Multivariable-based semiconductor diagnosis method, device and equipment and storage medium

Abstract

The invention provides a multi-variable-based semiconductor diagnosis method, device and equipment and a storage medium, and the method comprises the steps: obtaining grouping process variables of each target sensor group in a target manufacturing system in response to an abnormality detection request for the target manufacturing system; performing feature extraction on the grouped process variables to obtain grouped variable features corresponding to the grouped process variables; performing anomaly detection on the grouping variable characteristics by using an anomaly detection model corresponding to the target sensor group to obtain variable detection labels corresponding to the grouping variable characteristics; and determining a system detection result of the target manufacturing system based on the variable detection label corresponding to each target sensor group. According to the technical scheme, the influence of irrelevant variables can be effectively reduced in the semiconductor anomaly detection process, and the interpretability and accuracy of the anomaly detection process are improved.

Description

Technical Field

[0001] This application relates to the field of computer technology, and specifically to a semiconductor diagnostic method, apparatus, device, and storage medium based on multiple variables. Background Technology

[0002] Currently, the semiconductor manufacturing process involves numerous complex manufacturing equipment and process variables such as temperature, pressure, flow rate, and radio frequency power, which are highly coupled. To ensure the accuracy of semiconductor manufacturing, anomaly detection is required during the manufacturing process. However, existing anomaly detection methods are based on single-variable thresholds, which suffer from low detection accuracy, resulting in high false alarm and false negative rates, and cannot meet the needs of current semiconductor manufacturing inspection scenarios. Summary of the Invention

[0003] This application provides a semiconductor diagnostic method, apparatus, device, and storage medium based on multiple variables, aiming to solve the technical problem in the prior art that production anomalies cannot be accurately detected during semiconductor manufacturing.

[0004] On one hand, embodiments of this application provide a multivariate-based semiconductor diagnostic method, which includes the following steps: In response to an anomaly detection request for the target manufacturing system, obtain the grouping process variables of each target sensor group in the target manufacturing system; Feature extraction is performed on the grouping process variables to obtain the grouping variable features corresponding to the grouping process variables; Anomaly detection is performed on the grouped variable features using the anomaly detection model corresponding to the target sensor group, and variable detection labels corresponding to the grouped variable features are obtained. Based on the variable detection label corresponding to each target sensor group, the system detection result of the target manufacturing system is determined.

[0005] In one possible implementation of this application, the step of using the anomaly detection model corresponding to the target sensor group to perform anomaly detection on the grouped variable features and obtaining the variable detection label corresponding to the grouped variable features includes: Obtain the anomaly detection model corresponding to the target sensor group, wherein the anomaly detection model includes a first anomaly detection module, a second anomaly detection module, and a label fusion module; The first anomaly detection module is used to perform a first anomaly detection on the grouping variable features to obtain a first variable detection label; The second anomaly detection module is used to perform second anomaly detection on the grouping variable features to obtain the second variable detection label; The label fusion module is used to evaluate the first variable detection label and the second variable detection label to determine the variable detection label corresponding to the grouped variable feature.

[0006] In one possible implementation of this application, the second anomaly detection module includes at least two different anomaly detection units and a detection fusion unit; the input terminal of the anomaly detection unit is used to receive grouped variable features, the output terminal of the anomaly detection unit is connected to the input terminal of the detection fusion unit, and the output terminal of the detection fusion unit is connected to the tag fusion module as the output terminal of the second anomaly detection module. The step of using a second anomaly detection module to perform second anomaly detection on the grouped variable features to obtain a second variable detection label includes: The anomaly detection unit is used to calculate the initial feature distribution parameters of the grouped variable features; Based on the detection fusion unit and the initial feature distribution parameters, the variable feature distribution parameters corresponding to the grouped variable features are determined; The feature deviation parameters of the grouped variable features are determined based on the variable feature distribution parameters and the preset target feature distribution parameters; The second variable detection label is generated based on the feature deviation parameter and the preset deviation threshold.

[0007] In one possible implementation of this application, the step of evaluating the first variable detection label and the second variable detection label using the label fusion module to determine the variable detection label corresponding to the grouped variable feature includes: When both the first variable detection label and the second variable detection label are normal detection labels, the normal detection label is determined as the variable detection label corresponding to the grouped variable feature; When either the first variable detection label or the second variable detection label is an anomaly detection label, the anomaly detection label is determined as the variable detection label corresponding to the grouped variable feature.

[0008] In one possible implementation of this application, before performing anomaly detection on the grouped variable features using the anomaly detection model corresponding to the target sensor group, the method further includes: The sensor devices in the target manufacturing system are grouped to obtain the target sensor grouping corresponding to the target manufacturing system; Obtain the grouping variable samples corresponding to the target sensor group, and use the grouping variable samples to train the initial model to obtain the anomaly detection model corresponding to the target sensor group.

[0009] In one possible implementation of this application, the step of obtaining grouping variable samples corresponding to the target sensor group and training an initial model using the grouping variable samples to obtain an anomaly detection model corresponding to the target sensor group includes: Obtain the original variable samples corresponding to the target manufacturing system, and perform data separation on the original variable samples based on the machine modules and / or signal types of the target sensor group to obtain the grouped variable samples of the target sensor group; The grouping variable samples are augmented to obtain group-augmented samples; The target grouped samples are obtained by performing fault injection processing on the grouped enhanced samples using a fault injection model and a target fault mode. The initial model is trained using the target group samples to obtain the anomaly detection model corresponding to the target sensor group.

[0010] In one possible implementation of this application, the step of augmenting the grouping variable samples to obtain group-enhanced samples includes: Based on the grouping variable samples, a similarity search is performed on the sample database associated with the target manufacturing system to obtain the first enhanced sample associated with the grouping variable samples in the sample database; Using the data augmentation model and the grouping variable samples, a second augmented sample is generated. Grouped augmented samples are generated based on the grouping variable samples, the first augmented sample, and the second augmented sample.

[0011] In one possible implementation of this application, the step of performing fault injection processing on the grouped enhanced samples using a fault injection model and a target fault mode to obtain target grouped samples includes: Access the fault mode library and obtain the target fault mode corresponding to the sensor group in the fault mode library; The grouped enhanced samples are configured using the fault injection model and the target fault mode to obtain the grouped fault samples corresponding to the grouped enhanced samples; Target group samples are generated based on the group enhancement samples and the group failure samples.

[0012] In one possible implementation of this application, training the initial model using the target group samples to obtain the anomaly detection model corresponding to the target sensor group includes: Feature extraction is performed on the target grouped samples to obtain grouped sample features, which include group enhancement features corresponding to group enhancement samples and group fault features corresponding to group fault samples. The first initial detection module in the initial model is trained under supervision using the grouped enhancement features and the grouped fault features to obtain the first anomaly detection module. The first initial detection module in the initial model is semi-supervised and trained using the grouped enhancement features to obtain the second anomaly detection module. An anomaly detection model corresponding to the target sensor group is generated based on the first anomaly detection module and the second anomaly detection module.

[0013] In one possible implementation of this application, after determining the system detection result of the target manufacturing system based on the variable detection label corresponding to each target sensor group, the method further includes: In response to detection tagging events related to the system's detection results, the system counts the number of tagging events for each detection tagging event. If the number of markings is greater than or equal to the number of markings threshold, the sample is updated based on the detected marking event and the target group sample to obtain the updated group sample; The anomaly detection model is updated using the updated grouped samples to obtain an updated detection model; The update detection model is used to perform update detection on the grouping process variables to obtain update detection results.

[0014] In one possible implementation of this application, after determining the system detection result of the target manufacturing system based on the variable detection label corresponding to each target sensor group, the method further includes: Obtain the variable distribution data of the grouping process variables, and the training distribution data corresponding to the grouping process variables; Calculate the variable offset parameter of the grouping process variable based on the variable distribution data and the training distribution data; If the variable offset parameter is greater than the preset offset threshold, the anomaly detection model is updated to obtain an updated detection model.

[0015] On the other hand, this application provides a common detection device, the semiconductor diagnostic device comprising: The variable acquisition module is configured to acquire the grouping process variables of each target sensor group in the target manufacturing system in response to an anomaly detection request for the target manufacturing system. The feature extraction module is configured to extract features from the grouping process variables to obtain the grouping variable features corresponding to the grouping process variables; The anomaly detection module is configured to use the anomaly detection model corresponding to the target sensor group to perform anomaly detection on the group variable features and obtain the variable detection label corresponding to the group variable features. An anomaly detection module is configured to determine the system detection result of the target manufacturing system based on the variable detection label corresponding to each target sensor group.

