Abnormality detection method and device, equipment, processor and storage medium
Patent Information
- Application Number
- CN202311183788.X
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2023-09-13
- Publication Date
- 2025-03-14
AI Technical Summary
The accuracy of abnormal detection and recognition in the prior art still needs to be improved during the semiconductor process.
By obtaining the complete tracking data of the sensor to be tested, performing feature engineering processing, obtaining feature data, and using an abnormality detection model (such as the Transformer model) to identify the feature data to determine the location of the abnormality in the data.
The comprehensive detection of all process complete data in the semiconductor process is achieved, the accuracy of abnormal identification and positioning accuracy is improved, and the probability of abnormality not being discovered in time is reduced.
Smart Images
Figure CN119622560A_ABST
Abstract
Description
Technical Field
[0001] Embodiments of the present invention relate to the field of semiconductor manufacturing, and in particular to an abnormality detection method and apparatus, device, processor and storage medium. Background Art
[0002] Precision Fault Detection & Classification (FDC) is an anomaly detection and classification system. As a high-frequency process parameter early warning system, it is in rigid demand in the pan-semiconductor industry. Currently, it mainly uses single-variable and multi-variable control and statistical techniques to achieve real-time anomaly detection, data analysis and classification of equipment in the semiconductor process.
[0003] Compared with iEMS, FDC can provide richer statistical models and monitoring methods and has more powerful functions.
[0004] Currently, the accuracy of FDC in anomaly detection and recognition still needs to be improved. Summary of the invention
[0005] The problem solved by the embodiments of the present invention is to provide an anomaly detection method and apparatus, a device, a processor and a storage medium, which can improve the accuracy of anomaly recognition.
[0006] To solve the above problems, an embodiment of the present invention provides an anomaly detection method, including: obtaining complete tracking data of a sensor to be tested, wherein the complete tracking data includes complete process data in a semiconductor process; performing feature engineering processing on the complete tracking data to obtain feature data; performing identification processing on the feature data to obtain anomalies present in the feature data; and determining the position of the anomaly in the complete tracking data.
[0007] Optionally, the step of performing feature engineering processing on the complete tracking data includes: performing alignment processing on the complete tracking data to obtain an alignment value of the complete tracking data; performing drift processing on the alignment value to obtain a drift value between the alignment value and an ideal process parameter.
[0008] Optionally, the step of performing identification processing on the feature data to determine anomalies present in the feature data includes: performing identification processing on the feature data using an anomaly detection model to obtain anomalies present in the feature data.
[0009] Optionally, the anomaly detection model includes a Transformer anomaly detection model.
[0010] Optionally, the step of determining the position of the anomaly in the feature data includes: obtaining the position coordinates of the anomaly in the feature data, and using the position coordinates as the position of the anomaly in the complete tracking data.
[0011] Optionally, in the step of acquiring the feature data, the feature data includes multiple sub-feature data, the multiple sub-feature data constitute the feature data, and each of the sub-feature data corresponds to a position coordinate; the step of acquiring the position coordinates of the abnormality in the feature data includes: identifying the sub-feature data with the abnormality; and using the position coordinates of the sub-feature data with the abnormality as the position coordinates of the abnormality in the feature data.
[0012] Optionally, the anomalies present in the feature data include one or more of single-point anomalies, associated anomalies, collective anomalies, and multi-dimensional anomalies.
[0013] Optionally, the sensor to be tested is arranged on a semiconductor equipment platform, and the sensor to be tested is used to monitor and record feedback parameters of the semiconductor equipment platform, and the feedback parameters are used as the complete process data.
[0014] Correspondingly, an embodiment of the present invention also provides an anomaly detection device, including: a first acquisition unit, used to acquire complete tracking data of a sensor to be tested, wherein the complete tracking data includes complete process data in a semiconductor process; a data processing unit, used to perform feature engineering processing on the complete tracking data to acquire feature data; an anomaly identification unit, used to perform identification processing on the feature data to acquire anomalies existing in the feature data; and a second acquisition unit, used to determine the position of the anomaly in the complete tracking data based on the anomaly.
