Method and device for constructing intelligent smoke monitoring model based on wafer processing

Abstract

The present application proposes a method and device for constructing an intelligent smoke monitoring model based on wafer processing, including the following steps: The present invention proposes a method for constructing an intelligent smoke monitoring model based on wafer processing, including a smoke prediction subnetwork, an equipment prediction subnetwork, a process prediction subnetwork, an attention mechanism subnetwork and a prediction subnetwork; the smoke prediction subnetwork is connected in series by a first number of LSTM blocks, and the smoke data is processed to obtain a first feature; the equipment prediction subnetwork is connected in series by a second number of LSTM blocks to obtain a second feature; the process prediction subnetwork is connected in series by a third number of LSTM blocks to obtain a third feature. The attention mechanism subnetwork performs weighted fusion on the three features, and the prediction subnetwork predicts wafer processing problems and causes based on the fused features. This solution can effectively improve the safety and efficiency of wafer processing, and provide strong support for problem monitoring and cause analysis during wafer processing.

Description

Technical Field

[0001] The present application relates to the field of deep learning, and in particular to a method and device for constructing an intelligent smoke monitoring model based on wafer processing. Background Art

[0002] In the field of semiconductor manufacturing, wafer processing is a complex and critical process. Wafer processing involves multiple stages, including etching, deposition, doping, etc., and each stage requires precise process control to ensure the quality and performance of the final product. However, during the wafer processing, various problems may occur, such as unstable process parameters, equipment failure, environmental changes, etc. These problems may lead to reduced wafer quality, reduced production efficiency, and even product scrapping.

[0003] Traditional wafer processing monitoring relies mainly on manual experience and regular equipment inspections, which have certain limitations. Manual experience is often not accurate and timely enough to capture subtle changes and potential problems. Regular equipment inspections may not be able to detect ongoing problems in a timely manner, causing the problem to worsen.

[0004] With the development of sensor technology and data analysis technology, more and more wafer processing companies have begun to adopt real-time monitoring and data analysis methods to predict and prevent problems. Among them, time series analysis methods have great potential in wafer processing process prediction. As a powerful time series analysis model, long short-term memory network (LSTM) can effectively handle long-term dependencies in time series data, so it is widely used in wafer processing process prediction.

[0005] However, traditional LSTM still has some shortcomings when applied to wafer processing prediction. First, the wafer processing process involves multiple data sources, including smoke sensor data, process parameter data, equipment operation status data, etc. Traditional LSTM often has difficulty in effectively integrating these multimodal data. Secondly, problems in the wafer processing process often have complex dynamic characteristics. Even if a problem is predicted in a certain step, it is impossible to specifically locate which parameter caused it.

[0006] In summary, there is an urgent need for a method that can combine multiple data sources to predict problems that may occur during wafer processing and accurately locate erroneous parameters for correction. Summary of the invention

[0007] The embodiment of the present application provides a method and device for constructing an intelligent smoke monitoring model for wafer processing. By setting different numbers of LSTM blocks in different sub-networks, hierarchical feature processing is performed on the smoke, equipment and process data of wafer processing, so as to accurately predict processing problems and causes and improve the safety and efficiency of wafer processing.

[0008] In a first aspect, an embodiment of the present application provides a method for constructing an intelligent smoke monitoring model based on wafer processing, the method comprising:

[0009] Constructing a smoke prediction subnetwork, wherein the smoke prediction subnetwork is composed of a first number of LSTM blocks connected in series, and the smoke prediction subnetwork is trained to perform feature processing on smoke data acquired in real time to obtain a first feature;

[0010] Constructing a device prediction subnetwork, wherein the device prediction subnetwork is composed of a second number of LSTM blocks connected in series, and the device prediction subnetwork is trained to perform feature processing on a fusion result of the first feature and the device data acquired in real time to obtain a second feature;

[0011] Constructing a process prediction subnetwork, wherein the process prediction subnetwork is composed of a third number of LSTM blocks connected in series, and the process prediction subnetwork is trained to perform feature processing on a fusion result of the second feature and the process data acquired in real time to obtain a third feature;

[0012] Constructing an attention mechanism sub-network, wherein the attention mechanism fusion layer is trained to perform attention calculation on the first feature, the second feature, and the third feature and weightedly fuse them to obtain a fusion feature;

[0013] A prediction subnetwork is constructed, where the prediction subnetwork is trained to make predictions based on the fused features to obtain prediction results, where the prediction results are problems existing in the wafer processing process and the causes of the corresponding problems.

