Wafer test parameter optimization method and system

Abstract

The application provides a wafer test parameter optimization method and system, wherein all test parameters are divided into multiple to-be-optimized parameters and neighborhood parameters of each to-be-optimized parameter through a predicted value of a next round of wafer test indicators, then a difference gradient of parameter values between each to-be-optimized parameter and its neighborhood parameter is determined, then a neighborhood space is obtained through mapping, neighborhood aggregate features of all to-be-optimized parameters are determined through the neighborhood space, a parameter optimization model based on a gradient boosting tree algorithm is constructed according to the obtained adaptive parameters and the predicted value, an optimization confidence of the parameter optimization model is determined according to the neighborhood aggregate features, and then the parameter optimization model performs adaptive optimization on each to-be-optimized parameter based on the optimization confidence. By using the scheme of the application, the test parameters of the crystal grains in the wafer are optimized through the establishment of a computer-aided optimization model, so as to reduce the escape rate of wafer testing.

Description

Technical Field

[0001] This application relates to the field of wafer computer-aided testing technology, and more specifically, to a method and system for optimizing wafer testing parameters. Background Technology

[0002] Wafer testing refers to the process of testing the electrical performance and physical characteristics of each die in a semiconductor wafer. The purpose of the testing is to evaluate the quality and reliability of the wafer to ensure that the wafer performs well in manufacturing and practical applications. The parameters in the wafer testing process include current, voltage, frequency, etc.

[0003] In semiconductor manufacturing, optimizing wafer testing parameters is a crucial step in ensuring that die performance meets standards. Machine learning can analyze historical test data to identify complex patterns and relationships, thereby achieving more accurate parameter optimization. Traditional machine learning optimization methods typically use statistical methods to adjust test parameters based on historical test data. While this ensures the stability of test results to some extent, traditional machine learning statistical methods cannot statistically analyze the differences in test parameters corresponding to each die in the wafer, leading to serious test escape and making it impossible to accurately determine whether the wafer is qualified. Therefore, how to optimize the test parameters of the dies in the wafer by establishing a computer-aided optimization model to reduce the wafer test escape rate has become a challenge for the industry. Summary of the Invention

[0004] This application provides a method and system for optimizing wafer testing parameters. By establishing a computer-aided optimization model, the testing parameters of the dies in the wafer can be optimized to reduce the escape rate of wafer testing.

[0005] In a first aspect, this application provides a method for optimizing wafer testing parameters, comprising the following steps: Collect test parameters for all dies during wafer testing; Based on the predicted values ​​of the next round of wafer testing indicators, all testing parameters are divided into multiple parameters to be optimized and neighborhood parameters of each parameter to be optimized, thereby determining the gradient of the difference in parameter values ​​between each parameter to be optimized and its neighborhood parameters. A pre-defined spatial mapping network synchronously maps all the parameters to be optimized, all the neighborhood parameters, and all the differential gradients to the neighborhood space. Then, based on the mapping vector of each differential gradient in the neighborhood space, the neighborhood aggregation features of all the parameters to be optimized are determined. Perform the next round of wafer testing, obtain the adaptive parameters of the corresponding dies for each parameter to be optimized, construct a parameter optimization model based on gradient boosting tree algorithm based on all adaptive parameters and the predicted values, and determine the optimization confidence of the parameter optimization model by the neighborhood aggregation feature; The parameter optimization model adaptively optimizes each parameter to be optimized based on the optimization confidence level.

[0006] In some embodiments, based on the predicted values ​​of the wafer testing metrics for the next round, all testing parameters are divided into multiple parameters to be optimized and neighborhood parameters of each parameter to be optimized. Specifically, this includes: Determine the predicted values ​​for the next round of wafer testing metrics; Based on the predicted values, all test parameters that need to be optimized are selected from all test parameters, thus obtaining multiple parameters to be optimized; The neighborhood parameters of each parameter to be optimized are determined by the distribution characteristics of the grains in the wafer.

[0007] In some embodiments, dividing the neighborhood parameters of each parameter to be optimized based on the distribution characteristics of the grains in the wafer specifically includes: Obtain the distribution characteristics of all grains in the wafer; The neighborhood range of each parameter to be optimized is determined based on the distribution characteristics. The neighborhood parameters of each parameter to be optimized are determined based on the neighborhood range of each parameter to be optimized.

[0008] In some embodiments, determining the gradient of the difference between the parameter values ​​of each parameter to be optimized and its neighboring parameters specifically includes: For each parameter to be optimized, determine the difference between the parameter to be optimized and each test parameter in the corresponding neighborhood parameters; By determining the difference gradient between the parameter to be optimized and its corresponding neighboring parameters through all the difference values, the difference gradient between the parameter to be optimized and its corresponding neighboring parameters can be obtained.

[0009] In some embodiments, the pre-defined spatial mapping network synchronously maps all parameters to be optimized, all neighborhood parameters, and all differential gradients to the neighborhood space, specifically including: Obtain the distribution characteristics of all grains in the wafer; Construct a spatial mapping network based on Gaussian process regression according to the aforementioned distribution characteristics; The spatial mapping network is used to amplify all the parameters to be optimized and all the neighborhood parameters to obtain amplified parameters to be optimized and amplified neighborhood parameters. The spatial mapping network constructs a neighborhood space based on all magnified parameters to be optimized, all magnified neighborhood parameters, and all differential gradients.

