the L2 norm, represents the dimension of the encoded features generated by the pre-trained wafer feature encoder, They represent the variance and mean of the j-th dimension of the encoded feature generated by the pre-trained wafer feature encoder for the original sample i, represents the preset hyperparameters, Represents the original number of samples in the training set.

[0028] Furthermore, the supervision loss refers to the binary cross entropy loss between the prediction results of the original wafer coding features using the wafer failure classifier and the true label; the consistency loss refers to the binary cross entropy loss between the prediction results of the strongly enhanced wafer coding features and the weakly enhanced wafer coding features using the wafer failure classifier respectively.

[0029] Furthermore, when the data enhancement constraint loss, consistency loss and supervision loss are combined to perform semi-supervised training on the pre-trained wafer feature encoder and wafer failure classifier, the weight of the consistency loss gradually increases from 0 to a preset weight.

[0030] In a second aspect, the present invention proposes a self-supervised enhanced semi-supervised wafer failure prediction system for implementing the above-mentioned self-supervised enhanced semi-supervised wafer failure prediction method.

[0031] The beneficial effects of the present invention are:

[0032] The present invention proposes a self-supervised enhanced semi-supervised model framework for wafer failure prediction, which combines self-supervised and semi-supervised training processes to model the wafer acceptability test data of the wafer manufacturing process, and enhances the learning ability of the semi-supervised process by calculating the attention weights between data through self-supervised training. It further adopts strong noise and weak noise data enhancement methods and introduces them into the training of the semi-supervised process through the consistency principle between pseudo labels. It achieves important and effective data expansion with very little labor cost, and enhances the semi-supervised training process through self-supervised training, further improving the performance of the model in predicting wafer failure.

[0033] The present invention makes full use of wafer data that has been probed and wafer data that has not been probed, avoiding data waste, while improving the stability of the model, reducing resource waste in the manufacturing process, saving costs, and helping to speed up the root cause analysis of wafer failure and improve the yield rate. BRIEF DESCRIPTION OF THE DRAWINGS

[0034] Figure 1 A flow chart of wafer failure prediction data collection according to an embodiment of the present invention;

[0035] Figure 2 This is a flowchart for self-supervised training;

[0036] Figure 3 This is a flowchart for semi-supervised training;

[0037] Figure 4 Schematic diagram of semi-supervised application. DETAILED DESCRIPTION

[0038] The following description is used to disclose the present invention so that those skilled in the art can implement the present invention.

[0039] The accompanying drawings are only schematic illustrations of the present invention and are not necessarily drawn to scale. Some of the block diagrams shown in the accompanying drawings are functional entities that do not necessarily correspond to physically or logically independent entities. These functional entities can be implemented in software form, or in one or more hardware modules or integrated circuits, or in different networks and / or processor devices and / or microcontroller devices.

[0040] The flowcharts shown in the accompanying drawings are only exemplary and do not necessarily include all the steps. For example, some steps may be decomposed, while some steps may be combined or partially combined, so the actual execution order may change according to the actual situation.

[0041] The present invention proposes a self-supervised enhanced semi-supervised wafer failure prediction method, which mainly includes the following steps:

[0042] S1. Wafer data collection.

[0043] Collect and pre-process test data related to the wafer manufacturing process, wherein the wafer manufacturing process and its related test data include wafer acceptability test data and wafer probe test data; the wafer probe test data refers to the test result of whether the wafer has failed.

[0044] Data preprocessing methods include but are not limited to data merging, data cleaning, outlier processing, missing value processing, etc. The preprocessing process can be implemented according to the existing technology in the field and will not be described in detail here.

[0045] In a specific implementation of the present invention, Figure 1As shown in the figure, there are hundreds of steps in the wafer manufacturing process, such as chemical vapor deposition process, etching process, photolithography process, chemical mechanical polishing process, cleaning process, ion implantation process, etc. In order to ensure the stability of the process and the integrity of the chip function, the manufactured wafers need to be subjected to acceptability testing and probe testing. Wafer Acceptance Testing (WAT) is a key quality control step in the semiconductor manufacturing process. It is used to evaluate the performance and quality of wafers in the manufacturing process. The purpose of WAT is to detect the electrical characteristics of wafers at different manufacturing stages to ensure that they meet the design specifications and quality standards. Usually including resistance, capacitance, leakage current, threshold voltage, etc., these parameters help identify defects in the manufacturing process. Wafer probe testing (Circuit Probing Testing) is to measure the electrical properties of each chip through probe stations and automatic test equipment during the semiconductor manufacturing process to screen out unqualified chips. It evaluates parameters such as voltage, current and logic functions, generates yield reports, and provides feedback for process optimization, thereby improving product quality and production efficiency. By collecting wafer acceptability test data and probe test data, and preprocessing the data.