[0015] Optionally, the data processing unit includes: an alignment processing subunit, used to perform alignment processing on the complete tracking data to obtain an alignment value of the complete tracking data; and a drift processing subunit, used to perform drift processing on the alignment value to obtain a drift value between the alignment value and an ideal process parameter.
[0016] Correspondingly, an embodiment of the present invention also provides a device, comprising at least one memory and at least one processor, wherein the memory stores one or more computer instructions, wherein the one or more computer instructions are executed by the processor to implement any of the anomaly detection methods described above.
[0017] Correspondingly, an embodiment of the present invention further provides a processor, which is used to execute computer instructions, wherein the computer instructions execute any of the above-mentioned anomaly detection methods when executed.
[0018] Correspondingly, an embodiment of the present invention further provides a storage medium, wherein the storage medium stores one or more computer instructions, and the one or more computer instructions are used to implement any of the above-mentioned anomaly detection methods.
[0019] Compared with the prior art, the technical solution of the embodiment of the present invention has the following advantages:
[0020] The anomaly detection method provided by the embodiment of the present invention obtains complete tracking data of the sensor to be tested, performs feature engineering on the complete tracking data to obtain feature data, performs identification processing on the feature data, obtains anomalies existing in the feature data, and obtains the position of the anomaly in the complete tracking data. Compared with the existing anomaly detection method of obtaining summary data exceeding the control limit (Control Limit) through machine sensors, the embodiment of the present invention can detect all complete process data in the semiconductor process by obtaining the complete tracking data of the sensor to be tested, so as to comprehensively and accurately identify the anomalies appearing in the complete tracking data and the positions of the anomalies, thereby reducing the probability of not timely discovering the anomalies in the complete tracking data, thereby improving the accuracy of identifying the anomalies in the complete tracking data. BRIEF DESCRIPTION OF THE DRAWINGS
[0021] Figure 1 A schematic diagram of a process flow of an abnormality detection method in an embodiment of the present invention is shown;
[0022] Figure 2 A schematic diagram showing normal characteristic data is shown;
[0023] Figure 3 A schematic diagram showing anomalies in feature data;
[0024] Figure 4 A schematic diagram of the structure of an abnormality detection device in an embodiment of the present invention is shown;
[0025] Figure 5 An optional hardware structure diagram of a terminal device provided in an embodiment of the present invention is shown. DETAILED DESCRIPTION
[0026] As can be seen from the background technology, the existing FDC system obtains summary data from each sensor to be tested, and triggers an alarm when the summary data exceeds the control limit, indicating that there is an abnormality. However, in the process of obtaining summary data from each sensor to be tested, the summary data is only the average of a series of data and cannot fully cover all data, so that the FDC system cannot accurately identify the anomalies in all data, thereby reducing the accuracy of identifying anomalies.
[0027] In order to solve the above technical problems, an embodiment of the present invention provides an anomaly detection method, including: obtaining complete tracking data of a sensor to be tested, wherein the complete tracking data includes complete process data in a semiconductor process; performing feature engineering processing on the complete tracking data to obtain feature data; performing identification processing on the feature data to obtain anomalies present in the feature data; and determining the position of the anomaly in the complete tracking data.
[0028] The anomaly detection method provided by the embodiment of the present invention obtains complete tracking data of the sensor to be tested, performs feature engineering on the complete tracking data to obtain feature data, performs identification processing on the feature data, obtains anomalies existing in the feature data, and obtains the position of the anomaly in the complete tracking data. Compared with the existing anomaly detection method of obtaining summary data exceeding the control limit (Control Limit) through machine sensors, the embodiment of the present invention can detect all complete process data in the semiconductor process by obtaining the complete tracking data of the sensor to be tested, so as to comprehensively and accurately identify the anomalies appearing in the complete tracking data and the positions of the anomalies, thereby reducing the probability of not timely discovering the anomalies in the complete tracking data, thereby improving the accuracy of identifying the anomalies in the complete tracking data.
[0029] In order to make the above-mentioned purposes, features and advantages of the embodiments of the present invention more obvious and understandable, specific embodiments of the present invention are described in detail below with reference to the accompanying drawings.