[0010] In some embodiments, determining the optimization confidence of the parameter optimization model based on the neighborhood aggregation features specifically includes: Obtain the feature set input to the parameter optimization model; In the parameter optimization model, the feature importance of the feature set is evaluated through the neighborhood aggregation feature to obtain the feature importance of all adaptive parameters in the feature set; Based on the importance of all features, determine the attribution value and Shapley value of each adaptive parameter in the parameter optimization model; Weighted analysis is performed on all attribution values ​​and all Shapley values ​​to obtain the optimization variables of the parameter optimization model. Then, confidence analysis is performed on the optimization variables to obtain the optimization confidence level of the parameter optimization model.

[0011] In some embodiments, the parameter optimization model adaptively optimizes each parameter to be optimized based on the optimization confidence level, specifically including: Obtain the optimization variables of the parameter optimization model; If the optimization confidence level is greater than or equal to the preset confidence threshold, then for each parameter to be optimized, the parameter to be optimized is adjusted through the optimization variable to obtain the optimized test parameter, and then the optimized test parameters of all parameters to be optimized are obtained, thus completing the optimization of the wafer test parameters.

[0012] Secondly, this application provides a wafer testing parameter optimization system, comprising: The acquisition module is used to acquire the test parameters of all dies during the wafer testing process; The processing module is used to divide all test parameters into multiple parameters to be optimized and neighborhood parameters of each parameter to be optimized based on the predicted values ​​of the wafer test indicators for the next round, and then determine the gradient of the difference in parameter values ​​between each parameter to be optimized and its neighborhood parameters. The processing module is also used to synchronously map all the parameters to be optimized, all the neighborhood parameters, and all the differential gradients to the neighborhood space using a preset spatial mapping network, and then determine the neighborhood aggregation features of all the parameters to be optimized based on the mapping vectors of each differential gradient in the neighborhood space. The processing module is also used to execute the next round of wafer testing operations, obtain the adaptive parameters of the corresponding dies for each parameter to be optimized, construct a parameter optimization model based on gradient boosting tree algorithm based on all adaptive parameters and the predicted values, and determine the optimization confidence of the parameter optimization model by the neighborhood aggregation feature; An execution module is used for the parameter optimization model to adaptively optimize each parameter to be optimized based on the optimization confidence level.

[0013] Thirdly, this application provides a computer device, the computer device including a memory and a processor, the memory storing code, and the processor being configured to acquire the code and execute the above-described wafer test parameter optimization method.

[0014] Fourthly, this application provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the above-described wafer test parameter optimization method.

[0015] The technical solutions provided by the embodiments disclosed in this application have the following beneficial effects: The wafer testing parameter optimization method and system provided in this application, In this embodiment, firstly, test parameters of all dies during wafer testing are collected; based on the predicted values ​​of the next round of wafer testing indicators, all test parameters are divided into multiple parameters to be optimized and neighborhood parameters of each parameter to be optimized, thereby determining the difference gradient between the parameter values ​​of each parameter to be optimized and its neighborhood parameters; a preset spatial mapping network synchronously maps all parameters to be optimized, all neighborhood parameters, and all difference gradients to the neighborhood space, and then determines the neighborhood aggregation features of all parameters to be optimized based on the mapping vectors of each difference gradient in the neighborhood space; the next round of wafer testing is executed to obtain the adaptive parameters of the dies corresponding to each parameter to be optimized, and a parameter optimization model based on the gradient boosting tree algorithm is constructed based on all adaptive parameters and the predicted values, and the optimization confidence of the parameter optimization model is determined by the neighborhood aggregation features; the parameter optimization model adaptively optimizes each parameter to be optimized based on the optimization confidence.

[0016] Therefore, this application first divides test parameters into parameters to be optimized and neighborhood parameters based on the predicted values ​​of test indicators, thereby obtaining neighborhood aggregation features. This is beneficial for locating faulty dies and reducing test escape. When performing the next round of testing on faulty dies, the neighborhood aggregation features can also serve as a reference for parameter optimization to improve the adaptability of test parameter optimization. Second, performing the next round of testing and obtaining adaptive parameters allows for real-time feedback on the changes of each parameter to be optimized during the next round of testing. This enables the parameter optimization model to adjust its optimization strategy according to changes in test indicators, thereby effectively reducing the escape rate. Then, constructing a parameter optimization model based on adaptive parameters and predicted values ​​helps improve the accuracy and adaptability of the parameter optimization model, thereby further reducing the test escape rate. Finally, the optimization confidence of the parameter optimization model is determined through neighborhood aggregation features, and then adaptive optimization is performed on each parameter to be optimized based on the optimization confidence. This not only improves the accuracy of testing but also facilitates the identification and correction of test parameter deviations, reducing the escape rate. In summary, the solution of this application can optimize the test parameters of dies in wafers, thereby reducing the escape rate of wafer testing. Attached Figure Description

[0017] Figure 1 This is an exemplary flowchart of a wafer testing parameter optimization method according to some embodiments of this application; Figure 2 This is a schematic flowchart illustrating the process of determining the difference gradient according to some embodiments of this application; Figure 3 This is a flowchart illustrating the process of determining neighborhood aggregation features according to some embodiments of this application; Figure 4 This is a schematic diagram of the structure of a wafer testing parameter optimization system according to some embodiments of this application; Figure 5 This is a schematic diagram of the structure of a computer device for implementing a wafer testing parameter optimization method according to some embodiments of this application. Detailed Implementation

[0018] To better understand the technical solution of this application, the technical solution of this application will be described in detail below with reference to the accompanying drawings and specific embodiments.

[0019] refer to Figure 1 The figure is an exemplary flowchart of a wafer test parameter optimization method according to some embodiments of this application. The wafer test parameter optimization method 100 mainly includes the following steps: In step 101, test parameters for all dies are collected during the wafer testing process.