[0046] The pre-processed data is matched, and the wafer acceptability test data and the probe test data that are successfully matched are regarded as labeled wafer samples. The labels of the labeled wafer samples are the results of the probe test data, including failure and normal. The wafer acceptability test data that are not successfully matched are regarded as unlabeled wafer samples. The successful matching here means that the wafer acceptability test data and the probe test data are collected for the same wafer.

[0047] S2. Construct a self-supervised enhanced semi-supervised wafer failure prediction model.

[0048] The self-supervised enhanced semi-supervised wafer failure prediction model consists of a self-supervised network and a semi-supervised network.

[0049] In a specific implementation of the present invention, the self-supervised network includes a mask generator, a wafer feature encoder and a self-supervised prediction layer. The wafer feature encoder adopts an FTTransformer encoder, which itself integrates a Transformer function including an attention mechanism. The mask generator is used to generate a random mask proxy task for samples in a training set. The prediction layer is composed of two different fully connected layers, which serve as a mask position predictor and a mask value predictor, respectively, for two self-supervised proxy tasks.

[0050] The semi-supervised network includes a weak noise generator, a wafer feature encoder and a semi-supervised prediction layer. The wafer feature encoder is the same as the wafer feature encoder in the self-supervised network, and directly loads the model parameters trained by self-supervision; the semi-supervised prediction layer is composed of two different fully connected layers, which serve as a strong noise decoder and a wafer failure classifier, respectively, and are used to add strong noise to the original wafer coding features generated by the wafer feature encoder during the semi-supervised training process, and to predict whether the wafer fails; the weak noise generator is used to add weak noise to the original wafer coding features generated by the wafer feature encoder during the semi-supervised training process, and this embodiment uses weak Gaussian noise.

[0051] The FTTransformer encoder used in the present invention includes a feature segmenter, a multi-head self-attention calculation unit, an adder, a normalization unit, a residual network, a fully connected layer, and an activation unit. The feature segmenter is responsible for encoding the input wafer data, which is convenient for the multi-head self-attention calculation unit to calculate the attention weights between the data. The interaction information between the data can be effectively extracted through the attention weights, increasing the robustness of the model; the residual network and the adder are used to retain the original information of the data to avoid the phenomenon of information loss after passing through the multi-layer network; the normalization unit is responsible for normalizing the data, improving the stability and generalization ability of the model, and avoiding the phenomenon of too large or too small data, resulting in gradient explosion or small gradient; the activation function introduces nonlinear factors to increase the expression ability of the network. Common activation functions include ReLU, Sigmoid, Tanh, etc.; the fully connected layer flattens the extracted features and then connects to the output layer to generate wafer coding features. The FTTransformer encoder uses a residual network to add jump connections on the basis of the original network, which can solve the gradient explosion and gradient disappearance problems caused by the deepening of the network depth. The specific calculation process of the FTTransformer encoder is common knowledge in the art and will not be described in detail here.

[0052] S3. Self-supervised pre-training.

[0053] The labeled wafer samples and the unlabeled wafer samples together constitute a training set, and the training set is used to perform self-supervised mask pre-training on the wafer feature encoder to obtain a pre-trained wafer feature encoder.

[0054] In a specific implementation of the present invention, the mask m=[m1,…md]∈{0, 1} is generated by random sampling by a mask generator. B×D Mask the input sample data, where B is the batch size of the input sample and D is the dimension of the sample data. Figure 2 As shown, and Represent the labeled wafer samples and the unlabeled wafer samples respectively. The samples are masked by the mask generator. Each input sample has only one mask position. Input the wafer feature encoder code to generate the wafer code feature after masking.