[0030] Figure 1 FIG. 1 is a flow chart of an abnormality detection method in an embodiment of the present invention. Figure 1 , an anomaly detection method may specifically include the following steps:
[0031] Step S1: Acquire complete tracking data of the sensor to be tested, wherein the complete tracking data includes complete process data in a semiconductor process;
[0032] Step S2: performing feature engineering processing on the complete tracking data to obtain feature data;
[0033] Step S3: performing identification processing on the feature data to obtain anomalies existing in the feature data;
[0034] Step S4: Determine the location of the anomaly in the complete tracking data.
[0035] The embodiments of the present invention can detect all complete process data in a semiconductor process by acquiring complete tracking data of the sensor to be tested, thereby being able to comprehensively and accurately identify anomalies appearing in the complete tracking data and the locations of the anomalies, thereby reducing the probability of not promptly discovering anomalies in the complete tracking data, and thereby improving the accuracy of identifying anomalies in the complete tracking data.
[0036] Figure 2 A schematic diagram showing that the characteristic data are normal, Figure 3 A schematic diagram showing anomalies in feature data.
[0037] Combined with reference Figures 1 to 3 , execute step S1 to obtain complete tracking data of the sensor to be tested, wherein the complete tracking data includes complete process data in the semiconductor process.
[0038] Acquire the complete tracking data of the sensor to be tested, and provide a data basis for subsequent feature engineering of the complete tracking data and acquisition of feature data.
[0039] In this embodiment, the sensor to be tested is arranged on a semiconductor equipment platform, and the sensor to be tested is used to monitor and record feedback parameters of the semiconductor equipment platform, and the feedback parameters are used as the complete tracking data.
[0040] It should be noted that semiconductor equipment machines are used to perform various processing techniques on wafers to be processed in the semiconductor process. During the process of performing various processing techniques on the wafers, the sensors to be tested will monitor and record the feedback parameters of the semiconductor equipment machines, and use the feedback parameters as complete tracking data, so as to be able to perform abnormality detection on the complete tracking data, discover problems that specifically occur in a certain processing technique of the wafer to be processed, and further improve the production yield of the wafer to be processed.
[0041] It should also be noted that the feedback parameter refers to a process parameter in the semiconductor manufacturing process, and the feedback parameter is obtained directly from the semiconductor equipment machine by the sensor to be tested.
[0042] In some embodiments, the feedback parameter obtained by the sensor to be tested is initial tracking data. After the initial tracking data is obtained, a data integrity check can be performed on the initial tracking data to determine whether there is any data missing in the data and whether further tracking data acquisition is required.
[0043] In one embodiment, the data integrity check of the initial tracking data may be compared with the standard data of the semiconductor equipment and the semiconductor process technology to determine whether there is data missing. It can be understood by those skilled in the art that this is only an example, and in other embodiments, data integrity check may be performed in other ways, such as by machine learning, and the integrity of the acquired initial data may be determined by a machine learning model, and this application does not limit this.
[0044] In one embodiment, when it is determined that the acquired initial data is missing data, the corresponding data can be reacquired through the sensor based on the data identifier corresponding to the missing data, or the initial data can be reacquired through the sensor and the data integrity test can be performed to obtain complete tracking data.
[0045] Execute step S2 to perform feature engineering processing on the complete tracking data to obtain feature data.
[0046] Specifically, feature engineering is performed on the complete tracking data, which is to perform engineering processing on the complete tracking data to remove impurities and redundancies in the complete tracking data and refine it into feature data, which is used as input for anomaly detection models in subsequent recognition processing.
[0047] It should be noted that by performing feature engineering on the complete tracking data and refining it into feature data, the prediction accuracy of the anomaly detection model can be improved in the subsequent use of the anomaly detection model, thereby further improving the accuracy of identifying anomalies in the complete tracking data.
[0048] In this embodiment, the step of performing feature engineering on the complete tracking data includes: performing alignment processing on the complete tracking data to obtain the alignment value of the complete tracking data; performing drift processing on the alignment value to obtain the drift value between the alignment value and the ideal process parameter.
[0049] It should be noted that the data length of the complete tracking data is different. The alignment processing is performed on the complete tracking data to obtain the alignment value, which is convenient for the calculation of the alignment value and the ideal process parameters (the data length of the ideal process parameters is fixed). At the same time, in the subsequent process of using the anomaly detection model to identify and process the feature data, it is beneficial to the reasoning and prediction of the anomaly detection model.