[0016] On the other hand, this application also provides a semiconductor diagnostic device, the semiconductor diagnostic device comprising: One or more processors; Memory; and One or more applications, wherein the one or more applications are stored in the memory and configured to be executed by the processor to implement the steps of the semiconductor diagnostic method.

[0017] On the other hand, this application also provides a computer-readable storage medium having a computer program stored thereon, the computer program being loaded by a processor to perform the steps in the semiconductor diagnostic method.

[0018] This application, in response to an anomaly detection request for a target manufacturing system, acquires the grouping process variables of each target sensor group in the target manufacturing system; extracts features from the grouping process variables to obtain the corresponding grouping variable features; uses the anomaly detection model corresponding to each target sensor group to perform anomaly detection on the grouping variable features to obtain the variable detection labels corresponding to the grouping variable features; and determines the system detection result of the target manufacturing system based on the variable detection labels corresponding to each target sensor group. This achieves the following: during anomaly detection of the target manufacturing system, sensors in the target manufacturing system are grouped according to sensor mechanisms, and the grouping process variables corresponding to each target sensor group are collected. Then, the grouping variable features and anomaly detection model are used to perform grouped anomaly detection to determine the variable detection labels of each target sensor group. Based on these variable detection labels, the system detection result is determined to determine whether there are variable anomalies in the target manufacturing system. This effectively reduces the influence of irrelevant variables in the semiconductor anomaly detection process, improving the interpretability and accuracy of the anomaly detection process. Attached Figure Description

[0019] To more clearly illustrate the technical solutions in the embodiments of this application, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0020] Figure 1 This is a schematic diagram illustrating a scenario of the semiconductor diagnostic method according to an embodiment of this application; Figure 2 This is a flowchart illustrating one embodiment of the semiconductor diagnostic method in this application. Figure 3a This is a schematic diagram of the structure of the anomaly detection model in the semiconductor diagnostic method provided in the embodiments of this application; Figure 3b A schematic diagram of a scenario of an embodiment of the second anomaly detection module in the semiconductor diagnostic method provided in this application; Figure 4 A flowchart illustrating an embodiment of the semiconductor diagnostic method provided in this application for model training; Figure 5 This is a flowchart illustrating an embodiment of updating an anomaly detection model in a semiconductor diagnostic method provided in this application. Figure 6 A schematic diagram of the structure of one embodiment of the semiconductor diagnostic device provided in this application; Figure 7 This is a schematic diagram of one embodiment of the semiconductor diagnostic device provided in this application. Detailed Implementation

[0021] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0022] In the description of this invention, it should be understood that the terms "center," "longitudinal," "lateral," "length," "width," "thickness," "upper," "lower," "front," "rear," "left," "right," "vertical," "horizontal," "top," "bottom," "inner," and "outer," etc., indicating orientation or positional relationships based on the orientation or positional relationships shown in the accompanying drawings, are only for the convenience of describing the invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation, and therefore should not be construed as a limitation of the invention. Furthermore, the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of indicated technical features. Thus, features defined with "first" and "second" may explicitly or implicitly include one or more of the stated features. In the description of this invention, "a plurality of" means two or more, unless otherwise explicitly specified.

[0023] In this application, the term "exemplary" is used to mean "serving as an example, illustration, or description." Any embodiment described as "exemplary" in this application is not necessarily to be construed as being more preferred or advantageous than other embodiments. The following description is provided to enable any person skilled in the art to make and use the invention. Details are set forth in the following description for purposes of explanation. It should be understood that those skilled in the art will recognize that the invention can be made without using these specific details. In other instances, well-known structures and processes will not be described in detail to avoid obscuring the description of the invention with unnecessary detail. Therefore, the invention is not intended to be limited to the embodiments shown, but is consistent with the broadest scope of the principles and features disclosed in this application.

[0024] Currently, the semiconductor manufacturing process involves numerous complex manufacturing equipment and process variables such as temperature, pressure, flow rate, and radio frequency power, which are highly coupled. To ensure the accuracy of semiconductor manufacturing, anomaly detection is required during the manufacturing process. However, existing anomaly detection methods are based on single-variable thresholds, which suffer from low detection accuracy, resulting in high false alarm and false negative rates, and cannot meet the needs of current semiconductor manufacturing inspection scenarios.

[0025] Based on this, this application proposes a multivariate-based semiconductor diagnostic method, apparatus, device, and computer-readable storage medium to solve the technical problem in the prior art that production anomalies cannot be accurately detected during semiconductor manufacturing.

[0026] The semiconductor diagnostic method in this embodiment of the invention is applied to a semiconductor diagnostic device, which is disposed in a semiconductor diagnostic equipment. The semiconductor diagnostic equipment is provided with one or more processors, a memory, and one or more application programs, wherein the one or more application programs are stored in the memory and configured to be executed by the processor to implement the semiconductor diagnostic method. The semiconductor diagnostic equipment can be a smart terminal, such as a mobile phone, tablet computer, network device, and smart computer. Optionally, the semiconductor diagnostic equipment can also be a server or a service cluster composed of multiple servers.

[0027] like Figure 1 As shown, Figure 1This is a schematic diagram of a semiconductor diagnostic method according to an embodiment of this application. The anomaly detection scenario in this embodiment includes a semiconductor diagnostic device 100 (which integrates a semiconductor diagnostic apparatus) and a target manufacturing system 200. The semiconductor diagnostic device 100 runs a computer-readable storage medium corresponding to the semiconductor diagnostic method to execute the steps of the semiconductor diagnostic method. The target manufacturing system 200 is a semiconductor manufacturing system or other manufacturing system that is communicatively connected to the semiconductor diagnostic device 100 and is subject to anomaly detection.

[0028] Understandable, Figure 1 The semiconductor diagnostic equipment in the semiconductor diagnostic method scenario shown, or the devices included in the semiconductor diagnostic equipment, do not constitute a limitation on the embodiments of the present invention. That is, the number or type of semiconductor diagnostic equipment included in the semiconductor diagnostic method scenario, or the number or type of devices included in each device, do not affect the overall implementation of the technical solution in the embodiments of the present invention, and can all be considered as equivalent substitutions or derivatives of the technical solutions claimed in the embodiments of the present invention.

[0029] In this embodiment of the invention, the semiconductor diagnostic device 100 is mainly used to: respond to an anomaly detection request for a target manufacturing system and obtain the grouping process variables of each target sensor group in the target manufacturing system; Feature extraction is performed on the grouping process variables to obtain the grouping variable features corresponding to the grouping process variables; Anomaly detection is performed on the grouped variable features using the anomaly detection model corresponding to the target sensor group, and variable detection labels corresponding to the grouped variable features are obtained. Based on the variable detection label corresponding to each target sensor group, the system detection result of the target manufacturing system is determined.

[0030] The semiconductor diagnostic device 100 in this embodiment of the invention can be an independent semiconductor diagnostic device, such as a smart terminal like a mobile phone, tablet computer, network device, server, and smart computer, or it can be an anomaly detection network or anomaly detection cluster composed of multiple semiconductor diagnostic devices.

[0031] This application provides a multivariate-based semiconductor diagnostic method, apparatus, device, and computer-readable storage medium, which will be described in detail below.

[0032] It will be understood by those skilled in the art that Figure 1 The application environment shown is only one application scenario related to the solution of this application and does not constitute a limitation on the application scenario of this application. Other application environments may include more than one application scenario. Figure 1 The diagram shows more or fewer semiconductor diagnostic devices, or anomaly detection network connections, for example... Figure 1 Only one semiconductor diagnostic device is shown in the diagram. It is understood that the scenario of this semiconductor diagnostic method may also include one or more semiconductor diagnostic devices, which are not specifically limited here. The semiconductor diagnostic device 100 may also include a memory for storing grouping process variables and other data.

[0033] It should be noted that, Figure 1 The schematic diagram of the semiconductor diagnostic method shown is merely an example. The scenarios of the semiconductor diagnostic method described in the embodiments of the present invention are intended to more clearly illustrate the technical solutions of the embodiments of the present invention and do not constitute a limitation on the technical solutions provided in the embodiments of the present invention.

[0034] Based on the above-mentioned scenarios of semiconductor diagnostic methods, various embodiments of the semiconductor diagnostic methods disclosed in this invention are proposed.