[0055] Self-supervised pre-training uses two independent fully connected layers to implement the self-supervised pre-training task, and has an activation function and a dropout layer. The self-supervised pre-training task includes:

[0056] Task 1: Use the masked wafer encoding features to predict the masked position, which is a multi-class classification task. Multi-class cross entropy loss is used to calculate the loss function in the training process, which is recorded as :

[0057]

[0058] in, represents the mask position loss, N represents the number of samples, and D represents the data dimension. The position label representing whether the c-th dimension of the n-th sample is masked or not. Indicates that the cth dimension of the nth sample is masked. Indicates that the cth dimension of the nth sample is not masked; Represents the model's predicted value of the cth dimension of the nth sample after masking.

[0059] Task 2: Use the masked data to predict the value of the masked position, which is a regression task. The root mean square error is used as the loss function of the training process. :

[0060]

[0061] in, represents the mask value loss, Represents the true value of the i-th sample masked by the mask, Represents the predicted value of the i-th sample masked by the mask.

[0062] The total loss of the above self-supervised pre-training is defined as The weight α between the mask position loss and the mask value loss can be set based on the experience of those skilled in the art.

[0063] Through the above self-supervised pre-training task, the wafer feature encoder can effectively learn the relationship between sample data and attention weights, which is helpful for the convergence and prediction performance of the semi-supervised learning process. After the self-supervised pre-training is completed, the weight parameters of the wafer feature encoder are saved.

[0064] S4. Semi-supervised training fine-tuning.

[0065] like Figure 3 As shown in the figure, after the self-supervised pre-training is completed, the sample data in the training set is firstly used to calculate the original wafer coding features using the pre-trained wafer feature encoder, and the original wafer coding feature vector is recorded as , which can effectively extract the mutual relations and global information between sample data.

[0066] Secondly, strong noise and weak noise are added to the original wafer coding features respectively by means of data enhancement, and data enhancement constraints are introduced when adding strong noise to the original wafer coding features, and the data enhancement constraint loss is calculated.

[0067] In a specific implementation of the present invention, the weak noise added to the original wafer coding feature is weak Gaussian noise, and the strong noise is added using a strong noise decoder, specifically:

[0068] The original wafer encoding characteristics of the sample Input the initialized strong noise decoder, and generate strong enhanced noise samples after decoding ;

[0069] Strongly enhance the noise samples Generating Strongly Enhanced Wafer Encoding Features Using Pre-trained Wafer Feature Encoders , to achieve the addition of strong noise to the original wafer coding features;

[0070] The strong noise decoder participates in semi-supervised training together with the pre-trained wafer feature encoder and the wafer failure classifier.

[0071] In order to ensure the stability of strong noise addition, the data enhancement constraint introduced when adding strong noise to the original wafer coding feature using a strong noise decoder is calculated using KL divergence, which can be expressed as:

[0072]

[0073] in, represents the data augmentation constraint loss, represents the original sample i in the training set, represents a strongly enhanced noise sample, represents the square of the L2 norm, represents the dimension of the encoded features generated by the pre-trained wafer feature encoder, They represent the variance and mean of the j-th dimension of the encoded feature generated by the pre-trained wafer feature encoder for the original sample i, represents the preset hyperparameters, Represents the original number of samples in the training set.

[0074] After adding strong noise and weak noise to the original wafer coding features of each sample in the training set, the original wafer coding features, the strongly enhanced wafer coding features after adding strong noise, and the weakly enhanced wafer coding features after adding weak noise are respectively used to predict whether the wafer has failed using a wafer failure classifier. The consistency loss and supervision loss are calculated based on the failure prediction results, wherein the supervision loss requires the use of sample labels, that is, the supervision loss is only for labeled wafer samples in the training set.

[0075] The present invention introduces consistency regularization in the semi-supervised training part, and performs consistency regularization calculation on the wafer failure prediction results of the strong noise data and the weak noise data through the classifier to make the prediction results as consistent as possible. This embodiment uses binary cross entropy loss for calculation, which is recorded as , whose formula is defined as:

[0076]

[0077] in, is the predicted category corresponding to the weakly enhanced wafer encoding feature, is the predicted category corresponding to the strongly enhanced wafer encoding feature, and H is the binary cross entropy loss function.