[0050] It should also be noted that drift processing is to calculate the error value between the alignment value and the ideal process parameters, and use the obtained error value as the drift value. The drift value is a kind of characteristic data. In the subsequent process of using the anomaly detection model to identify and process the characteristic data, it helps to accelerate the learning of the anomaly detection model, so that the anomaly detection model can obtain better recognition accuracy.
[0051] Specifically, the ideal process parameters refer to the most ideal process parameters corresponding to each process in the steps of performing various processes on the wafer to be processed.
[0052] In this embodiment, the obtained drift value is used as feature data corresponding to the complete tracking data.
[0053] In this embodiment, the feature data includes a plurality of sub-feature data, the plurality of sub-feature data constitute the feature data, and each of the sub-feature data corresponds to a position coordinate.
[0054] It should be noted that in the semiconductor process, the wafer to be processed will undergo multiple processing steps, and the sensor to be tested will record the feedback parameters of each processing step. By performing feature engineering on the feedback parameters of each processing step, the sub-feature data of each processing step can be obtained.
[0055] It should also be noted that the multiple processing techniques that the wafer to be processed undergoes have a sequence. Therefore, the sub-feature data corresponding to each processing technique will correspond to a position coordinate. Accordingly, in the subsequent step of obtaining the position of the anomaly in the complete tracking data, the position of the anomaly in the complete tracking data can be determined based on the position coordinates corresponding to the sub-feature data, thereby improving the accuracy of identifying the abnormal position.
[0056] In some embodiments, before aligning the complete tracking data, the complete tracking data may also be standardized, for example, the complete tracking data may be normalized, so that the processed data is standardized and accurate, thereby reducing the difficulty of subsequent data processing and improving the accuracy and efficiency of subsequent data processing.
[0057] Execute step S3 to perform identification processing on the feature data to obtain anomalies existing in the feature data.
[0058] It should be noted that since the feature data is extracted from the complete tracking data, anomalies in the feature data can be discovered by identifying and processing the feature data, which means that the anomalies in the complete tracking data and the locations of the anomalies can be comprehensively and accurately identified, reducing the probability of not discovering anomalies in the complete tracking data in a timely manner, thereby improving the accuracy of identifying anomalies in the complete tracking data.
[0059] In this embodiment, the step of performing identification processing on the feature data and obtaining anomalies existing in the feature data includes: performing identification processing on the feature data using an anomaly detection model and obtaining anomalies existing in the feature data.
[0060] Specifically, correct data is preset in the anomaly detection model. The feature data is identified and processed by using the anomaly detection model to identify inconsistencies between the feature data and the correct data, thereby obtaining anomalies existing in the feature data.
[0061] It should be noted that the anomaly detection model used to identify and process the feature data can be implemented and deployed using a preset artificial intelligence machine learning framework, including the Google TensorFlow framework, the Microsoft Cognitive Toolkit framework, and the like.
[0062] As an example, the anomaly detection model includes a Transformer anomaly detection model.
[0063] Specifically, the Transformer anomaly detection model can identify and process feature data, and can obtain anomalies existing in the feature data and the locations of the anomalies.
[0064] The Transformer model is a deep learning model that uses a self-attention mechanism. The self-attention mechanism can assign different weights according to the importance of each part of the input data (i.e., feature data). At the same time, the Transformer model has the characteristics of unrestricted position-related operations, strong modeling capabilities, strong versatility and strong scalability, and can better perform parallel operations. In other embodiments, the anomaly detection model can also adopt a Kmeans model, an Exponentially Weighted Moving Average (EWMA) model, or an Encoder-Decoder model.
[0065] As an example, the anomalies existing in the feature data include one or more of single-point anomalies, associated anomalies, collective anomalies, and multi-element anomalies.
[0066] Step S4 is executed to determine the position of the anomaly in the complete tracking data.
[0067] Specifically, by obtaining the location of the anomaly in the complete tracking data, in the semiconductor process, it is possible to determine the problem that occurs in a specific processing link of the wafer to be processed, so that the process technology can be adjusted according to the problem, thereby improving the production yield of the wafer to be processed.