[0035] like Figure 2 As shown, Figure 2 This is a flowchart illustrating one embodiment of the semiconductor diagnostic method in this application. The semiconductor diagnostic method includes the following steps 201 to 204: 201. In response to an anomaly detection request for the target manufacturing system, obtain the grouping process variables of each target sensor group in the target manufacturing system; The semiconductor diagnostic method in this embodiment is applied to a semiconductor diagnostic device. The type and number of semiconductor diagnostic devices are not specifically limited. That is, the semiconductor diagnostic device can be one or more smart terminals or servers. In one specific embodiment, the semiconductor diagnostic device is a smart device such as a mobile phone, tablet computer, network device, service, and smart computer that can perform anomaly monitoring on the target manufacturing system. Optionally, in other embodiments, the semiconductor diagnostic device can also be other smart terminals or components in the target manufacturing system used for anomaly detection. This embodiment does not make specific limitations here.

[0036] Optionally, the target manufacturing system is a combination of manufacturing equipment and production lines comprising multiple process steps for producing a specified product from raw materials and / or components. For example, in one specific embodiment, the target manufacturing system is a semiconductor manufacturing system. This semiconductor manufacturing system includes process steps such as etching, deposition, ion implantation, and oxidation, with different sensor devices installed at each process step to collect corresponding process variables. Optionally, in other embodiments, the target manufacturing system can also be other manufacturing systems.

[0037] Optionally, since there are a large number of complex equipment and process variables in the process links of the target manufacturing system, and there are high coupling relationships between the variables, the changes in the process state are often the result of the combined effect of multiple variables. Therefore, during the operation of the target manufacturing system, it is necessary to perform anomaly detection on the target manufacturing system in order to identify abnormal situations in the initial target manufacturing system production process in a timely and effective manner.

[0038] Optionally, an anomaly detection request is an operational event that drives the semiconductor diagnostic equipment to group process variables in the target manufacturing system and monitor the grouped process variables. Optionally, in a specific embodiment, the anomaly detection request can be actively triggered by the user or automatically triggered by the semiconductor diagnostic equipment. For example, when the semiconductor diagnostic equipment detects that the target manufacturing system is in operation, it automatically triggers the anomaly detection request to perform grouped anomaly detection on the target manufacturing system.

[0039] Optionally, the target sensor group is a sensor combination obtained by grouping sensor devices in each process step of the target manufacturing system according to preset grouping rules. For example, in one specific embodiment, a semiconductor diagnostic device can group sensor devices in the target manufacturing system based on preset grouping rules to construct multiple sensor groups, and then perform anomaly monitoring on the grouping process variables of the corresponding sensor groups in subsequent steps. The preset grouping rules are rules for grouping sensor devices according to the physical mechanism and process action mechanism of each sensor device in the target manufacturing system to construct the target sensor group. Optionally, in one specific embodiment, the target sensor group includes a vacuum and gas transport group, a heat pump sensor group, a radio frequency energy sensor group, and a mechanical motion sensor group, etc.

[0040] Optionally, the grouped process variables are the process variables collected by the target sensor group during the production process of the target manufacturing system. These process variables are physical quantities or state quantities that need to be measured, monitored, and / or regulated during production or control. Different target sensor groups correspond to different grouped process variables, and the grouped process variables within different target sensor groups have strong physical correlations. Therefore, independent modeling and diagnosis of the grouped process variables within each target sensor group can be performed to reduce noise coupling between variables and improve diagnostic accuracy.

[0041] For example, in one specific embodiment, the process variables corresponding to the vacuum and gas delivery group are process variables such as pressure, flow rate, and valve opening; in another specific embodiment, the process variables corresponding to the thermal control sensor group are process variables such as temperature, cooling flow rate, and heating power; in yet another specific embodiment, the process variables corresponding to the radio frequency energy sensor group are process variables such as radio frequency power, reflected power, and matching state; and the process variables corresponding to the mechanical motion sensor group are process variables corresponding to signals such as motion time, displacement, and current.

[0042] Optionally, during the manufacturing process of the target manufacturing system, the semiconductor diagnostic equipment pre-groups the sensor devices in the target manufacturing system. After obtaining the target sensor groups, it also responds to the anomaly detection request for the target manufacturing system and obtains the grouping process variables of each target sensor group in the target manufacturing system. Then, in subsequent steps, it identifies the operating status of the target manufacturing system through the grouping process variables, thereby performing group anomaly detection on the target manufacturing system.

[0043] 202. Perform feature extraction on the grouping process variables to obtain the grouping variable features corresponding to the grouping process variables; Optionally, after acquiring the grouping process variables corresponding to each target sensor group, the semiconductor diagnostic equipment also performs feature extraction on the grouping process variables to obtain grouping variable features.

[0044] Optionally, the grouping variable features are feature information extracted from the grouping process variables to describe the attributes and meanings of the grouping process variables. Optionally, in one specific embodiment, the grouping variable features include any one or more of the statistical features, dynamic features, frequency domain features, and correlation features extracted from the grouping process variables. In one specific embodiment, the statistical feature can be any one or more of the mean, variance, and skewness corresponding to the grouping process variable. The dynamic feature can be any one or more of the rolling mean, lag, and difference features corresponding to the grouping process variable. The frequency domain feature can be any one or more of the FFT amplitude, power spectral density, etc., corresponding to the grouping process variable. The correlation feature can be any one or more of the cross-correlation coefficient and PCA principal components corresponding to the grouping process variable.

[0045] Optionally, in one specific embodiment, after acquiring the grouping process variables corresponding to each target sensor group, the semiconductor diagnostic device further preprocesses the grouping process variables to improve data quality. That is, it performs any one or more preprocessing operations on the grouping process variables, such as cleaning, outlier removal, missing value imputation, collinearity detection, and feature standardization, to obtain preprocessed grouping process variables. Then, it extracts features from the grouping process variables to obtain grouping variable features.

[0046] Optionally, the semiconductor diagnostic equipment uses a pre-set feature extraction module to perform multi-dimensional feature extraction on the grouped process variables to obtain the statistical features, dynamic features, frequency domain features and correlation features corresponding to the grouped process variables, and determines the statistical features, dynamic features, frequency domain features and correlation features as the grouping variable features of the grouped process variables.

[0047] 203. Use the anomaly detection model corresponding to the target sensor group to perform anomaly detection on the group variable features, and obtain the variable detection label corresponding to the group variable features; Optionally, after obtaining the grouping process variables and corresponding grouping variable features corresponding to each target sensor group, the semiconductor diagnostic equipment also uses the anomaly detection model corresponding to the target sensor group to perform anomaly detection on the grouping variable features, and obtains the variable detection label corresponding to the grouping variable features.

[0048] Optionally, in one specific embodiment, the semiconductor diagnostic device pre-trains an initial model on the group variable samples corresponding to the target sensor group to obtain an anomaly detection model corresponding to the target sensor group, and uses the anomaly detection model to perform anomaly detection on the group variable features corresponding to the target sensor group to obtain variable detection labels corresponding to the group variable features.

[0049] Optionally, the anomaly detection model is an artificial intelligence model associated with the target sensor grouping, used to detect grouping anomalies in the grouping process variables corresponding to the target sensor grouping. Optional, such as... Figure 3a As shown, Figure 3aThis is a schematic diagram of the anomaly detection model in the semiconductor diagnostic method provided in this application embodiment. The anomaly detection model includes a first anomaly detection model, a second anomaly detection module, and a label fusion module. The first and second anomaly detection modules respectively receive grouped variable features and independently perform anomaly detection based on these features. The first anomaly detection module outputs a first variable detection label representing the detection result of the grouped variable features, and the second anomaly detection module outputs a second variable detection label representing another detection result of the grouped variable features. The first and second variable detection labels are then input into the label fusion module for final evaluation to determine the variable detection label representing whether anomalies have occurred in the grouped variable features. Optionally, in one specific embodiment, the first anomaly detection module can be any one or more supervised models. The second anomaly detection module can be an artificial intelligence model composed of multiple unsupervised models.

[0050] Optionally, the variable detection label is detection label information used to characterize whether there are abnormal variables in the grouping process variables of the target sensor group that deviate from the normal range. In one specific embodiment, the variable detection label includes a normal detection label and an abnormal detection label. The normal detection label indicates that the grouping process variables in the detected target sensor group are normal process variables that do not exhibit abnormalities. The abnormal detection label indicates that the grouping process variables in the detected target sensor group are abnormal process variables that exhibit abnormalities.

[0051] Optionally, after acquiring the grouping variable features, the semiconductor diagnostic device also acquires the anomaly detection model corresponding to the target sensor group. The anomaly detection model includes a first anomaly detection module, a second anomaly detection module, and a tag fusion module.