[0020] It should be noted that the test parameters in this application refer to the comprehensive value of historical test parameters of the dies in the wafer from the start of wafer testing to the current testing stage. The test parameters may be static leakage current, reverse breakdown voltage, output voltage, or other electrical performance parameters.

[0021] In specific implementation, the acquisition of test parameters for all dies during wafer testing can be achieved in the following way: for each die in the wafer, the testing of that die is monitored, and historical test parameters for all testing stages from the start of wafer testing to the current moment are automatically collected. Then, the comprehensive value obtained by analyzing all historical test parameters is used as the test parameter for that die, thereby obtaining the test parameters for each die in the wafer. As a preferred embodiment, the comprehensive value can be obtained by analyzing all the tested test parameters using an autoregressive analysis model in the prior art. Other existing models can also be used for analysis in other embodiments, which will not be elaborated here. It should also be noted that the wafer testing in this application includes multiple testing stages.

[0022] In step 102, based on the predicted values ​​of the wafer testing indicators for the next round, all testing parameters are divided into multiple parameters to be optimized and neighborhood parameters of each parameter to be optimized, thereby determining the gradient of the difference in parameter values ​​between each parameter to be optimized and its neighborhood parameters.

[0023] In some embodiments, dividing all test parameters into multiple parameters to be optimized and neighborhood parameters of each parameter to be optimized based on the predicted values ​​of the wafer test indicators for the next round can be achieved by the following steps; Determine the predicted values ​​for the next round of wafer testing metrics; Based on the predicted values, all test parameters that need to be optimized are selected from all test parameters, thus obtaining multiple parameters to be optimized; The neighborhood parameters of each parameter to be optimized are determined by the distribution characteristics of the grains in the wafer.

[0024] In specific implementation, the test parameters that need to be optimized refer to the test parameters whose parameter values ​​have not reached the predicted values. Therefore, based on the predicted values, all test parameters that need to be optimized are selected from all test parameters, thereby obtaining multiple parameters to be optimized. That is, the parameter values ​​of all test parameters are compared with the test values, and all test parameters whose parameter values ​​are less than the predicted values ​​are taken as parameters to be optimized, thereby obtaining multiple parameters to be optimized.

[0025] The distribution characteristics can be obtained during wafer fabrication. In some embodiments, the neighborhood parameters of each parameter to be optimized can be divided based on the distribution characteristics of the grains in the wafer in the following manner: Obtain the distribution characteristics of all grains in the wafer; The neighborhood range of each parameter to be optimized is determined based on the distribution characteristics. The neighborhood parameters of each parameter to be optimized are determined based on the neighborhood range of each parameter to be optimized.

[0026] It should be noted that in this application, one grain corresponds to one test parameter, and one parameter to be optimized corresponds to one test parameter. Therefore, one parameter to be optimized corresponds to one grain.

[0027] In specific implementation, the distribution characteristics include the spatial coordinates of each grain in the wafer and the regional density of the grains in the wafer. Obtaining the distribution characteristics of all grains in the wafer involves using existing techniques such as grain boundary density measurement, grain size measurement, and electron backscattering diffraction to acquire the spatial coordinates and regional density of each grain in the wafer. These acquired spatial coordinates and regional density are then used as the distribution characteristics of all grains in the wafer. Based on these distribution characteristics, the neighborhood range of each parameter to be optimized is determined, i.e., the spatial coordinates of each grain in the wafer are... The region density of the dies in the wafer is input into the visual segmentation algorithm in the existing machine learning algorithm to obtain the neighborhood range of the die corresponding to each parameter to be optimized in the wafer, wherein the region density and the neighborhood range are linearly related; based on the neighborhood range of each parameter to be optimized, the neighborhood parameters of each parameter to be optimized are determined, that is: for each die corresponding to each parameter to be optimized, the neighborhood range of the die is divided into multiple neighboring dies by the visual segmentation algorithm, and then the test parameters of all neighboring dies are used as the neighborhood parameters of the parameter to be optimized, thereby obtaining the neighborhood parameters of each parameter to be optimized.

[0028] Preferably, in some embodiments, determining the predicted values ​​of the wafer testing metrics for the next round can be achieved using the following steps: Construct an indicator prediction model based on a BP neural network; The predicted values ​​of the next round of wafer testing indicators are obtained by analyzing the trained indicator prediction model based on BP neural network.

[0029] It should be noted that the construction of the index prediction model based on BP neural network in this application specifically includes: data preparation, design of BP neural network architecture based on backpropagation algorithm, model training, model evaluation, and model deployment. The BP neural network architecture includes: input layer, hidden layer, output layer, and activation function of each network layer. Using the index prediction model based on BP neural network can improve the confidence of the predicted value.

[0030] In specific implementation, the following steps can be taken to obtain the predicted values ​​of the next round of wafer testing indicators using a trained BP neural network-based indicator prediction model: First, collect a set of historical wafer testing indicators and divide the set into a training dataset and a validation dataset; second, train and validate the BP neural network-based indicator prediction model using the training dataset and validation dataset respectively, and obtain the confidence level and prediction deviation value of the indicator prediction model through model validation; then, analyze and obtain the first predicted value of the next round of wafer testing indicators using the trained BP neural network-based indicator prediction model; finally, obtain the prediction deviation value of the indicator prediction model, and determine the predicted value of the next round of wafer testing indicators using the prediction deviation value and the first predicted value.

[0031] It should be noted that the first predicted value is the preliminary predicted value of the wafer testing index for the next round. The predicted value of the wafer testing index for the next round can be obtained by combining the first predicted value with the deviation of the index prediction model. That is, as a preferred embodiment, the predicted value of the wafer testing index for the next round can be determined by the following formula through the prediction deviation value and the first predicted value: Predicted value = N * first predicted value - (1-N) * prediction deviation value, where N is the confidence level of the trained index prediction model based on BP neural network, and the confidence level is between 0.6 and 0.9.