[0068] In this embodiment, the step of determining the position of the abnormality in the feature data includes: obtaining the position coordinates of the abnormality in the feature data, and using the position coordinates as the position of the abnormality in the complete tracking data.
[0069] As an example, the step of obtaining the position coordinates of the abnormality in the feature data includes: identifying the sub-feature data with the abnormality; and using the position coordinates of the sub-feature data with the abnormality as the position coordinates of the abnormality in the feature data.
[0070] In this embodiment, the Transformer anomaly detection model is used to identify the sub-feature data with anomalies.
[0071] like Figure 2 , which shows a schematic diagram showing that the characteristic data are all normal.
[0072] The step in the figure represents: the position coordinates corresponding to the sub-feature data.
[0073] As an example, the figure shows the first sub-feature data, the second sub-feature data and the third sub-feature data, the position coordinates step1 corresponding to the first sub-feature data, the position coordinates step2 corresponding to the second sub-feature data, and the position coordinates step3 corresponding to the third sub-feature data.
[0074] like Figure 3 , which shows a schematic diagram of anomalies in feature data.
[0075] The step in the figure represents: the position coordinates corresponding to the sub-feature data.
[0076] As an example, the figure shows the first sub-feature data, the second sub-feature data and the third sub-feature data, the position coordinates step1 corresponding to the first sub-feature data, the position coordinates step2 corresponding to the second sub-feature data, and the position coordinates step3 corresponding to the third sub-feature data.
[0077] It can be seen that there is a deviation between the second sub-feature data and the normal data, that is, there is an anomaly in the second sub-feature data. Since the position coordinate corresponding to the second sub-feature data is step2, the position coordinate step2 can be used as the position of the anomaly in the complete tracking data.
[0078] Accordingly, an embodiment of the present invention further provides an abnormality detection device. Figure 4 A schematic structural diagram of an abnormality detection device in an embodiment of the present invention is shown.
[0079] The anomaly detection device includes: a first acquisition unit 300, used to acquire complete tracking data of the sensor to be tested, wherein the complete tracking data includes complete process data in the semiconductor process; a data processing unit 310, used to perform feature engineering processing on the complete tracking data to acquire feature data; an anomaly recognition unit 320, used to perform recognition processing on the feature data to acquire anomalies existing in the feature data; and a second acquisition unit 330, used to determine the position of the anomaly in the complete tracking data based on the anomaly.
[0080] The first acquisition unit 300 is used to acquire complete tracking data of the sensor to be tested, so as to provide a data basis for subsequent feature engineering of the complete tracking data and acquisition of feature data.
[0081] In this embodiment, the sensor to be tested is arranged on a semiconductor equipment platform, and the sensor to be tested is used to monitor and record feedback parameters of the semiconductor equipment platform, and the feedback parameters are used as the complete tracking data.
[0082] It should be noted that semiconductor equipment machines are used to perform various processing techniques on wafers to be processed in the semiconductor process. During the process of performing various processing techniques on the wafers, the sensors to be tested will monitor and record the feedback parameters of the semiconductor equipment machines, and use the feedback parameters as complete tracking data, so as to be able to perform abnormality detection on the complete tracking data, discover problems that specifically occur in a certain processing technique of the wafer to be processed, and further improve the production yield of the wafer to be processed.
[0083] It should also be noted that the feedback parameter refers to a process parameter in the semiconductor manufacturing process, and the feedback parameter is obtained directly from the semiconductor equipment machine by the sensor to be tested.
[0084] In some embodiments, the feedback parameter obtained by the sensor to be tested is initial tracking data. After the initial tracking data is obtained, a data integrity check can be performed on the initial tracking data to determine whether there is any data missing in the data and whether further tracking data acquisition is required.
[0085] In one embodiment, the data integrity check of the initial tracking data may be compared with the standard data of the semiconductor equipment and the semiconductor process technology to determine whether there is data missing. It can be understood by those skilled in the art that this is only an example, and in other embodiments, data integrity check may be performed in other ways, such as by machine learning, and the integrity of the acquired initial data may be determined by a machine learning model, and this application does not limit this.
[0086] In one embodiment, when it is determined that the acquired initial data is missing data, the corresponding data can be reacquired through the sensor based on the data identifier corresponding to the missing data, or the initial data can be reacquired through the sensor and the data integrity test can be performed to obtain complete tracking data.