[0052] Optionally, after acquiring the grouped variable features, the semiconductor diagnostic device inputs these features into a first anomaly detection module. This module then performs a first anomaly detection on the grouped variable features to obtain a first variable detection label. That is, the semiconductor diagnostic device uses the first anomaly detection module, which represents a supervised model, to detect and classify anomalies in the grouped variable features, thereby determining the first variable detection label for the grouped variable features. Optionally, in one specific embodiment, the first anomaly detection module can select different supervised models based on the sample size of the grouped variable samples during training. For example, when the sample size is small, a support vector machine model is selected as the first anomaly detection module. When the sample size is large, an ensemble learning model based on gradient boosting is selected as the first anomaly detection module.

[0053] Optionally, the semiconductor diagnostic device also inputs the grouped variable features into a second anomaly detection module, which performs a second anomaly detection on the grouped variable features to obtain a second variable detection label. That is, the semiconductor diagnostic device uses the second anomaly detection module to evaluate the feature distribution of the grouped variable features and the target feature distribution parameters corresponding to the feature distribution representing the normal state to determine the degree of feature deviation corresponding to the grouped variable features, thereby generating a second variable feature label.

[0054] Optional, such as Figure 3b As shown, Figure 3b This is a schematic diagram of a scenario of an embodiment of the second anomaly detection module in the semiconductor diagnostic method provided in this application. Figure 3b In this embodiment, the second anomaly detection module includes at least two different anomaly detection units and a detection fusion unit. The input of each anomaly detection unit receives grouped variable features, and its output is connected to the input of the detection fusion unit. The output of the detection fusion unit serves as the output of the second anomaly detection module and is connected to the label fusion module. The anomaly detection unit is a semi-supervised model used to determine the degree of deviation between the grouped variable features and the target feature distribution parameters. The detection fusion unit is a detection unit that determines the second variable detection label based on the feature deviation parameters output by each anomaly detection unit. Optionally, in one specific embodiment, the anomaly detection unit can be at least two or more of the following: a model based on feature space dimensionality reduction (PCA), a model based on sample similarity or neighborhood distance (KNN), and a model based on statistical distribution (COPOD). Optionally, in other embodiments, the anomaly detection unit can also be other semi-supervised models.

[0055] Optionally, after acquiring the grouped variable features, the semiconductor diagnostic device inputs these features into each anomaly detection unit of the second anomaly detection module, and uses these anomaly detection units to calculate the initial feature distribution parameters of the grouped variable features. That is, the semiconductor diagnostic device independently calculates the initial feature distribution parameters of the grouped variable features using each anomaly detection unit in the second anomaly detection module, and inputs these initial feature distribution parameters into the detection fusion unit. Here, the initial feature distribution parameters are parameter information output by each anomaly detection unit that characterizes the feature distribution of the grouped variable features.

[0056] Optionally, the semiconductor diagnostic equipment further determines the variable feature distribution parameters corresponding to the grouped variable feature based on the detection fusion unit and the initial feature distribution parameters. That is, after each anomaly detection unit calculates the initial feature distribution parameters and inputs them into the detection fusion unit, the detection fusion unit determines the final variable feature distribution parameters corresponding to the grouped variable feature through methods such as voting and / or weighted fusion of the initial feature distribution parameters. Here, the variable feature distribution parameters are parameter information characterizing the feature distribution state corresponding to the grouped variable feature.

[0057] Optionally, after generating the variable feature distribution parameters, the detection fusion unit further determines the feature deviation parameters of the grouped variable features based on these variable feature distribution parameters and preset target feature distribution parameters. Specifically, the detection fusion unit determines the degree to which the grouped variable features deviate from the normal distribution state by comparing the variable feature distribution parameters with the target feature distribution parameters that characterize the normal feature distribution state, and generates a second variable detection label based on these feature deviation parameters and a preset deviation threshold. The preset deviation threshold is an evaluation threshold used to determine whether the grouped variable features deviate from the normal variable features based on the feature deviation parameters.

[0058] Optionally, if the feature deviation parameter is less than the preset deviation threshold, it is determined that the grouped variable feature has not deviated from the normal variable feature, and the normal detection label is determined as the second variable detection label of the grouped variable feature.

[0059] Optionally, if the feature deviation parameter is greater than the preset deviation threshold, it is determined that the grouped variable feature has deviated from the normal variable feature, and the anomaly detection label is determined as the second variable detection label of the grouped variable feature.

[0060] Optionally, after acquiring the first variable detection label and the second variable detection label of the anomaly detection model, the semiconductor diagnostic equipment inputs the first variable detection label and the second variable detection label into the label fusion module. The label fusion module then fuses the first variable detection label and the second variable detection label to determine the variable detection label corresponding to the grouped variable feature. That is, after receiving the first variable detection label and the second variable detection label, the label fusion module performs a bitwise OR operation on the first variable detection label and the second variable detection label to obtain the variable detection label. In other words, if either the first variable detection label or the second variable detection label contains an anomaly detection label, then the anomaly detection label is determined as the variable detection label.

[0061] Optionally, when both the first variable detection label and the second variable detection label are normal detection labels, the label fusion module determines the normal detection label as the variable detection label corresponding to the grouped variable feature; Optionally, when the first variable detection label or the second variable detection label is an anomaly detection label, the label fusion module determines the anomaly detection label as the variable detection label corresponding to the grouped variable feature.

[0062] 204. Based on the variable detection label corresponding to each target sensor group, determine the system detection result of the target manufacturing system.

[0063] Optionally, after generating corresponding variable detection labels by performing anomaly detection on the group variable features corresponding to each target sensor group through the anomaly detection model corresponding to each target sensor group, the semiconductor diagnostic equipment also determines the system detection result based on the variable detection labels corresponding to each target sensor group.

[0064] The system detection results are those indicating whether the target manufacturing system has encountered any abnormal states during the production process. In one specific embodiment, the system detection results include normal detection results indicating that the target manufacturing system is operating normally and abnormal detection results indicating that the target manufacturing system is operating abnormally.

[0065] Optionally, in one specific embodiment, after acquiring the variable detection tag corresponding to each target sensor group, the semiconductor diagnostic device performs a logical OR operation on the variable detection tag corresponding to each target sensor group to determine the system detection result of the target manufacturing system. That is, when the variable detection tag corresponding to any target sensor group is an anomaly detection tag, the system detection result of the target manufacturing system is determined to be an anomaly detection result, and the abnormal sensor group and corresponding abnormal process variable associated with the anomaly detection result in the target sensor group are determined, as well as the feature contribution and anomaly score of the abnormal process variable, thereby effectively reducing the interference of irrelevant variables and improving the interpretability and accuracy of anomaly detection.

[0066] Optionally, in one specific embodiment, the semiconductor diagnostic device can also update the anomaly detection model based on the detection marker events and data distribution reported by the user, obtain an updated detection model, and use the updated detection model to perform updated detection on the target manufacturing system to obtain updated detection results.

[0067] In this embodiment, the semiconductor diagnostic equipment, in response to an anomaly detection request for a target manufacturing system, acquires the grouping process variables of each target sensor group in the target manufacturing system; extracts features from the grouping process variables to obtain the corresponding grouping variable features; uses the anomaly detection model corresponding to each target sensor group to perform anomaly detection on the grouping variable features to obtain the variable detection labels corresponding to the grouping variable features; and determines the system detection result of the target manufacturing system based on the variable detection labels corresponding to each target sensor group. This achieves the following: during the anomaly detection process of the target manufacturing system, sensors in the target manufacturing system are grouped according to sensor mechanisms, and the grouping process variables corresponding to each target sensor group are collected. Then, the grouping variable features and anomaly detection model are used to perform grouped anomaly detection to determine the variable detection labels of each target sensor group. Based on these variable detection labels, the system detection result is determined to determine whether there are variable anomalies in the target manufacturing system. This effectively reduces the influence of irrelevant variables and improves the interpretability and accuracy of the anomaly detection process.

[0068] like Figure 4 As shown, Figure 4 A flowchart illustrating an embodiment of model training in the semiconductor diagnostic method provided in this application is shown below. Figure 4 As shown in the embodiment, the semiconductor diagnostic method further includes steps 301 to 302: 301. Group the sensor devices in the target manufacturing system to obtain the target sensor grouping corresponding to the target manufacturing system; 302. Obtain the grouping variable samples corresponding to the target sensor group, and use the grouping variable samples to train the initial model to obtain the anomaly detection model corresponding to the target sensor group.