[0078] In order to make full use of the labels of labeled wafer samples, the present invention introduces supervision loss, that is, the difference between the predicted value of the labeled data and the original label calculation. This embodiment uses binary cross entropy loss for calculation, which is recorded as , and its calculation formula is:

[0079]

[0080] in, is the label category of the labeled wafer sample, is the predicted value output by the model, is the number of labeled wafer samples.

[0081] The data augmentation constraint loss, consistency loss and supervision loss are combined to perform semi-supervised training on the pre-trained wafer feature encoder and wafer failure classifier. The total loss of the above semi-supervised training is defined as L semi =L s +βL c +λL r , where the value of β gradually increases from 0 to the specified value γ during the training process to avoid the instability of the classifier prediction value caused by random weight initialization at the beginning of training.

[0082] S5, predicts whether a wafer fails using the pre-trained wafer feature encoder and wafer failure classifier after semi-supervised training.

[0083] In the process of wafer failure prediction, wafer acceptability test data is collected, wafer coding features are generated using a wafer feature encoder, and then a wafer failure classifier is used to predict whether the wafer has failed.

[0084] In a specific implementation of the present invention, in order to optimize the applicability of the trained wafer feature encoder and wafer failure classifier in the IC industrial environment, the model generated in the initial stage is used to predict new wafers, the prediction results are saved, failure samples are collected to update the training set, and the wafer feature encoder and wafer failure classifier are regularly trained based on self-supervised reinforcement. By providing more labeled data, the continuous expansion of the data set is allowed, and the performance of the model can be sustainably improved.

[0085] like Figure 4 As shown, a complete process is as follows:

[0086] Obtain initial labeled wafer samples and unlabeled wafer samples, train the model using the self-supervised enhanced semi-supervised method and save the model;

[0087] Collect the data to be tested and input it into the model to obtain the wafer failure prediction results. The results can be used by engineers to further analyze the batch of wafers, speed up the root cause analysis of wafer failure and improve the yield rate;

[0088] The online wafer failure prediction results are used to expand the initial training set, and the model is regularly trained with semi-supervised training based on self-supervised reinforcement to continuously optimize the model.

[0089] In order to quantitatively evaluate the performance of the present invention, a real factory data set was used for verification. 128 tested wafers were used as labeled data and more than 1,000 untested wafers were used as unlabeled data. The data were compared with the machine learning methods commonly used in wafer failure prediction. The results are shown in Table 1. The results show that the method proposed in the present invention has achieved the best performance in many indicators, and at the same time improved the performance of the basic encoder FTTransformer, proving the effectiveness of the method of the present invention.

[0090] Table 1 Comparison of the self-supervised enhanced semi-supervised method of the present invention with the existing fully supervised method

[0091]

[0092] In addition, Table 2 also shows the comparison between the proposed method and the most advanced self-supervised and semi-supervised models. The results show that the proposed method outperforms all the most advanced self-supervised and semi-supervised models in the field of tables on industrial datasets.

[0093] Table 2 Comparison of the self-supervised enhanced semi-supervised method of the present invention with existing self-supervised and semi-supervised methods

[0094]

[0095] The present invention also provides a self-supervised enhanced semi-supervised wafer failure prediction system, which is used to implement the above-mentioned embodiments. The terms "module", "unit", etc. used below can implement a combination of software and / or hardware for predetermined functions. Although the system described in the following embodiments is preferably implemented in software, it is also possible to implement hardware, or a combination of software and hardware.

[0096] This embodiment provides a self-supervised enhanced semi-supervised wafer failure prediction system, including:

[0097] A data acquisition module, which is used to obtain wafer acceptability test data and probe test data in the training phase, preprocess and match the two parts of data, and obtain labeled wafer samples and unlabeled wafer samples as training sets; and, used to collect wafer acceptability test data to be predicted in the prediction phase;

[0098] A self-supervised training module, which is used to perform self-supervised mask pre-training on the wafer feature encoder using the training set to obtain a pre-trained wafer feature encoder;