[0014] Among them, the intelligent smoke monitoring model includes a smoke prediction subnetwork, an equipment prediction subnetwork, a process prediction subnetwork, an attention mechanism subnetwork and a prediction subnetwork. The intelligent smoke monitoring model is trained with historically recorded smoke data, equipment data and process data during the training process.

[0015] In a second aspect, an embodiment of the present application provides a device for constructing an intelligent smoke monitoring model based on wafer processing, comprising:

[0016] A smoke prediction subnetwork construction module, used to construct a smoke prediction subnetwork, wherein the smoke prediction subnetwork is composed of a first number of LSTM blocks connected in series, and the smoke prediction subnetwork is trained to perform feature processing on smoke data acquired in real time to obtain a first feature;

[0017] A device prediction subnetwork construction module, used to construct a device prediction subnetwork, wherein the device prediction subnetwork is composed of a second number of LSTM blocks connected in series, and the device prediction subnetwork is trained to perform feature processing on a fusion result of the first feature and the device data acquired in real time to obtain a second feature;

[0018] A process prediction subnetwork construction module, used to construct a process prediction subnetwork, wherein the process prediction subnetwork is composed of a third number of LSTM blocks connected in series, and the process prediction subnetwork is trained to perform feature processing on a fusion result of the second feature and the process data acquired in real time to obtain a third feature;

[0019] An attention mechanism sub-network construction module is used to construct an attention mechanism sub-network, wherein the attention mechanism fusion layer is trained to perform attention calculation on the first feature, the second feature, and the third feature and weightedly fuse them to obtain a fusion feature;

[0020] A prediction subnetwork construction module is used to construct a prediction subnetwork, which is trained to make predictions based on fusion features to obtain prediction results, wherein the prediction results are problems existing in the wafer processing process and the causes of the corresponding problems. The intelligent smoke monitoring model includes a smoke prediction subnetwork, an equipment prediction subnetwork, a process prediction subnetwork, an attention mechanism subnetwork and a prediction subnetwork. The intelligent smoke monitoring model is trained with historically recorded smoke data, equipment data and process data during the training process.

[0021] In a third aspect, an embodiment of the present application provides an electronic device, including a memory and a processor, wherein the memory stores a computer program, and the processor is configured to run the computer program to execute a method for constructing an intelligent smoke monitoring model based on wafer processing.

[0022] In a fourth aspect, an embodiment of the present application provides a readable storage medium, in which a computer program is stored. The computer program includes a program code for controlling a process to execute a process, and the process includes a method for constructing an intelligent smoke monitoring model based on wafer processing.

[0023] The main contributions and innovations of the present invention are as follows:

[0024] The embodiment of the present application constructs an intelligent smoke monitoring model including a smoke prediction subnetwork, an equipment prediction subnetwork, a process prediction subnetwork, an attention mechanism subnetwork, and a prediction subnetwork, which effectively integrates multimodal data such as smoke sensor data, process parameter data, and equipment operation status data, and overcomes the problem that traditional LSTM is difficult to integrate multiple data sources; this solution sets different numbers of LSTM blocks in different subnetworks to perform hierarchical feature processing on the data. The smoke prediction subnetwork uses more LSTM blocks to capture subtle changes in smoke, and the LSTM blocks of the equipment and process prediction subnetworks are gradually reduced, extracting more representative features, reducing detailed interference, improving the generalization ability of the model, and more accurately predicting processing problems and causes; this solution constructs a selective attention module between subnetworks to filter out features that are useless for prediction, ensuring that the data input into subsequent subnetworks is valid, and accurate prediction can be made even when there are fewer LSTM blocks, ensuring the real-time nature of the prediction results.