[0069] Based on the above embodiments, in this embodiment, the semiconductor diagnostic device pre-trains the initial model on the group variable samples corresponding to the target sensor group to obtain the anomaly detection model corresponding to the target sensor group, and uses the anomaly detection model to perform anomaly detection on the group variable features corresponding to the target sensor group to obtain the variable detection label corresponding to the group variable features.

[0070] Optionally, the sensor device is a sensor component used to collect various process variables in the target manufacturing system. The target sensor group is a sensor combination obtained by grouping the sensor devices of each process step in the target manufacturing system according to preset grouping rules. For example, in one specific embodiment, a semiconductor diagnostic device can group the sensor devices in the target manufacturing system based on preset grouping rules to construct multiple sensor groups, and then perform anomaly monitoring on the grouped process variables of the corresponding sensor groups in subsequent steps. Optionally, different sensor groups correspond to different grouped process variables and grouped variable samples.

[0071] Optionally, after generating the target sensor group, the semiconductor diagnostic device also acquires the grouping variable samples corresponding to the target sensor group. These grouping variable samples are training sample data used to train the initial model corresponding to the target sensor group.

[0072] Optionally, in one specific embodiment, the semiconductor diagnostic device acquires the original variable samples corresponding to the target manufacturing system, performs data separation on the original variable samples based on the machine modules and / or signal types of the target sensor group, obtains the grouped variable samples of the target sensor group, and then in subsequent steps, the initial model is trained by grouping the grouped variable samples to obtain an anomaly detection model, thereby reducing the curse of dimensionality and reducing the training difficulty of the model.

[0073] Optionally, the original variable samples are tagged variable sample information associated with each target sensor group in the target manufacturing system. Optionally, in one specific embodiment, the semiconductor diagnostic device obtains tagged variable samples associated with the target manufacturing system from the variable database through the data access module as the original variable samples.

[0074] Optionally, the machine module is a mechanical-electrical-control unit that is functionally packaged in the target manufacturing system, can be independently installed / replaced, and has a standardized interface. Optionally, in other embodiments, the semiconductor diagnostic device can also perform data separation on the original variable samples through physical mechanisms to obtain grouped variable samples corresponding to each target sensor group.

[0075] Optionally, in one specific embodiment, after acquiring the original variable samples, the semiconductor diagnostic device further preprocesses the original variable samples to improve data quality, resulting in preprocessed original variable samples. The preprocessing operations include cleaning the original variable samples, removing outliers, imputing missing values, detecting collinearity, and standardizing features to ensure data quality and consistency.

[0076] Optionally, after dividing the original variable samples into grouped variable samples, the semiconductor diagnostic equipment further enhances the number of grouped variable samples to obtain grouped enhanced samples, thereby expanding the number of samples and alleviating the shortage of normal samples.

[0077] Optionally, in one specific embodiment, the semiconductor diagnostic device performs a similarity search on a sample database associated with the target manufacturing system based on the grouping variable sample, and obtains a first enhanced sample associated with the grouping variable sample in the sample database. The sample database stores sample data corresponding to samples belonging to the same or similar equipment and processes as the target manufacturing system. The semiconductor diagnostic device can retrieve the sample database using a similarity retrieval algorithm to obtain normal variable samples whose statistical distribution is similar to the grouping variable sample as the first enhanced sample. Optionally, in one specific embodiment, the similarity retrieval algorithm includes retrieval algorithms such as Dynamic Time Warping (DTW) and clustering sharing. The first enhanced sample is a normal variable sample whose statistical distribution data is similar to the grouping variable sample obtained through similarity retrieval. In one specific embodiment, the first enhanced sample is a variable sample in the sample database whose statistical distribution data is greater than a preset distribution threshold.

[0078] For example, in one specific embodiment, the semiconductor diagnostic device can acquire a high-quality target variable sample from the grouped variable sample and perform a similarity search based on the target variable sample to obtain a first enhanced sample.

[0079] Optionally, in another specific embodiment, the semiconductor diagnostic device further pre-sets a data augmentation model for data augmentation, and uses the data augmentation model and the grouping variable sample to generate a second augmented sample. That is, the semiconductor diagnostic device uses the data augmentation model to learn the sample distribution data of the grouping variable sample, and generates a generative sample with a similarity greater than a preset threshold to the sample distribution data as the second augmented sample. The second augmented sample is an augmented sample generated by the data augmentation model and associated with the grouping variable sample.

[0080] Optionally, after acquiring the first enhanced sample and the second enhanced sample, the semiconductor diagnostic device generates a grouped enhanced sample based on the grouping variable sample, the first enhanced sample, and the second enhanced sample. That is, the anomaly detection device combines the grouping variable sample, the first enhanced sample, and the second enhanced sample to obtain the grouped enhanced sample of the target sensor group.

[0081] Optionally, after acquiring the grouped enhanced samples, the semiconductor diagnostic equipment also performs fault injection on the grouped enhanced samples to generate controllable labeled anomaly data. That is, the anomaly detection uses a fault injection model and a target fault mode to perform fault injection processing on the grouped enhanced samples to obtain target grouped samples.

[0082] Optionally, the semiconductor diagnostic equipment pre-establishes an abnormal mode experience library to record the mapping relationship between common target fault modes and variable characteristics of grouped variable samples. After generating grouped enhanced samples, the grouped enhanced samples are configured through target fault modes to obtain grouped fault samples. Here, the target fault mode is the manifestation of the fault type and characteristics that may occur in the process variables of the target sensor group.

[0083] Optionally, in one specific embodiment, the semiconductor diagnostic device accesses a fault mode library, obtains the target fault mode corresponding to the sensor group in the fault mode library, configures the enhanced samples of the group using the fault injection model and the target fault mode, obtains the group fault samples corresponding to the group enhanced samples, and generates target group samples based on the group enhanced samples and the group fault samples. That is, the semiconductor diagnostic device inputs the target fault mode into a pre-trained fault injection model, so that the fault injection model learns the mapping relationship between the target fault mode and variable features based on the anomaly description of the target fault model, and configures the group enhanced samples based on the mapping relationship to generate group fault samples obtained by changing the group enhanced samples. Here, the group fault samples are anomaly variable samples corresponding to the target fault mode obtained by modifying the group enhanced samples.

[0084] Optionally, after generating the grouped fault samples, the semiconductor diagnostic device further generates target grouped samples based on the grouped enhanced samples and the grouped fault samples. Optionally, in one specific embodiment, the target grouped samples include a first grouped sample containing grouped enhanced samples representing normal sample data and grouped fault samples containing grouped fault samples representing abnormal sample data, and a second grouped sample containing only the grouped enhanced samples.

[0085] Optionally, after acquiring the target group sample, the semiconductor diagnostic device uses the target group sample to train the initial model to obtain the anomaly detection model corresponding to the target sensor group.

[0086] That is, the semiconductor diagnostic equipment extracts features from the target group samples to obtain group sample features. These group sample features include group enhancement features corresponding to the enhanced group samples and group fault features corresponding to the faulty group samples. Specifically, the semiconductor diagnostic equipment uses a pre-set feature extraction module to perform multi-dimensional feature extraction on the target group samples to obtain statistical features, dynamic features, frequency domain features, and correlation features corresponding to the target group samples, and determines these statistical features, dynamic features, frequency domain features, and correlation features as the group sample features of the target group samples.

[0087] Optionally, in one specific embodiment, after acquiring the grouped sample features, the semiconductor diagnostic device also performs feature selection on the grouped sample features. That is, the semiconductor diagnostic device eliminates redundant or low-information grouped sample features through collinearity detection and constant value detection, thereby reducing dimensionality and improving model computation efficiency.

[0088] Optionally, the semiconductor diagnostic device uses the grouped enhancement features and the grouped fault features to perform supervised training on the first initial detection module in the initial model to obtain the first anomaly detection module. That is, the semiconductor diagnostic device uses grouped enhanced samples representing normal sample data and grouped fault samples containing data representing abnormal samples as model inputs and inputs them into the first initial detection module for training to generate the first anomaly detection module.

[0089] Optionally, the semiconductor diagnostic device also utilizes the grouped enhancement features to perform semi-supervised training on the first initial detection module in the initial model to obtain the second anomaly detection module. That is, the semiconductor diagnostic device uses grouped enhanced samples containing only normal sample data to perform semi-supervised training on the first initial detection module in the initial model to generate the second anomaly detection module.