[0087] The data processing unit 310 is used to perform feature engineering on the complete tracking data to obtain feature data.
[0088] Specifically, feature engineering is performed on the complete tracking data, which is to perform engineering processing on the complete tracking data to remove impurities and redundancies in the complete tracking data and refine it into feature data, which is used as input for anomaly detection models in subsequent recognition processing.
[0089] It should be noted that by performing feature engineering on the complete tracking data and refining it into feature data, the prediction accuracy of the anomaly detection model can be improved in the subsequent use of the anomaly detection model, thereby further improving the accuracy of identifying anomalies in the complete tracking data.
[0090] In this embodiment, the data processing unit 310 includes: an alignment processing subunit (not shown in the figure), which is used to perform alignment processing on the complete tracking data and obtain the alignment value of the complete tracking data; a drift processing subunit (not shown in the figure), which is used to perform drift processing on the alignment value and obtain the drift value between the alignment value and the ideal process parameter.
[0091] It should be noted that the data length of the complete tracking data is different. The alignment processing is performed on the complete tracking data to obtain the alignment value, which is convenient for the calculation of the alignment value and the ideal process parameters (the data length of the ideal process parameters is fixed). At the same time, in the subsequent process of using the anomaly detection model to identify and process the feature data, it is beneficial to the reasoning and prediction of the anomaly detection model.
[0092] It should also be noted that drift processing is to calculate the error value between the alignment value and the ideal process parameters, and use the obtained error value as the drift value. The drift value is a kind of characteristic data. In the subsequent process of using the anomaly detection model to identify and process the characteristic data, it helps to accelerate the learning of the anomaly detection model, so that the anomaly detection model can obtain better recognition accuracy.
[0093] Specifically, the ideal process parameters refer to the most ideal process parameters corresponding to each process in the steps of performing various processes on the wafer to be processed.
[0094] In this embodiment, the obtained drift value is used as feature data corresponding to the complete tracking data.
[0095] In this embodiment, the feature data includes a plurality of sub-feature data, the plurality of sub-feature data constitute the feature data, and each of the sub-feature data corresponds to a position coordinate.
[0096] It should be noted that in the semiconductor process, the wafer to be processed will undergo multiple processing steps, and the sensor to be tested will record the feedback parameters of each processing step. By performing feature engineering on the feedback parameters of each processing step, the sub-feature data of each processing step can be obtained.
[0097] It should also be noted that the multiple processing techniques that the wafer to be processed undergoes have a sequence. Therefore, the sub-feature data corresponding to each processing technique will correspond to a position coordinate. Accordingly, in the subsequent step of obtaining the position of the anomaly in the complete tracking data, the position of the anomaly in the complete tracking data can be determined based on the position coordinates corresponding to the sub-feature data, thereby improving the accuracy of identifying the abnormal position.
[0098] In some embodiments, before aligning the complete tracking data, the complete tracking data may also be standardized, for example, the complete tracking data may be normalized, so that the processed data is standardized and accurate, thereby reducing the difficulty of subsequent data processing and improving the accuracy and efficiency of subsequent data processing.
[0099] The abnormality identification unit 320 is used to identify and process the feature data to obtain abnormalities existing in the feature data.
[0100] It should be noted that since the feature data is extracted from the complete tracking data, anomalies in the feature data can be discovered by identifying and processing the feature data, which means that the anomalies in the complete tracking data and the locations of the anomalies can be comprehensively and accurately identified, reducing the probability of not discovering anomalies in the complete tracking data in a timely manner, thereby improving the accuracy of identifying anomalies in the complete tracking data.
[0101] In this embodiment, the step of performing identification processing on the feature data and obtaining anomalies existing in the feature data includes: performing identification processing on the feature data using an anomaly detection model and obtaining anomalies existing in the feature data.
[0102] Specifically, correct data is preset in the anomaly detection model. The feature data is identified and processed by using the anomaly detection model to identify inconsistencies between the feature data and the correct data, thereby obtaining anomalies existing in the feature data.
[0103] It should be noted that the anomaly detection model used to identify and process the feature data can be implemented and deployed using a preset artificial intelligence machine learning framework, including the Google TensorFlow framework, the Microsoft Cognitive Toolkit framework, and the like.