[0025] Details of one or more embodiments of the present application are set forth in the following drawings and description to make other features, objects, and advantages of the present application more readily apparent. BRIEF DESCRIPTION OF THE DRAWINGS

[0026] The drawings described herein are used to provide a further understanding of the present application and constitute a part of the present application. The illustrative embodiments of the present application and their descriptions are used to explain the present application and do not constitute an improper limitation on the present application. In the drawings:

[0027] Figure 1 is a flow chart of a method for constructing an intelligent smoke monitoring model based on wafer processing according to an embodiment of the present application;

[0028] Figure 2 is a structural diagram of an intelligent smoke monitoring model based on wafer processing according to an embodiment of the present application;

[0029] Figure 3 is a structural diagram of a selection attention module according to an embodiment of the present application;

[0030] Figure 4 is a structural block diagram of a device for constructing an intelligent smoke monitoring model based on wafer processing according to an embodiment of the present application;

[0031] Figure 5 It is a schematic diagram of the hardware structure of an electronic device according to an embodiment of the present application. DETAILED DESCRIPTION

[0032] Exemplary embodiments will be described in detail herein, examples of which are shown in the accompanying drawings. When the following description refers to the drawings, the same numbers in different drawings represent the same or similar elements unless otherwise indicated. The implementations described in the following exemplary embodiments do not represent all implementations consistent with one or more embodiments of this specification. Instead, they are merely examples of devices and methods consistent with some aspects of one or more embodiments of this specification as detailed in the appended claims.

[0033] It should be noted that: in other embodiments, the steps of the corresponding method are not necessarily performed in the order shown and described in this specification. In some other embodiments, the steps included in the method may be more or less than those described in this specification. In addition, a single step described in this specification may be decomposed into multiple steps for description in other embodiments; and multiple steps described in this specification may be combined into a single step for description in other embodiments.

[0034] Embodiment 1

[0035] The embodiment of the present application provides a method for constructing an intelligent smoke monitoring model based on wafer processing, which performs hierarchical feature processing on the smoke, equipment and process data of wafer processing by setting different numbers of LSTM blocks in different sub-networks, accurately predicts processing problems and causes, and improves the safety and efficiency of wafer processing. Specifically, reference Figure 1 as well as Figure 2 , the method comprising:

[0036] Constructing a smoke prediction subnetwork, wherein the smoke prediction subnetwork is composed of a first number of LSTM blocks connected in series, and the smoke prediction subnetwork is trained to perform feature processing on smoke data acquired in real time to obtain a first feature;

[0037] Constructing a device prediction subnetwork, wherein the device prediction subnetwork is composed of a second number of LSTM blocks connected in series, and the device prediction subnetwork is trained to perform feature processing on a fusion result of the first feature and the device data acquired in real time to obtain a second feature;

[0038] Constructing a process prediction subnetwork, wherein the process prediction subnetwork is composed of a third number of LSTM blocks connected in series, and the process prediction subnetwork is trained to perform feature processing on a fusion result of the second feature and the process data acquired in real time to obtain a third feature;

[0039] Constructing an attention mechanism sub-network, wherein the attention mechanism fusion layer is trained to perform attention calculation on the first feature, the second feature, and the third feature and weightedly fuse them to obtain a fusion feature;

[0040] A prediction subnetwork is constructed, where the prediction subnetwork is trained to make predictions based on the fused features to obtain prediction results, where the prediction results are problems existing in the wafer processing process and the causes of the corresponding problems.

[0041] Among them, the intelligent smoke monitoring model includes a smoke prediction subnetwork, an equipment prediction subnetwork, a process prediction subnetwork, an attention mechanism subnetwork and a prediction subnetwork. The intelligent smoke monitoring model is trained with historically recorded smoke data, equipment data and process data during the training process.

[0042] In this solution, in order to train and use the intelligent smoke monitoring model, smoke sensors are installed in the etching, deposition, doping and other stages of wafer processing where smoke is generated to obtain smoke data. Temperature sensors, pressure sensors and speed sensors are installed on the processing equipment to collect the equipment data during the operation of the processing equipment, and the process parameters during the wafer processing process are obtained from the control system of the processing equipment.