[0090] Optionally, the semiconductor diagnostic equipment also generates an anomaly detection model corresponding to the target sensor group based on the first anomaly detection module and the second anomaly detection module. That is, after training the first and second anomaly detection modules, the semiconductor diagnostic equipment configures the initial model with the model parameters corresponding to the first and second anomaly detection modules to obtain the anomaly detection model. In subsequent steps, the anomaly detection model is used to perform grouped anomaly detection on the target manufacturing system, so as to effectively reduce the influence of irrelevant variables and improve the interpretability and accuracy of the anomaly detection process.

[0091] like Figure 5 As shown, Figure 5 This is a flowchart illustrating an embodiment of updating the anomaly detection model in the semiconductor diagnostic method provided in this application. Figure 5 In the illustrated embodiment, the semiconductor diagnostic method further includes steps 401 to 404: 401. Respond to the detection marking event for the detection result of the system, and count the number of times the detection marking event is marked; 402. If the number of markings is greater than or equal to the number of markings threshold, the sample is updated based on the detected marking event and the target group sample to obtain the updated group sample; 403. Update the anomaly detection model using the updated grouped samples to obtain an updated detection model; 404. Use the update detection model to perform update detection on the grouping process variables to obtain update detection results.

[0092] Based on the above embodiments, in this embodiment, the semiconductor diagnostic equipment can also update the anomaly detection model through the detection marker events and data distribution feedback from the user, obtain an updated detection model, and use the updated detection model to perform updated detection on the target manufacturing system to obtain updated detection results.

[0093] Optionally, after outputting the system test results, the semiconductor diagnostic equipment also responds to the user's test marking event for the system test results and counts the number of times the test marking event is marked.

[0094] Optionally, a detection marking event is a marking feedback event triggered when a user disagrees with the system's detection results. The marking count data represents the number of detection marking events that occur within a statistical period using the semiconductor diagnostic device.

[0095] Optionally, the semiconductor diagnostic device also pre-sets a threshold for the number of times a label is used to trigger a model update of the anomaly detection model, and determines whether the anomaly detection model needs to be updated by comparing the number of labeling data with the threshold number of labeling.

[0096] Optionally, if the number of times the label is recorded is greater than or equal to the labeling threshold, it is determined that the anomaly detection model needs to be updated. The semiconductor diagnostic device updates the samples based on the detected labeling event and the target group samples to obtain updated group samples. That is, the semiconductor diagnostic device generates updated label tags by the labeling results of the detected labeling event, maps the updated label tags to the corresponding group variable features to obtain updated label samples, and merges the updated label samples and the target group samples to form updated group samples.

[0097] Optionally, after acquiring the updated grouped samples, the semiconductor diagnostic device updates the anomaly detection model using the updated grouped samples to obtain an updated detection model. This updated detection model is an artificial intelligence model obtained by online updating the anomaly detection model, used to improve anomaly detection accuracy. Optionally, the training method of this updated detection model is the same as the anomaly detection model training method in the above embodiments, and will not be repeated in this embodiment.

[0098] Optionally, after generating the updated detection model, the semiconductor diagnostic equipment uses the updated detection model to update the variables of the grouping process and obtain the updated detection results.

[0099] Optionally, in another specific embodiment, to identify potential distribution drift risks and avoid data distribution drift caused by equipment aging and environmental changes, leading to model performance degradation over time, the semiconductor diagnostic equipment also periodically detects the variable distribution data of the grouped process variables and the corresponding training distribution data. The variable distribution data consists of data parameters characterizing the data distribution of the grouped process variables. The training distribution data consists of data parameters characterizing the data distribution of historical process variables under the same production conditions as the grouped process variables. This variable distribution data includes distribution data information such as mean drift and variance changes.

[0100] Optionally, after acquiring the variable distribution data and training distribution data, the semiconductor diagnostic device further calculates the variable offset parameter between the variable distribution data and the training distribution data. That is, the semiconductor diagnostic device compares the distribution similarity between the variable distribution data and the training distribution data to determine the variable offset parameter. Optionally, in a specific embodiment, the variable offset parameter can be the cosine similarity or distribution divergence (e.g., KL divergence and JS divergence) between the variable distribution data and the training distribution data.

[0101] Optionally, after obtaining the variable offset parameter, if the variable offset parameter is greater than a preset offset threshold, the semiconductor diagnostic equipment confirms that the grouping process variable has significantly drifted. Therefore, the semiconductor diagnostic equipment updates the anomaly detection model to obtain an updated detection model, thereby enabling the anomaly detection model to adapt to the data distribution shift caused by equipment aging and process drift, thereby improving the anomaly identification accuracy.

[0102] To better implement the semiconductor diagnostic method in the embodiments of this application, based on the semiconductor diagnostic method, the embodiments of this application also provide a semiconductor diagnostic apparatus, such as... Figure 6 As shown, Figure 6 This is a schematic diagram of one embodiment of the semiconductor diagnostic apparatus provided in this application. Specifically, the semiconductor diagnostic apparatus 500 includes: The variable acquisition module 501 is configured to acquire the grouping process variables of each target sensor group in the target manufacturing system in response to an anomaly detection request for the target manufacturing system. Feature extraction module 502 is configured to extract features from the grouping process variables to obtain grouping variable features corresponding to the grouping process variables; Anomaly identification module 503 is configured to use the anomaly detection model corresponding to the target sensor group to perform anomaly detection on the group variable features and obtain variable detection labels corresponding to the group variable features; Anomaly detection module 504 is configured to determine the system detection result of the target manufacturing system based on the variable detection label corresponding to each target sensor group.

[0103] In one possible implementation of this embodiment, the semiconductor diagnostic device uses an anomaly detection model corresponding to the target sensor group to perform anomaly detection on the grouped variable features, and obtains variable detection labels corresponding to the grouped variable features, including: Obtain the anomaly detection model corresponding to the target sensor group, wherein the anomaly detection model includes a first anomaly detection module, a second anomaly detection module, and a label fusion module; The first anomaly detection module is used to perform a first anomaly detection on the grouping variable features to obtain a first variable detection label; The second anomaly detection module is used to perform second anomaly detection on the grouping variable features to obtain the second variable detection label; The label fusion module is used to evaluate the first variable detection label and the second variable detection label to determine the variable detection label corresponding to the grouped variable feature.

[0104] In one possible implementation of this embodiment, the second anomaly detection module in the semiconductor diagnostic device includes at least two different anomaly detection units and a detection fusion unit; the input terminal of the anomaly detection unit is used to receive grouped variable features, the output terminal of the anomaly detection unit is connected to the input terminal of the detection fusion unit, and the output terminal of the detection fusion unit is connected to the tag fusion module as the output terminal of the second anomaly detection module. The semiconductor diagnostic device uses a second anomaly detection module to perform second anomaly detection on the grouping variable features to obtain a second variable detection label, including: The anomaly detection unit is used to calculate the initial feature distribution parameters of the grouped variable features; Based on the detection fusion unit and the initial feature distribution parameters, the variable feature distribution parameters corresponding to the grouped variable features are determined; The feature deviation parameters of the grouped variable features are determined based on the variable feature distribution parameters and the preset target feature distribution parameters; The second variable detection label is generated based on the feature deviation parameter and the preset deviation threshold.

[0105] In one possible implementation of this embodiment, the semiconductor diagnostic device uses the tag fusion module to evaluate the first variable detection tag and the second variable detection tag to determine the variable detection tag corresponding to the grouped variable feature, including: When both the first variable detection label and the second variable detection label are normal detection labels, the normal detection label is determined as the variable detection label corresponding to the grouped variable feature; When either the first variable detection label or the second variable detection label is an anomaly detection label, the anomaly detection label is determined as the variable detection label corresponding to the grouped variable feature.

[0106] In one possible implementation of this embodiment, before the semiconductor diagnostic device performs anomaly detection on the grouped variable features using the anomaly detection model corresponding to the target sensor group, it further includes: The sensor devices in the target manufacturing system are grouped to obtain the target sensor grouping corresponding to the target manufacturing system; Obtain the grouping variable samples corresponding to the target sensor group, and use the grouping variable samples to train the initial model to obtain the anomaly detection model corresponding to the target sensor group.

[0107] In one possible implementation of this embodiment, the semiconductor diagnostic device acquires grouping variable samples corresponding to the target sensor group, and uses the grouping variable samples to train an initial model to obtain an anomaly detection model corresponding to the target sensor group, including: Obtain the original variable samples corresponding to the target manufacturing system, and perform data separation on the original variable samples based on the machine modules and / or signal types of the target sensor group to obtain the grouped variable samples of the target sensor group; The grouping variable samples are augmented to obtain group-augmented samples; The target grouped samples are obtained by performing fault injection processing on the grouped enhanced samples using a fault injection model and a target fault mode. The initial model is trained using the target group samples to obtain the anomaly detection model corresponding to the target sensor group.