[0032] In some embodiments, reference Figure 2 As shown in the figure, this is a flowchart illustrating the process of determining the difference gradient in some embodiments of this application. In this embodiment, the difference gradient between the parameter value to be optimized and its neighboring parameters can be determined by the following steps: In step 1021, for each parameter to be optimized, the difference value between the parameter to be optimized and each test parameter in the corresponding neighborhood parameters is determined; In step 1022, the difference gradient between the parameter value to be optimized and the corresponding neighboring parameter is determined by all the difference values, thereby obtaining the difference gradient between the parameter value to be optimized and the corresponding neighboring parameter.

[0033] In addition, in specific implementation, the difference between the parameter to be optimized and the test parameters in the corresponding neighborhood parameters can be determined in the following way: for each test parameter in the corresponding neighborhood parameters, the difference between the parameter value of the parameter to be optimized and the test parameter is taken as the difference between the parameter to be optimized and the test parameter, and then the difference between the parameter to be optimized and the test parameters in the corresponding neighborhood parameters is calculated.

[0034] It should be noted that the difference gradient in this application measures the degree of difference between the parameter value to be optimized and the corresponding neighboring parameter. The larger the difference gradient, the greater the difference between the parameter value to be optimized and the corresponding neighboring parameter; the smaller the difference gradient, the smaller the difference between the parameter value to be optimized and the corresponding neighboring parameter. The difference gradient allows for real-time adjustment of the optimization strategy for test parameters to reduce the escape rate of wafer testing. As a preferred embodiment, the difference gradient between the parameter value to be optimized and the corresponding neighboring parameter can be determined by the following method: First, sort all the difference values ​​in ascending order; then, calculate the difference between every two adjacent difference values; finally, use the average of all calculated differences as the difference gradient between the parameter value to be optimized and the corresponding neighboring parameter. Other methods can be used to determine the difference gradient in other embodiments, which are not limited here.

[0035] In step 103, a preset spatial mapping network synchronously maps all the parameters to be optimized, all the neighborhood parameters, and all the differential gradients to the neighborhood space, and then determines the neighborhood aggregation features of all the parameters to be optimized based on the mapping vectors of each differential gradient in the neighborhood space.

[0036] In some embodiments, the synchronous mapping of all parameters to be optimized, all neighborhood parameters, and all differential gradients to the neighborhood space by a preset spatial mapping network can be achieved through the following steps: Obtain the distribution characteristics of all grains in the wafer; Construct a spatial mapping network based on Gaussian process regression according to the aforementioned distribution characteristics; The spatial mapping network is used to amplify all the parameters to be optimized and all the neighborhood parameters to obtain amplified parameters to be optimized and amplified neighborhood parameters. The spatial mapping network constructs a neighborhood space based on all magnified parameters to be optimized, all magnified neighborhood parameters, and all differential gradients.

[0037] Specifically, in implementation, constructing a spatial mapping network based on Gaussian process regression according to the distribution characteristics can be achieved in the following way: A Gaussian process regression model is established, and the regional density of the grains in the distribution characteristics is used to constrain the Gaussian process regression model (i.e., the regional density is used as a constraint on the Gaussian distribution function in the Gaussian process regression model), resulting in a spatial mapping network. The spatial mapping network includes an input layer, a hidden layer, an output layer, and a Gaussian distribution function. Amplifying all the parameters to be optimized and all the neighborhood parameters through the spatial mapping network can be achieved in the following way: In the input layer of the spatial mapping network, the output weight of the Gaussian distribution function is amplified by the regional density of the grains in the wafer, and all the parameters to be optimized and all the neighborhood parameters are input into the input layer of the spatial mapping network for amplification. All the parameters to be optimized and all the neighborhood parameters are then forward-propagated through the spatial mapping network, and the amplified parameters to be optimized and all the neighborhood parameters are obtained through the output layer of the spatial mapping network. Other methods can also be used in other embodiments, which are not limited here.

[0038] In addition, in a specific implementation, the neighborhood space is a magnified scatter plot of information of all parameters to be optimized and all neighboring parameters. The neighborhood space can amplify the differences between the parameters to be optimized and the neighboring parameters, which facilitates subsequent processing. As a preferred embodiment, the spatial mapping network can construct the neighborhood space based on all magnified parameters to be optimized, all magnified neighboring parameters, and all differential gradients in the following way: the distribution weights of the spatial mapping network can be set according to the proportion of each differential gradient, and the positions of all parameters to be optimized and all neighboring parameters in the spatial mapping network can be adjusted by setting the distribution weights, thereby drawing an information scatter plot, and using the drawn information scatter plot as the neighborhood space. In other embodiments, other methods can be used to determine the neighborhood space, which is not limited here.

[0039] In some embodiments, reference Figure 3 As shown in the figure, this is a flowchart illustrating the process of determining neighborhood aggregation features according to some embodiments of this application. In this embodiment, determining the neighborhood aggregation features of all parameters to be optimized based on the mapping vectors of each difference gradient in the neighborhood space can be achieved by the following steps: In step 1031, all differential gradients are plotted as gradient change curves; In step 1032, the mapping vector of each differential gradient in the neighborhood space is determined based on the gradient change curve; In step 1033, the distance between different parameters to be optimized in the neighborhood space is obtained; In step 1034, neighborhood aggregation features for all parameters to be optimized are determined based on all distances and all mapping vectors.