[0043] In some specific embodiments, the smoke data, equipment data and process data are time series data. The smoke data is the change of smoke during the wafer processing process, including changes in smoke concentration, changes in smoke generation rate, etc. The equipment data is equipment operating status information, including equipment temperature, equipment pressure, equipment speed, etc. The process data is process parameters during the processing, such as processing time, processing temperature, processing pressure, etc.

[0044] Specifically, the smoke data, equipment data and process data are cleansed, and the cleaned smoke data, equipment data and process data are mapped to the [0, 1] interval and then input into the intelligent smoke monitoring model.

[0045] Specifically, data cleaning can remove outliers and noise in smoke data, equipment data, and process data. Since sensors may fail or have data transmission errors when working, this junk data will affect the training and application of the intelligent smoke monitoring model. Therefore, data cleaning is required to remove this data to ensure the accuracy of the intelligent smoke monitoring model.

[0046] Specifically, mapping the cleaned smoke data, equipment data, and process data to the [0, 1] interval for normalization can eliminate the dimensional differences between different data and improve the training effect of the intelligent smoke monitoring model.

[0047] In some embodiments, the first number is greater than the second number, and the second number is greater than the third number.

[0048] Specifically, due to the large amount of data and the variety of data involved in the wafer processing process, if the smoke prediction subnetwork, the equipment prediction subnetwork and the process prediction subnetwork are all connected in series by a large number of LSTM blocks, it will definitely cause overfitting of the entire intelligent smoke monitoring model and fail to achieve the effect of real-time prediction. Therefore, in this scheme, the smoke prediction subnetwork used first uses the most LSTM blocks to process a large amount of smoke data to capture the most subtle changes and detailed information, while the number of LSTM blocks in the subsequent equipment prediction subnetwork and process prediction subnetwork gradually decreases, so that the equipment prediction subnetwork and the process prediction subnetwork can extract more representative features, reduce unnecessary detail interference, and by gradually reducing the number of LSTM blocks, the model can more concisely represent the essential characteristics of the data, improve the generalization ability, and better adapt to different wafer processing scenarios and data changes.

[0049] In some specific embodiments, the first number is 64, the second number is 32, and the third number is 16.

[0050] Furthermore, the smoke prediction subnetwork obtains a first feature by processing the features of the smoke data. The first feature accurately represents abnormal conditions displayed by the smoke during the processing, such as a sudden increase or decrease in the amount of smoke, and provides an important basis for judging the problem.

[0051] Furthermore, since there may be different processes, materials or environments in the wafer processing process, it is difficult to judge the problem based on simple equipment information. This solution uses equipment data in the equipment prediction subnetwork combined with the first feature output by the smoke prediction subnetwork to make a comprehensive judgment, which can more comprehensively reflect the situation in the wafer processing process. Therefore, the second feature output by the equipment prediction subnetwork can accurately represent abnormal conditions related to the equipment during the processing.

[0052] Furthermore, the same is true for the process prediction sub-network. It is also impossible to use pure process data to judge the problem. However, since process data is closely related to equipment data, this solution uses process data in the process prediction sub-network combined with the second feature output by the equipment prediction sub-network to make a comprehensive judgment. Since the second feature combines the judgment of smoke data, the third feature output by the process prediction sub-network is the most comprehensive. The third feature represents process-related abnormal conditions during the processing process.

[0053] In this solution, a selective attention module is constructed between the smoke prediction subnetwork and the equipment prediction subnetwork, and between the equipment prediction subnetwork and the process prediction subnetwork. The selective attention module is used to fuse the first feature with the equipment data acquired in real time, and to fuse the second feature with the process data acquired in real time.

[0054] Furthermore, the structure of the selected attention module is as follows Figure 3 As shown, the selective attention module is composed of a self-attention calculation layer, a linear calculation layer, a screening layer and a splicing layer. The first feature or the second feature is self-attention calculated in the self-attention calculation layer, and the linear calculation layer maps the self-attention calculation result to the [0, 1] screening interval. The screening layer removes feature vectors less than a set threshold in the screening interval to obtain screening features, and the screening features are fused and spliced ​​with the equipment data or process data in the splicing layer.