[0108] In one possible implementation of this embodiment, the semiconductor diagnostic device performs quantity enhancement on the grouped variable samples to obtain grouped enhanced samples, including: Based on the grouping variable samples, a similarity search is performed on the sample database associated with the target manufacturing system to obtain the first enhanced sample associated with the grouping variable samples in the sample database; Using the data augmentation model and the grouping variable samples, a second augmented sample is generated. Grouped augmented samples are generated based on the grouping variable samples, the first augmented sample, and the second augmented sample.

[0109] In one possible implementation of this embodiment, the semiconductor diagnostic device uses a fault injection model and a target fault mode to perform fault injection processing on the grouped enhanced samples to obtain target grouped samples, including: Access the fault mode library and obtain the target fault mode corresponding to the sensor group in the fault mode library; The grouped enhanced samples are configured using the fault injection model and the target fault mode to obtain the grouped fault samples corresponding to the grouped enhanced samples; Target group samples are generated based on the group enhancement samples and the group failure samples.

[0110] In one possible implementation of this embodiment, the semiconductor diagnostic device uses the target group samples to train an initial model to obtain an anomaly detection model corresponding to the target sensor group, including: Feature extraction is performed on the target grouped samples to obtain grouped sample features, which include group enhancement features corresponding to group enhancement samples and group fault features corresponding to group fault samples. The first initial detection module in the initial model is trained under supervision using the grouped enhancement features and the grouped fault features to obtain the first anomaly detection module. The first initial detection module in the initial model is semi-supervised and trained using the grouped enhancement features to obtain the second anomaly detection module. An anomaly detection model corresponding to the target sensor group is generated based on the first anomaly detection module and the second anomaly detection module.

[0111] In one possible implementation of this embodiment, after the semiconductor diagnostic device determines the system detection result of the target manufacturing system based on the variable detection tag corresponding to each target sensor group, it further includes: In response to detection tagging events related to the system's detection results, the system counts the number of tagging events for each detection tagging event. If the number of markings is greater than or equal to the number of markings threshold, the sample is updated based on the detected marking event and the target group sample to obtain the updated group sample; The anomaly detection model is updated using the updated grouped samples to obtain an updated detection model; The update detection model is used to perform update detection on the grouping process variables to obtain update detection results.

[0112] In one possible implementation of this embodiment, after the semiconductor diagnostic device determines the system detection result of the target manufacturing system based on the variable detection tag corresponding to each target sensor group, it further includes: Obtain the variable distribution data of the grouping process variables, and the training distribution data corresponding to the grouping process variables; Calculate the variable offset parameter of the grouping process variable based on the variable distribution data and the training distribution data; If the variable offset parameter is greater than the preset offset threshold, the anomaly detection model is updated to obtain an updated detection model.

[0113] In this embodiment, the semiconductor diagnostic device, in response to an anomaly detection request for a target manufacturing system, acquires the grouping process variables of each target sensor group in the target manufacturing system; extracts features from the grouping process variables to obtain grouping variable features corresponding to the grouping process variables; performs anomaly detection on the grouping variable features using the anomaly detection model corresponding to the target sensor group to obtain variable detection labels corresponding to the grouping variable features; and determines the system detection result of the target manufacturing system based on the variable detection labels corresponding to each target sensor group. This achieves the following: during the anomaly detection process of the target manufacturing system, sensors in the target manufacturing system are grouped according to sensor mechanisms, and the grouping process variables corresponding to each target sensor group are collected. Then, the grouping variable features and anomaly detection model are used to perform grouped anomaly detection to determine the variable detection labels of each target sensor group. Based on these variable detection labels, the system detection result is determined to determine whether there are variable anomalies in the target manufacturing system. This effectively reduces the influence of irrelevant variables and improves the interpretability and accuracy of the anomaly detection process.

[0114] This invention also provides a semiconductor diagnostic device, such as... Figure 7 As shown, Figure 7 This is a schematic diagram of one embodiment of the semiconductor diagnostic device provided in this application.

[0115] The semiconductor diagnostic device integrates any of the semiconductor diagnostic apparatuses provided in the embodiments of the present invention, and the semiconductor diagnostic device includes: One or more processors; Memory; and One or more applications, wherein the one or more applications are stored in the memory and configured by the processor to perform the steps of the semiconductor diagnostic method described in any of the embodiments of the above semiconductor diagnostic method.

[0116] Specifically, a semiconductor diagnostic device may include components such as a processor 601 with one or more processing cores, a memory 602 with one or more computer-readable storage media, a power supply 603, and an input unit 604. Those skilled in the art will understand that... Figure 7 The semiconductor diagnostic device structure shown does not constitute a limitation on the semiconductor diagnostic device and may include more or fewer components than shown, or combine certain components, or have different component arrangements. Wherein: The processor 601 is the control center of the semiconductor diagnostic equipment. It connects various parts of the equipment via various interfaces and lines, and performs various functions and processes data by running or executing software programs and / or modules stored in the memory 602, and by calling data stored in the memory 602, thereby providing overall monitoring of the semiconductor diagnostic equipment. Optionally, the processor 601 may include one or more processing cores; preferably, the processor 601 may integrate an application processor and a modem processor, wherein the application processor mainly handles the operating system, user interface, and applications, and the modem processor mainly handles wireless communication. It is understood that the modem processor may not be integrated into the processor 601.

[0117] The memory 602 can be used to store software programs and modules. The processor 601 executes various functional applications and performs anomaly detection by running the software programs and modules stored in the memory 602. The memory 602 may mainly include a program storage area and a data storage area. The program storage area may store the operating system, application programs required for at least one function (such as sound playback function, image playback function, etc.), etc.; the data storage area may store data created based on the use of the semiconductor diagnostic equipment, etc. In addition, the memory 602 may include high-speed random access memory, and may also include non-volatile memory, such as at least one disk storage device, flash memory device, or other volatile solid-state storage device. Accordingly, the memory 602 may also include a memory controller to provide the processor 601 with access to the memory 602.

[0118] The semiconductor diagnostic equipment also includes a power supply 603 that supplies power to the various components. Preferably, the power supply 603 can be logically connected to the processor 601 through a power management system, thereby enabling functions such as charging, discharging, and power consumption management through the power management system. The power supply 603 may also include one or more DC or AC power supplies, recharging systems, power fault detection circuits, power converters or inverters, power status indicators, and other arbitrary components.

[0119] The semiconductor diagnostic device may also include an input unit 604, which can be used to receive input digital or character information and generate keyboard, mouse, joystick, optical or trackball signal inputs related to user settings and function control.

[0120] Although not shown, the semiconductor diagnostic device may also include a display unit, etc., which will not be described in detail here. Specifically, in this embodiment, the processor 601 in the semiconductor diagnostic device loads the executable files corresponding to the processes of one or more application programs into the memory 602 according to the following instructions, and the processor 601 runs the application programs stored in the memory 602 to realize various functions, as follows: In response to an anomaly detection request for the target manufacturing system, obtain the grouping process variables of each target sensor group in the target manufacturing system; Feature extraction is performed on the grouping process variables to obtain the grouping variable features corresponding to the grouping process variables; Anomaly detection is performed on the grouped variable features using the anomaly detection model corresponding to the target sensor group, and variable detection labels corresponding to the grouped variable features are obtained. Based on the variable detection label corresponding to each target sensor group, the system detection result of the target manufacturing system is determined.

[0121] Therefore, embodiments of the present invention provide a computer-readable storage medium, which may include: read-only memory (ROM), random access memory (RAM), magnetic disk, or optical disk, etc. A computer program is stored thereon, which is loaded by a processor to execute the steps in any of the semiconductor diagnostic methods provided in the embodiments of the present invention. For example, the computer program loaded by the processor can execute the following steps: In response to an anomaly detection request for the target manufacturing system, obtain the grouping process variables of each target sensor group in the target manufacturing system; Feature extraction is performed on the grouping process variables to obtain the grouping variable features corresponding to the grouping process variables; Anomaly detection is performed on the grouped variable features using the anomaly detection model corresponding to the target sensor group, and variable detection labels corresponding to the grouped variable features are obtained. Based on the variable detection label corresponding to each target sensor group, the system detection result of the target manufacturing system is determined.