[0099] A semi-supervised training module is used to generate original wafer coding features of each sample in a training set using a pre-trained wafer feature encoder, and to add strong noise and weak noise to the original wafer coding features respectively; a data enhancement constraint is introduced when adding strong noise to the original wafer coding features, and a data enhancement constraint loss is calculated;

[0100] The original wafer coding features, the strongly enhanced wafer coding features after adding strong noise, and the weakly enhanced wafer coding features after adding weak noise are respectively used by the wafer failure classifier to predict whether the wafer fails, and the consistency loss and the supervision loss are calculated according to the failure prediction results; the supervision loss is only for the labeled wafer samples in the training set;

[0101] Semi-supervised training of the pre-trained wafer feature encoder and wafer failure classifier is performed by combining the data augmentation constraint loss, consistency loss and supervision loss;

[0102] The wafer failure prediction module is used to predict whether a wafer fails by using a pre-trained wafer feature encoder and a wafer failure classifier after semi-supervised training.

[0103] For the system embodiment, since it basically corresponds to the method embodiment, the relevant parts can be referred to the partial description of the method embodiment, and the implementation method of the module will not be repeated here. The system embodiment described above is only schematic, wherein the units described as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units, that is, they may be located in one place, or they may be distributed on multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the scheme of the present invention. Ordinary technicians in this field can understand and implement it without paying creative labor.

[0104] The embodiments of the system of the present invention can be applied to any device with data processing capabilities, and the device with data processing capabilities can be a device or apparatus such as a computer. The system embodiments can be implemented by software, or by hardware or a combination of software and hardware. Taking software implementation as an example, as a device in a logical sense, the corresponding computer program instructions in the non-volatile memory are read into the memory by the processor of any device with data processing capabilities and run.

[0105] Obviously, the above-described embodiments and drawings are only some examples of the present application. For those of ordinary skill in the art, the present application can also be applied to other similar situations based on these drawings without creative work. In addition, it is understandable that, although the work done in this development process may be complicated and lengthy, for those of ordinary skill in the art, certain changes in design, manufacturing or production based on the technical content disclosed in this application are only conventional technical means and should not be regarded as insufficient content disclosed in this application. Without departing from the concept of the present application, several variations and improvements can also be made, which all belong to the scope of protection of the present application. Therefore, the scope of protection of the present application shall be subject to the attached claims.

Claims

1. A self-supervised enhanced semi-supervised wafer failure prediction method, characterized in that: include: Obtain wafer acceptability test data and probe test data, preprocess and match the data, and obtain labeled wafer samples and unlabeled wafer samples as training sets; Using the training set to perform self-supervised mask pre-training on the wafer feature encoder to obtain a pre-trained wafer feature encoder; The pre-trained wafer feature encoder is used to generate the original wafer coding features of each sample in the training set, and strong noise and weak noise are added to the original wafer coding features respectively; when adding strong noise to the original wafer coding features, a data enhancement constraint is introduced, and the data enhancement constraint loss is calculated; The original wafer coding features, the strongly enhanced wafer coding features after adding strong noise, and the weakly enhanced wafer coding features after adding weak noise are respectively used by the wafer failure classifier to predict whether the wafer fails, and the consistency loss and the supervision loss are calculated according to the failure prediction results; the supervision loss is only for the labeled wafer samples in the training set; The data enhancement constraint loss, consistency loss and supervision loss are combined to perform semi-supervised training on the pre-trained wafer feature encoder and wafer failure classifier, and the pre-trained wafer feature encoder and wafer failure classifier after the semi-supervised training are used to predict whether the wafer fails.

2. The self-supervised enhanced semi-supervised wafer failure prediction method according to claim 1, characterized in that: The preprocessing and matching of data includes: Handle outliers and missing values ​​on wafer acceptability test data; Match the pre-processed wafer acceptability test data and probe test data, use a group of successfully matched data as labeled wafer samples, use unsuccessfully matched wafer acceptability test data as unlabeled wafer samples, and discard unsuccessfully matched probe test data; The labels of the labeled wafer samples are the results of probe test data, including failure and normal.