[0104] As an example, the anomaly detection model includes a Transformer anomaly detection model.
[0105] Specifically, the Transformer anomaly detection model can identify and process feature data, and can obtain anomalies existing in the feature data and the locations of the anomalies.
[0106] The Transformer model is a deep learning model that uses a self-attention mechanism. The self-attention mechanism can assign different weights according to the importance of each part of the input data (i.e., feature data). At the same time, the Transformer model has the characteristics of unrestricted position-related operations, strong modeling capabilities, strong versatility and strong scalability, and can better perform parallel operations.
[0107] In other embodiments, the anomaly detection model may also adopt a Kmeans model, an Exponentially Weighted Moving Average (EWMA) model, or an Encoder-Decoder model.
[0108] As an example, the anomalies existing in the feature data include one or more of single-point anomalies, associated anomalies, collective anomalies, and multi-element anomalies.
[0109] The second acquiring unit 330 is used to determine the position of the exception in the complete tracking data according to the exception.
[0110] Specifically, by obtaining the location of the anomaly in the complete tracking data, in the semiconductor process, it is possible to determine the problem that occurs in a specific processing link of the wafer to be processed, so that the process technology can be adjusted according to the problem, thereby improving the production yield of the wafer to be processed.
[0111] In this embodiment, the step of obtaining the position of the abnormality in the feature data includes: obtaining the position coordinates of the abnormality in the feature data, and using the position coordinates as the position of the abnormality in the complete tracking data.
[0112] As an example, the step of determining the position coordinates of the abnormality in the feature data includes: identifying the sub-feature data with the abnormality; and using the position coordinates of the sub-feature data with the abnormality as the position coordinates of the abnormality in the feature data.
[0113] In this embodiment, the Transformer anomaly detection model is used to identify the sub-feature data with anomalies.
[0114] Accordingly, an embodiment of the present invention further provides a device, comprising at least one memory and at least one processor, wherein the memory stores one or more computer instructions, wherein the one or more computer instructions are executed by the processor to implement the anomaly detection method. The anomaly detection method is described in the above section and will not be described in detail.
[0115] An optional hardware structure of the terminal device provided by the embodiment of the present invention can be as follows Figure 5 As shown, it includes: at least one processor 01, at least one communication interface 02, at least one memory 03 and at least one communication bus 04.
[0116] In the embodiment of the present invention, the number of the processor 01 , the communication interface 02 , the memory 03 , and the communication bus 04 is at least one, and the processor 01 , the communication interface 02 , and the memory 03 communicate with each other through the communication bus 04 .
[0117] The communication interface 02 may be an interface of a communication module for network communication, such as an interface of a GSM module.
[0118] The processor 01 may be a central processing unit CPU, or an application specific integrated circuit ASIC (Application Specific Integrated Circuit), or one or more integrated circuits configured to implement the embodiments of the present invention.
[0119] The memory 03 may include a high-speed RAM memory, and may also include a non-volatile memory (non-volatile memory), such as at least one disk memory.
[0120] The memory 03 stores one or more computer instructions, and the one or more computer instructions are executed by the processor 01 to implement the defect detection method of the embodiment of the present invention.
[0121] It should be noted that the above-mentioned terminal device may also include other devices (not shown) that may not be necessary for understanding the contents disclosed in the embodiments of the present invention; given that these other devices may not be necessary for understanding the contents disclosed in the embodiments of the present invention, the embodiments of the present invention will not introduce them one by one.
[0122] Accordingly, an embodiment of the present invention further provides a processor, the processor is used to run computer instructions, wherein the computer instructions execute an anomaly detection method when they are run. The anomaly detection method is described in the above section and will not be described in detail.
[0123] Accordingly, an embodiment of the present invention further provides a storage medium, wherein the storage medium stores one or more computer instructions, wherein the one or more computer instructions are used to implement the anomaly detection method provided by the embodiment of the present invention. The anomaly detection method is described in the above section and will not be described in detail.
[0124] The embodiments of the present invention can be implemented by various means such as hardware, firmware, software or a combination thereof. In a hardware configuration, the method according to the exemplary embodiment of the present invention can be implemented by one or more application specific integrated circuits (ASICs), digital signal processors (DSPs), digital signal processing devices (DSPDs), programmable logic devices (PLDs), field programmable gate arrays (FPGAs), processors, controllers, microcontrollers, microprocessors, etc.