[0122] In the above embodiments, the descriptions of each embodiment have different focuses. For parts not described in detail in a certain embodiment, please refer to the detailed descriptions of other embodiments above, which will not be repeated here.

[0123] In practice, each of the above units or structures can be implemented as an independent entity or can be arbitrarily combined to be implemented as the same or several entities. For the specific implementation of each of the above units or structures, please refer to the previous method embodiments, which will not be repeated here.

[0124] For details on the implementation of each of the above operations, please refer to the previous examples, which will not be repeated here.

[0125] The above provides a detailed description of a multivariate-based semiconductor diagnostic method provided by the embodiments of this application. Specific embodiments have been used to illustrate the principles and implementation methods of the present invention. The description of the above embodiments is only for the purpose of helping to understand the method and core ideas of the present invention. At the same time, for those skilled in the art, there will be changes in the specific implementation methods and application scope based on the ideas of the present invention. Therefore, the content of this specification should not be construed as a limitation of the present invention.

Claims

1. A semiconductor diagnostic method based on multivariate analysis, characterized in that, The semiconductor diagnostic method includes: In response to an anomaly detection request for the target manufacturing system, obtain the grouping process variables of each target sensor group in the target manufacturing system; Feature extraction is performed on the grouping process variables to obtain the grouping variable features corresponding to the grouping process variables; Anomaly detection is performed on the grouped variable features using the anomaly detection model corresponding to the target sensor group, and variable detection labels corresponding to the grouped variable features are obtained. Based on the variable detection label corresponding to each target sensor group, the system detection result of the target manufacturing system is determined.

2. The semiconductor diagnostic method according to claim 1, characterized in that, The step of using the anomaly detection model corresponding to the target sensor group to perform anomaly detection on the grouped variable features, and obtaining the variable detection label corresponding to the grouped variable features, includes: Obtain the anomaly detection model corresponding to the target sensor group, wherein the anomaly detection model includes a first anomaly detection module, a second anomaly detection module, and a label fusion module; The first anomaly detection module is used to perform a first anomaly detection on the grouping variable features to obtain a first variable detection label; The second anomaly detection module is used to perform second anomaly detection on the grouping variable features to obtain the second variable detection label; The label fusion module is used to evaluate the first variable detection label and the second variable detection label to determine the variable detection label corresponding to the grouped variable feature.

3. The semiconductor diagnostic method according to claim 2, characterized in that, The second anomaly detection module includes at least two different anomaly detection units and a detection fusion unit; the input of the anomaly detection unit is used to receive grouped variable features, the output of the anomaly detection unit is connected to the input of the detection fusion unit, and the output of the detection fusion unit is connected to the tag fusion module as the output of the second anomaly detection module. The step of using a second anomaly detection module to perform second anomaly detection on the grouped variable features to obtain a second variable detection label includes: The anomaly detection unit is used to calculate the initial feature distribution parameters of the grouped variable features; Based on the detection fusion unit and the initial feature distribution parameters, the variable feature distribution parameters corresponding to the grouped variable features are determined; The feature deviation parameters of the grouped variable features are determined based on the variable feature distribution parameters and the preset target feature distribution parameters; The second variable detection label is generated based on the feature deviation parameter and the preset deviation threshold.

4. The semiconductor diagnostic method according to claim 2, characterized in that, The step of evaluating the first variable detection label and the second variable detection label using the label fusion module to determine the variable detection label corresponding to the grouped variable feature includes: When both the first variable detection label and the second variable detection label are normal detection labels, the normal detection label is determined as the variable detection label corresponding to the grouped variable feature; When either the first variable detection label or the second variable detection label is an anomaly detection label, the anomaly detection label is determined as the variable detection label corresponding to the grouped variable feature.

5. The semiconductor diagnostic method according to any one of claims 1-4, characterized in that, Before using the anomaly detection model corresponding to the target sensor group to perform anomaly detection on the grouped variable features, the method further includes: The sensor devices in the target manufacturing system are grouped to obtain the target sensor grouping corresponding to the target manufacturing system; Obtain the grouping variable samples corresponding to the target sensor group, and use the grouping variable samples to train the initial model to obtain the anomaly detection model corresponding to the target sensor group.

6. The semiconductor diagnostic method according to claim 5, characterized in that, The step of obtaining the grouping variable samples corresponding to the target sensor group, and using the grouping variable samples to train the initial model to obtain the anomaly detection model corresponding to the target sensor group includes: Obtain the original variable samples corresponding to the target manufacturing system, and perform data separation on the original variable samples based on the machine modules and / or signal types of the target sensor group to obtain the grouped variable samples of the target sensor group; The grouping variable samples are augmented to obtain group-augmented samples; The target grouped samples are obtained by performing fault injection processing on the grouped enhanced samples using a fault injection model and a target fault mode. The initial model is trained using the target group samples to obtain the anomaly detection model corresponding to the target sensor group.

7. The semiconductor diagnostic method according to claim 6, characterized in that, The step of augmenting the grouping variable samples to obtain group-augmented samples includes: Based on the grouping variable samples, a similarity search is performed on the sample database associated with the target manufacturing system to obtain the first enhanced sample associated with the grouping variable samples in the sample database; Using the data augmentation model and the grouping variable samples, a second augmented sample is generated. Grouped augmented samples are generated based on the grouping variable samples, the first augmented sample, and the second augmented sample.

8. The semiconductor diagnostic method according to claim 6, characterized in that, The process of performing fault injection processing on the grouped enhanced samples using a fault injection model and a target fault mode to obtain target grouped samples includes: Access the fault mode library and obtain the target fault mode corresponding to the target sensor group in the fault mode library; The grouped enhanced samples are configured using the fault injection model and the target fault mode to obtain the grouped fault samples corresponding to the grouped enhanced samples; Target group samples are generated based on the group enhancement samples and the group failure samples.

9. The semiconductor diagnostic method according to claim 6, characterized in that, The step of training the initial model using the target group samples to obtain the anomaly detection model corresponding to the target sensor group includes: Feature extraction is performed on the target grouped samples to obtain grouped sample features, which include group enhancement features corresponding to group enhancement samples and group fault features corresponding to group fault samples. The first initial detection module in the initial model is trained under supervision using the grouped enhancement features and the grouped fault features to obtain the first anomaly detection module. The first initial detection module in the initial model is semi-supervised and trained using the grouped enhancement features to obtain the second anomaly detection module. An anomaly detection model corresponding to the target sensor group is generated based on the first anomaly detection module and the second anomaly detection module.

10. The semiconductor diagnostic method according to claim 1, characterized in that, After determining the system detection result of the target manufacturing system based on the variable detection label corresponding to each target sensor group, the method further includes: In response to detection tagging events related to the system's detection results, the system counts the number of tagging events for each detection tagging event. If the number of markings is greater than or equal to the number of markings threshold, the sample is updated based on the detected marking event and the target group sample to obtain the updated group sample; The anomaly detection model is updated using the updated grouped samples to obtain an updated detection model; The update detection model is used to perform update detection on the grouping process variables to obtain update detection results.

11. The semiconductor diagnostic method according to claim 1, characterized in that, After determining the system detection result of the target manufacturing system based on the variable detection label corresponding to each target sensor group, the method further includes: Obtain the variable distribution data of the grouping process variables, and the training distribution data corresponding to the grouping process variables; Calculate the variable offset parameter of the grouping process variable based on the variable distribution data and the training distribution data; If the variable offset parameter is greater than the preset offset threshold, the anomaly detection model is updated to obtain an updated detection model.

12. A semiconductor diagnostic device, characterized in that, The semiconductor diagnostic device includes: The variable acquisition module is configured to acquire the grouping process variables of each target sensor group in the target manufacturing system in response to an anomaly detection request for the target manufacturing system. The feature extraction module is configured to extract features from the grouping process variables to obtain the grouping variable features corresponding to the grouping process variables; The anomaly detection module is configured to use the anomaly detection model corresponding to the target sensor group to perform anomaly detection on the group variable features and obtain the variable detection label corresponding to the group variable features. An anomaly detection module is configured to determine the system detection result of the target manufacturing system based on the variable detection label corresponding to each target sensor group.

13. A semiconductor diagnostic device, characterized in that, The semiconductor diagnostic device includes: One or more processors; Memory; and One or more applications, wherein the one or more applications are stored in the memory and configured to be executed by the processor to implement the steps of the semiconductor diagnostic method according to any one of claims 1 to 11.

14. A computer-readable storage medium, characterized in that, It stores a computer program, which is loaded by a processor to execute the steps of the semiconductor diagnostic method according to any one of claims 1 to 11.

Images