[0040] Specifically, plotting all differential gradients as gradient change curves can be achieved in the following way: using the existing Matplotlib library, the curves obtained from plotting all differential gradients can be used as gradient change curves. These curves visually represent all differential gradients. Furthermore, in the neighborhood space, the differential gradients amplify with the amplification of the parameters to be optimized and the neighborhood parameters, which is beneficial for real-time monitoring of the differences in test parameters of the wafer during testing. This facilitates the screening of faulty wafers, thereby improving the optimization of faulty wafers and reducing test escape. Determining the mapping vector of each differential gradient in the neighborhood space based on the gradient change curves can be achieved in the following way: using the `gradient()` function in the Matplotlib library to obtain the tangent and slope of each differential gradient in the gradient change curves, and then determining the direction of the mapping vector of each differential gradient in the neighborhood space based on the obtained tangent and slope. The value corresponding to each differential gradient is then used as the value of the mapping vector of the corresponding differential gradient in the neighborhood space, thus obtaining the mapping vector of each differential gradient in the neighborhood space.

[0041] In addition, in specific implementation, the distance between different parameters to be optimized in the neighborhood space is obtained. For example, the Euclidean distance between each parameter to be optimized in the neighborhood space can be used as the distance between two corresponding parameters to be optimized, thereby obtaining the distance between different parameters to be optimized. Other methods can also be used to obtain the distance in other embodiments, which will not be elaborated here.

[0042] In addition, in a specific implementation, determining the neighborhood aggregation features of all parameters to be optimized based on all distances and all mapping vectors can be achieved in the following way: For every two parameters to be optimized in the neighborhood space, firstly, the distance between the two parameters to be optimized is obtained; secondly, the sum of the absolute values ​​of the adjacent mappings of the two parameters to be optimized is calculated; then, the product of the distance between the two parameters to be optimized and the calculated sum is used as the aggregation feature between the two parameters to be optimized, thereby obtaining the aggregation feature between every two parameters to be optimized in the neighborhood space; finally, the variance of all aggregation features is used as the neighborhood aggregation feature of all parameters to be optimized. In other embodiments, other methods can also be used to determine the neighborhood aggregation features of all parameters to be optimized, which are not limited here.

[0043] It should be noted that the neighborhood aggregation feature in this application is an index value that measures the degree of aggregation between different parameters to be optimized. In addition, the neighborhood aggregation feature can also inversely represent the degree of difference between different parameters to be optimized. That is, the larger the neighborhood aggregation feature, the smaller the difference between the parameters to be optimized, and the smaller the neighborhood aggregation feature, the larger the difference between the parameters to be optimized. By identifying the changes in the difference between test parameters through the neighborhood aggregation feature, it is helpful to locate faulty dies. When performing the next round of testing on faulty dies, the neighborhood aggregation feature can be used as a reference for parameter optimization to optimize the accuracy of test parameters in wafer testing.

[0044] In step 104, the next round of wafer testing is performed to obtain the adaptive parameters of the corresponding dies for each parameter to be optimized. Based on all the adaptive parameters and the predicted values, a parameter optimization model based on the gradient boosting tree algorithm is constructed, and the optimization confidence of the parameter optimization model is determined by the neighborhood aggregation feature.

[0045] It should be noted that performing the next round of wafer testing refers to conducting adaptive testing on the wafer. Obtaining the adaptive parameters of each die corresponding to the parameters to be optimized means using the collected test parameters as the adaptive parameters of the dies in the wafer during the adaptive testing. Adaptive testing is a testing process that optimizes subsequent test conditions by analyzing historical test data, thereby improving the accuracy and efficiency of testing. In this application, adaptive testing is used in the second round of wafer testing to ensure test quality and reduce test costs.

[0046] In some embodiments, constructing a parameter optimization model based on the gradient boosting tree algorithm according to all the fitness parameters and the predicted values ​​can be achieved by the following steps: All fitness parameters are normalized, and the normalized fitness test parameters are used as the feature set. The predicted value is used as an evaluation metric, and a parameter optimization model based on the gradient boosting tree algorithm is initialized using the evaluation metric and the feature set.

[0047] It should be noted that in this embodiment, normalizing all adaptive parameters is to ensure that different features of the input model are on the same scale. Normalization (Z-score) can be used to normalize the parameter values ​​of all adaptive parameters, and all normalized adaptive test parameters can be used as a feature set. This can reduce the scale difference between different features of the input model, thereby improving the stability and efficiency of model training.

[0048] In specific implementation, the predicted value is used as the evaluation metric, and a parameter optimization model based on the gradient boosting tree algorithm is initialized using the evaluation metric and the feature set. This can be achieved in the following way: First, a basic decision tree model is created, and the loss function of the decision tree model is established using mean squared error. Second, the evaluation metric and the feature set are divided into training sets, and the decision tree model is trained using the training sets. Then, iterative training is performed, and new trees are added in each iteration based on the gradient boosting tree algorithm. Finally, during the iteration process, the loss function value of the decision tree model is continuously verified using the partitioned validation set. When the loss function value is minimized, the trained model is used as the parameter optimization model based on the gradient boosting tree algorithm.

[0049] It should be noted that the parameter optimization model in this application is a machine learning model used for optimizing test parameters during wafer testing. The test parameters of the dies in the wafer can be adjusted through this parameter optimization model. This parameter optimization model can identify and correct deviations between test parameters by analyzing the relationship between historical test parameters and current test parameters of wafer testing, so as to improve the accuracy of testing and reduce test escape.