3. The self-supervised enhanced semi-supervised wafer failure prediction method according to claim 1, characterized in that: The method of using the training set to perform self-supervisory mask pre-training on the wafer feature encoder includes: Randomly mask the samples in the training set; The masked samples use the wafer feature encoder to generate wafer coding features, and the mask position and mask value are predicted based on the wafer coding features. The wafer feature encoder parameters are iteratively updated according to the prediction results to complete self-supervised pre-training.

4. The self-supervised enhanced semi-supervised wafer failure prediction method according to claim 1, characterized in that: The weak noise added to the original wafer coding feature is weak Gaussian noise.

5. The self-supervised enhanced semi-supervised wafer failure prediction method according to claim 1, characterized in that: Methods to add strong noise to the original wafer code features include: Inputting the original wafer coding feature of the sample into the initialized strong noise decoder, and generating a strong enhanced noise sample after decoding; The strongly enhanced noise samples are used to generate strongly enhanced wafer coding features using a pre-trained wafer feature encoder, thereby adding strong noise to the original wafer coding features; The strong noise decoder participates in semi-supervised training together with the pre-trained wafer feature encoder and the wafer failure classifier.

6. The self-supervised enhanced semi-supervised wafer failure prediction method according to claim 5, characterized in that: The data enhancement constraint loss is as follows: ； in, represents the data augmentation constraint loss, represents the original sample i in the training set, represents a strongly enhanced noise sample, represents the square of the L2 norm, represents the dimension of the encoded features generated by the pre-trained wafer feature encoder, They represent the variance and mean of the j-th dimension of the encoded feature generated by the pre-trained wafer feature encoder for the original sample i, represents the preset hyperparameters, Represents the original number of samples in the training set.

7. The self-supervised enhanced semi-supervised wafer failure prediction method according to claim 1, characterized in that: The supervised loss refers to the binary cross entropy loss between the original wafer coding features using the prediction results of the wafer failure classifier and the true label; the consistency loss refers to the binary cross entropy loss between the prediction results of the strongly enhanced wafer coding features and the weakly enhanced wafer coding features using the wafer failure classifier respectively.

8. The self-supervised enhanced semi-supervised wafer failure prediction method according to claim 1, characterized in that: When the data enhancement constraint loss, consistency loss and supervision loss are combined to perform semi-supervised training on the pre-trained wafer feature encoder and wafer failure classifier, the weight of the consistency loss is gradually increased from 0 to a preset weight.

9. The self-supervised enhanced semi-supervised wafer failure prediction method according to claim 1, characterized in that: Also includes: After predicting whether a wafer has failed using the wafer feature encoder and the wafer failure classifier, the prediction results are saved, failure samples are collected to update the training set, and semi-supervised training based on self-supervised reinforcement is performed regularly on the wafer feature encoder and the wafer failure classifier.

10. A self-supervised enhanced semi-supervised wafer failure prediction system, used to implement the self-supervised enhanced semi-supervised wafer failure prediction method according to any one of claims 1 to 9, characterized in that: The wafer failure prediction system comprises: A data acquisition module, which is used to obtain wafer acceptability test data and probe test data in the training phase, preprocess and match the two parts of data, and obtain labeled wafer samples and unlabeled wafer samples as training sets; and, used to collect wafer acceptability test data to be predicted in the prediction phase; A self-supervised training module, which is used to perform self-supervised mask pre-training on the wafer feature encoder using the training set to obtain a pre-trained wafer feature encoder; A semi-supervised training module is used to generate original wafer coding features of each sample in a training set using a pre-trained wafer feature encoder, and to add strong noise and weak noise to the original wafer coding features respectively; a data enhancement constraint is introduced when adding strong noise to the original wafer coding features, and a data enhancement constraint loss is calculated; The original wafer coding features, the strongly enhanced wafer coding features after adding strong noise, and the weakly enhanced wafer coding features after adding weak noise are respectively used by the wafer failure classifier to predict whether the wafer fails, and the consistency loss and the supervision loss are calculated according to the failure prediction results; the supervision loss is only for the labeled wafer samples in the training set; Semi-supervised training of the pre-trained wafer feature encoder and wafer failure classifier is performed by combining the data augmentation constraint loss, consistency loss and supervision loss; The wafer failure prediction module is used to predict whether a wafer fails by using a pre-trained wafer feature encoder and a wafer failure classifier after semi-supervised training.

Images