[0125] In a firmware or software configuration, the embodiments of the present invention may be implemented in the form of modules, processes, functions, etc. The software code may be stored in a memory module and executed by a processor. The memory module is located inside or outside the processor and may send data to and receive data from the processor via various known means.
[0126] Although the present invention is disclosed as above, the present invention is not limited thereto. Any person skilled in the art can make various changes and modifications without departing from the spirit and scope of the present invention. Therefore, the protection scope of the present invention shall be subject to the scope defined by the claims.
Claims
1. An anomaly detection method, characterized in that: include: Acquire complete tracking data of the sensor to be tested, wherein the complete tracking data includes complete process data in a semiconductor process; Performing feature engineering processing on the complete tracking data to obtain feature data; Performing identification processing on the feature data to obtain anomalies present in the feature data; A location of the anomaly in the complete trace data is determined.
2. The anomaly detection method according to claim 1, characterized in that: The step of performing feature engineering processing on the complete tracking data includes: performing alignment processing on the complete tracking data to obtain an alignment value of the complete tracking data; The alignment value is subjected to drift processing to obtain a drift value between the alignment value and an ideal process parameter.
3. The anomaly detection method according to claim 1, characterized in that: The step of performing identification processing on the feature data and obtaining anomalies existing in the feature data includes: performing identification processing on the feature data using an anomaly detection model and obtaining anomalies existing in the feature data.
4. The anomaly detection method according to claim 3, characterized in that: The anomaly detection model includes a Transformer anomaly detection model.
5. The anomaly detection method according to claim 1, characterized in that: The step of determining the position of the abnormality in the feature data includes: obtaining the position coordinates of the abnormality in the feature data, and using the position coordinates as the position of the abnormality in the complete tracking data.
6. The anomaly detection method according to claim 5, characterized in that: In the step of acquiring the feature data, the feature data includes a plurality of sub-feature data, the plurality of sub-feature data constitute the feature data, and each of the sub-feature data corresponds to a position coordinate; The step of determining the position coordinates of the anomaly in the feature data includes: identifying the sub-feature data where the anomaly exists; The position coordinates of the sub-feature data where the abnormality exists are used as the position coordinates of the abnormality in the feature data.
7. The abnormality detection method according to claim 1, characterized in that: The anomalies existing in the feature data include one or more of single-point anomalies, associated anomalies, collective anomalies, and multivariate anomalies.
8. The abnormality detection method according to any one of claims 1 to 7, characterized in that: The sensor to be tested is arranged on a semiconductor equipment platform, and the sensor to be tested is used to monitor and record feedback parameters of the semiconductor equipment platform, and the feedback parameters are used as the complete process data.
9. An abnormality detection device, characterized in that: include: A first acquisition unit, configured to acquire complete tracking data of the sensor to be tested, wherein the complete tracking data includes complete process data in a semiconductor process; A data processing unit, used for performing feature engineering processing on the complete tracking data to obtain feature data; An abnormality identification unit, used to identify and process the characteristic data to obtain abnormalities existing in the characteristic data; The second acquiring unit is configured to determine, according to the exception, a position of the exception in the complete tracking data.
10. The abnormality detection device according to claim 9, characterized in that: The data processing unit comprises: an alignment processing subunit, configured to perform alignment processing on the complete tracking data to obtain an alignment value of the complete tracking data; The drift processing subunit is used to perform drift processing on the alignment value to obtain a drift value between the alignment value and an ideal process parameter.
11. A device, characterized in that: The method comprises at least one memory and at least one processor, wherein the memory stores one or more computer instructions, wherein the one or more computer instructions are executed by the processor to implement the anomaly detection method according to any one of claims 1 to 8.
12. A processor, characterized in that: The processor is used to execute computer instructions, wherein the computer instructions, when executed, execute the anomaly detection method according to any one of claims 1 to 8.
13. A storage medium, characterized in that: The storage medium stores one or more computer instructions, and the one or more computer instructions are used to implement the anomaly detection method according to any one of claims 1 to 8.