[0050] In some embodiments, determining the optimization confidence of the parameter optimization model based on the neighborhood aggregation features can be achieved through the following steps: Obtain the feature set input to the parameter optimization model; In the parameter optimization model, the feature importance of the feature set is evaluated through the neighborhood aggregation feature to obtain the feature importance of all adaptive parameters in the feature set; Based on the importance of all features, determine the attribution value and Shapley value of each adaptive parameter in the parameter optimization model; Weighted analysis is performed on all attribution values ​​and all Shapley values ​​to obtain the optimization variables of the parameter optimization model. Then, confidence analysis is performed on the optimization variables to obtain the optimization confidence level of the parameter optimization model.

[0051] In specific implementation, the feature importance of the feature set is evaluated using the neighborhood aggregation feature. This evaluation, which assesses the feature importance of all adaptive parameters in the feature set, can be achieved as follows: In the parameter optimization model, for each branch node of the gradient boosting tree (each branch node corresponds to an adaptive parameter in the feature set), the difference between the Gini index of the branch node before iteration and the Gini index of the branch node after iteration is used as the weight for evaluating the feature importance of the feature set. Then, the neighborhood aggregation feature is introduced. This neighborhood aggregation feature is an indicator that measures the degree of aggregation of different parameters to be optimized within the neighborhood set. Therefore, the neighborhood... The product of the aggregated feature and the weight is used as the feature importance of the corresponding feature value of the branch node, thereby obtaining the feature importance of all adaptive parameters in the feature set. The Gini index is an indicator that measures the importance of branch features in the corresponding model. In the parameter optimization model, the smaller the Gini index, the more concentrated and less differentiated the feature values ​​of the branch nodes are. Furthermore, the feature importance is an indicator that measures the predictive role of the feature values ​​of the branch nodes in the parameter optimization model. Based on the ranking of feature importance, redundant node features can be eliminated to reduce overfitting in the parameter optimization model and improve its optimization capability.

[0052] Furthermore, in specific implementation, determining the attribution value and Shapley value of each adaptive parameter in the parameter optimization model based on the importance of all features can be achieved in the following way: In the parameter optimization model, for each adaptive parameter, the feature importance of the adaptive parameter is quantified into an attribution value through the feature importance evaluation mechanism in the gradient boosting tree algorithm, thereby obtaining the attribution value of each adaptive parameter in the parameter optimization model. Here, the attribution value refers to the specific contribution of the features of the branch node to the output of the tree model, which can be quantified by feature importance; in addition, the Shapley value is an index for fairly allocating feature importance. The determination of the Shapley value is based on cooperative game theory and can be used to determine the contribution of each adaptive parameter on the branch in the parameter optimization model. The Shapley value of each adaptive parameter can be solved using the SHAP library in Python to obtain the Shapley value of each adaptive parameter in the parameter optimization model.

[0053] In specific implementation, the optimization variable refers to the dynamic adjustment amount of the parameter optimization model for different test parameters. As a preferred embodiment, the optimization variable of the parameter optimization model can be obtained by performing a weighted analysis on all attribution values ​​and all Shapley values ​​in the following way: weights can be assigned to the attribution values ​​and Shapley values ​​of each fitness parameter according to the feature importance evaluation mechanism in the gradient boosting tree algorithm, and then all attribution values ​​and all Shapley values ​​can be weighted and summed according to the assigned weights. The value obtained by weighted summation is then used as the optimization variable of the parameter optimization model. In other embodiments, other methods can be used to determine the optimization variable, which will not be elaborated here.

[0054] It should be noted that the parameter optimization model in this application optimizes different test parameters by adjusting the weights of the attribution value and the Shapley value. That is, different test parameters correspond to different adjustment amounts, which is beneficial for the parameter optimization model to adaptively optimize different parameters to be optimized.

[0055] Furthermore, in specific implementation, the optimization confidence level reflects the degree of confidence of the parameter optimization model in the dynamic adjustment of different test parameters. As a preferred embodiment, the optimization confidence level of the parameter optimization model can be obtained by performing confidence analysis on the optimization variables in the following ways: for example, the cross-validation method in the prior art can be used to perform confidence analysis on the optimization variables; or, grid search or random search in parameter optimization techniques can be used to determine the performance of the parameter optimization model in dynamically adjusting different test parameters according to the optimization variables, thereby evaluating the confidence level of the optimization variables. In other embodiments, other methods can also be used to perform confidence analysis on the optimization variables to obtain the optimization confidence level of the parameter optimization model, which is not limited here.

[0056] In step 105, the parameter optimization model adaptively optimizes each parameter to be optimized based on the optimization confidence level.

[0057] In some embodiments, the parameter optimization model can be implemented by adaptively optimizing each parameter to be optimized based on the optimization confidence level using the following steps: Obtain the optimization variables of the parameter optimization model; If the optimization confidence level is greater than or equal to the preset confidence threshold, then for each parameter to be optimized, the parameter to be optimized is adjusted through the optimization variable to obtain the optimized test parameter, and then the optimized test parameters of all parameters to be optimized are obtained, thus completing the optimization of the wafer test parameters.

[0058] In specific implementation, adjusting the parameter to be optimized through the optimization variable to obtain the optimized test parameter means that the parameter optimization model adds the optimization variable to the parameter value of the parameter to be optimized to obtain the optimized test parameter; the confidence threshold is the evaluation metric for the optimization confidence. When the optimization confidence is greater than or equal to the preset confidence threshold, the optimization variable corresponding to the optimization confidence can be used to adjust the parameter value of the parameter to be optimized.

[0059] In addition, in some embodiments, if the optimization confidence level is less than a preset confidence threshold, the corresponding parameters in the parameter optimization model are updated, and the optimization variables of the parameter optimization model are redefined until the optimization confidence level is greater than or equal to the preset confidence threshold.

[0060] It should be noted that in this application, the partial derivative (gradient) of the loss function with respect to the corresponding parameters can be calculated by gradient descent, and then the parameters are updated along the opposite direction of the gradient. In other embodiments, existing techniques such as stochastic gradient descent and AdaGrad (Adaptive Gradient) can also be used to update the corresponding parameters in the parameter optimization model, which will not be elaborated here.

[0061] On the other hand, in some embodiments, this application provides a wafer testing parameter optimization system, referring to... Figure 4 The figure is a schematic diagram of the structure of a wafer test parameter optimization system according to some embodiments of this application. The wafer test parameter optimization system 400 includes: an acquisition module 401, a processing module 402, and an execution module 403, which are described below: The acquisition module 401 in this application is mainly used to acquire the test parameters of all dies during the wafer testing process; Processing module 402 in this application is mainly used to divide all test parameters into multiple parameters to be optimized and neighborhood parameters of each parameter to be optimized based on the predicted value of the wafer test index of the next round, and then determine the difference gradient of parameter values ​​between each parameter to be optimized and its neighborhood parameters. In this application, the processing module 402 is also used to synchronously map all the parameters to be optimized, all the neighborhood parameters and all the differential gradients to the neighborhood space by a preset spatial mapping network, and then determine the neighborhood aggregation features of all the parameters to be optimized based on the mapping vector of each differential gradient in the neighborhood space. In this application, the processing module 402 is also used to perform the next round of wafer testing operations, obtain the adaptive parameters of the corresponding dies for each parameter to be optimized, construct a parameter optimization model based on the gradient boosting tree algorithm based on all the adaptive parameters and the predicted values, and determine the optimization confidence of the parameter optimization model by the neighborhood aggregation feature; The execution module 403 in this application is mainly used for the parameter optimization model to adaptively optimize each parameter to be optimized based on the optimization confidence level.

[0062] In addition, this application also provides a computer device, the computer device including a memory and a processor, the memory storing code, and the processor being configured to acquire the code and execute the above-described wafer test parameter optimization method.

[0063] In some embodiments, reference Figure 5 The figure is a schematic diagram of a computer device for implementing a wafer testing parameter optimization method according to some embodiments of this application. The wafer testing parameter optimization method in the above embodiments can be implemented through... Figure 5 The computer device shown is used to implement this, and the computer device 500 includes at least one processor 501, a communication bus 502, a memory 503, and at least one communication interface 504.

[0064] Processor 501 can be a general-purpose central processing unit (CPU) or an application-specific integrated circuit (ASIC).

[0065] The communication bus 502 can be used to transmit information between the aforementioned components.

[0066] Memory 503 may be a read-only memory (ROM) or other type of static storage device capable of storing static information and instructions, random access memory (RAM) or other type of dynamic storage device capable of storing information and instructions, or electrically erasable programmable read-only memory (EEPROM), compact disc read-only memory (CD-ROM) or other optical disc storage, optical disc storage (including compressed optical discs, laser discs, optical discs, digital versatile optical discs, Blu-ray discs, etc.), magnetic disks or other magnetic storage devices, or any other medium capable of carrying or storing desired program code in the form of instructions or data structures and accessible by a computer, but not limited thereto. Memory 503 may exist independently and be connected to processor 501 via communication bus 502. Memory 503 may also be integrated with processor 501.

[0067] The memory 503 stores program code for executing the scheme of this application, and its execution is controlled by the processor 501. The processor 501 executes the program code stored in the memory 503. The program code may include one or more software modules. In the above embodiments, the wafer test parameter optimization method can be implemented by the processor 501 and one or more software modules in the program code in the memory 503.

[0068] Communication interface 504 uses any transceiver-like device to communicate with other devices or communication networks, such as Ethernet, radio access network (RAN), wireless local area networks (WLAN), etc.

[0069] In a specific implementation, as one example, a computer device may include multiple processors, each of which may be a single-core (single-CPU) processor or a multi-core (multi-CPU) processor. Here, a processor may refer to one or more devices, circuits, and / or processing cores used to process data (e.g., computer program instructions).

[0070] The aforementioned computer device can be a general-purpose computer device or a special-purpose computer device. In specific implementations, the computer device can be a desktop computer, a portable computer, a network server, a handheld digital assistant (PDA), a mobile phone, a tablet computer, a wireless terminal device, a communication device, or an embedded device. This application does not limit the type of computer device.

[0071] In addition, this application also provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the above-described wafer test parameter optimization method.

[0072] Although preferred embodiments of this application have been described, those skilled in the art, upon learning the basic inventive concept, can make other changes and modifications to these embodiments. Therefore, the appended claims are intended to be interpreted as including the preferred embodiments as well as all changes and modifications falling within the scope of this application.

[0073] Obviously, those skilled in the art can make various modifications and variations to this application without departing from the spirit and scope of this application. Therefore, if such modifications and variations fall within the scope of the claims of this application and their equivalents, this application also intends to include such modifications and variations.

Claims

1. A method for optimizing wafer testing parameters, characterized in that, Includes the following steps: Collect test parameters for all dies during wafer testing; Based on the predicted values ​​of the next round of wafer testing indicators, all testing parameters are divided into multiple parameters to be optimized and neighborhood parameters of each parameter to be optimized, thereby determining the gradient of the difference in parameter values ​​between each parameter to be optimized and its neighborhood parameters. A pre-defined spatial mapping network synchronously maps all the parameters to be optimized, all the neighborhood parameters, and all the differential gradients to the neighborhood space. Then, based on the mapping vector of each differential gradient in the neighborhood space, the neighborhood aggregation features of all the parameters to be optimized are determined. Perform the next round of wafer testing, obtain the adaptive parameters of the corresponding dies for each parameter to be optimized, construct a parameter optimization model based on gradient boosting tree algorithm based on all adaptive parameters and the predicted values, and determine the optimization confidence of the parameter optimization model by the neighborhood aggregation feature; The parameter optimization model adaptively optimizes each parameter to be optimized based on the optimization confidence level. The determination of the neighborhood aggregation features of all parameters to be optimized based on the mapping vectors of each difference gradient in the neighborhood space is achieved through the following steps: Plot all differential gradients as gradient change curves; Based on the gradient change curve, determine the mapping vector of each differential gradient in the neighborhood space; Obtain the distance between different parameters to be optimized in the neighborhood space; Based on all distances and all mapping vectors, determine the neighborhood aggregation features of all parameters to be optimized.

2. The method as described in claim 1, characterized in that, Based on the predicted values ​​of the next round of wafer testing metrics, all testing parameters are divided into multiple parameters to be optimized and neighborhood parameters of each parameter to be optimized. Specifically, these include: Determine the predicted values ​​for the next round of wafer testing metrics; Based on the predicted values, all test parameters that need to be optimized are selected from all test parameters, thus obtaining multiple parameters to be optimized; The neighborhood parameters of each parameter to be optimized are determined by the distribution characteristics of the grains in the wafer.

3. The method as described in claim 2, characterized in that, The neighborhood parameters of each parameter to be optimized are determined by the distribution characteristics of the grains in the wafer. Specifically, these include: Obtain the distribution characteristics of all grains in the wafer; The neighborhood range of each parameter to be optimized is determined based on the distribution characteristics. The neighborhood parameters of each parameter to be optimized are determined based on the neighborhood range of each parameter to be optimized.

4. The method as described in claim 1, characterized in that, Determining the gradient of the difference between the parameter values ​​of each parameter to be optimized and its neighboring parameters specifically includes: For each parameter to be optimized, determine the difference between the parameter to be optimized and each test parameter in the corresponding neighborhood parameters; By determining the difference gradient between the parameter to be optimized and its corresponding neighboring parameters through all the difference values, the difference gradient between the parameter to be optimized and its corresponding neighboring parameters can be obtained.

5. The method as described in claim 1, characterized in that, The pre-defined spatial mapping network synchronously maps all parameters to be optimized, all neighborhood parameters, and all differential gradients to the neighborhood space, specifically including: Obtain the distribution characteristics of all grains in the wafer; Construct a spatial mapping network based on Gaussian process regression according to the aforementioned distribution characteristics; The spatial mapping network is used to amplify all the parameters to be optimized and all the neighborhood parameters to obtain amplified parameters to be optimized and amplified neighborhood parameters. The spatial mapping network constructs a neighborhood space based on all magnified parameters to be optimized, all magnified neighborhood parameters, and all differential gradients.

6. The method as described in claim 1, characterized in that, Determining the optimization confidence of the parameter optimization model based on the neighborhood aggregation features specifically includes: Obtain the feature set input to the parameter optimization model; In the parameter optimization model, the feature importance of the feature set is evaluated through the neighborhood aggregation feature to obtain the feature importance of all adaptive parameters in the feature set; Based on the importance of all features, determine the attribution value and Shapley value of each adaptive parameter in the parameter optimization model; Weighted analysis is performed on all attribution values ​​and all Shapley values ​​to obtain the optimization variables of the parameter optimization model. Then, confidence analysis is performed on the optimization variables to obtain the optimization confidence level of the parameter optimization model.

7. The method as described in claim 1, characterized in that, The parameter optimization model adaptively optimizes each parameter to be optimized based on the optimization confidence level, specifically including: Obtain the optimization variables of the parameter optimization model; If the optimization confidence level is greater than or equal to the preset confidence threshold, then for each parameter to be optimized, the parameter to be optimized is adjusted through the optimization variable to obtain the optimized test parameter, and then the optimized test parameters of all parameters to be optimized are obtained, thus completing the optimization of the wafer test parameters.

8. A wafer testing parameter optimization system, wherein the wafer testing parameters are optimized using the method described in any one of claims 1 to 7, characterized in that, The system includes: The acquisition module is used to acquire the test parameters of all dies during the wafer testing process; The processing module is used to divide all test parameters into multiple parameters to be optimized and neighborhood parameters of each parameter to be optimized based on the predicted values ​​of the wafer test indicators for the next round, and then determine the gradient of the difference in parameter values ​​between each parameter to be optimized and its neighborhood parameters. The processing module is also used to synchronously map all the parameters to be optimized, all the neighborhood parameters, and all the differential gradients to the neighborhood space using a preset spatial mapping network, and then determine the neighborhood aggregation features of all the parameters to be optimized based on the mapping vectors of each differential gradient in the neighborhood space. The processing module is also used to execute the next round of wafer testing operations, obtain the adaptive parameters of the corresponding dies for each parameter to be optimized, construct a parameter optimization model based on gradient boosting tree algorithm based on all adaptive parameters and the predicted values, and determine the optimization confidence of the parameter optimization model by the neighborhood aggregation feature; An execution module is used for the parameter optimization model to adaptively optimize each parameter to be optimized based on the optimization confidence level.

9. A computer device comprising a memory and a processor, the memory storing code, characterized in that, The processor is configured to acquire the code and execute the wafer test parameter optimization method as described in any one of claims 1 to 7.

10. A computer-readable storage medium storing a computer program, characterized in that, When the computer program is executed by the processor, it implements the wafer test parameter optimization method as described in any one of claims 1 to 7.

Images