[0055] Specifically, since the amount of smoke data, equipment data and process data used in this scheme is very large, and only a part of it can effectively determine the cause of problems in the wafer processing process, this scheme constructs a selective attention module to filter out some features that are useless for prediction to ensure that the data input into the equipment prediction subnetwork and the process prediction subnetwork are valid data. Only in this way can it be ensured that even if there are fewer LSTM blocks in the equipment prediction subnetwork and the process prediction subnetwork, accurate predictions can be made, thereby ensuring the real-time nature of the prediction results.

[0056] In this solution, the attention mechanism subnetwork assigns different weights to the first feature, the second feature, and the third feature through attention calculation and performs weighted fusion. Specifically, the fusion feature obtained by the attention mechanism subnetwork adjusts the importance of the first feature, the second feature, and the third feature according to the actual application situation, so that the prediction result of the intelligent smoke monitoring model is more accurate.

[0057] In this solution, the prediction subnetwork is a fully connected neural network, and the output of the prediction subnetwork is the specific problems that occur during wafer processing and the causes of the corresponding problems, wherein the various causes of the corresponding problems are expressed in the form of percentage probability. For example, the specific problem predicted by the prediction subnetwork is: irregular scratches appear on the wafer surface, and the cause of the problem is: the force of the robot arm in the processing equipment is unstable when grabbing the wafer 60%, the wafer transfer track is uneven 15%, the chemical mechanical polishing (CMP) process parameter setting is unreasonable 15%, and the concentration of the cleaning agent used in the cleaning process is too high or the cleaning time is too long 10%.

[0058] In this solution, the intelligent smoke monitoring model is optimized using the stochastic gradient descent method. Stochastic gradient descent is a commonly used optimization algorithm that minimizes the model's loss function by continuously adjusting the model's parameters. During the training process, appropriate learning rate, batch size and other parameters can be set to improve the training effect. At the same time, techniques such as momentum and adaptive learning rate can be used to speed up the training and improve the stability of the model.

[0059] Specifically, this solution uses methods such as cross-validation to evaluate the performance of the model, adjust the parameters of the model, and improve the accuracy and generalization ability of the model. The cross-validation is a commonly used method for evaluating model performance. The data set is divided into multiple subsets, and each subset is selected as the test set, and the remaining subsets are used as the training set. Multiple training and testing are performed, and finally the average performance is taken as the evaluation result of the model. The generalization ability of the model can be evaluated through cross-validation to avoid overfitting. In the cross-validation process, the parameters of the model, such as the parameters of the LSTM block and the parameters of the attention mechanism, can be adjusted according to the evaluation results to improve the accuracy and generalization ability of the model.

[0060] Embodiment 2

[0061] Based on the same idea, refer to Figure 4 , the present application also proposes a device for constructing an intelligent smoke monitoring model based on wafer processing, comprising:

[0062] A smoke prediction subnetwork construction module, used to construct a smoke prediction subnetwork, wherein the smoke prediction subnetwork is composed of a first number of LSTM blocks connected in series, and the smoke prediction subnetwork is trained to perform feature processing on smoke data acquired in real time to obtain a first feature;

[0063] A device prediction subnetwork construction module, used to construct a device prediction subnetwork, wherein the device prediction subnetwork is composed of a second number of LSTM blocks connected in series, and the device prediction subnetwork is trained to perform feature processing on a fusion result of the first feature and the device data acquired in real time to obtain a second feature;

[0064] A process prediction subnetwork construction module, used to construct a process prediction subnetwork, wherein the process prediction subnetwork is composed of a third number of LSTM blocks connected in series, and the process prediction subnetwork is trained to perform feature processing on a fusion result of the second feature and the process data acquired in real time to obtain a third feature;

[0065] An attention mechanism sub-network construction module is used to construct an attention mechanism sub-network, wherein the attention mechanism fusion layer is trained to perform attention calculation on the first feature, the second feature, and the third feature and weightedly fuse them to obtain a fusion feature;

[0066] A prediction subnetwork construction module is used to construct a prediction subnetwork, which is trained to make predictions based on fusion features to obtain prediction results, wherein the prediction results are problems existing in the wafer processing process and the causes of the corresponding problems. The intelligent smoke monitoring model includes a smoke prediction subnetwork, an equipment prediction subnetwork, a process prediction subnetwork, an attention mechanism subnetwork and a prediction subnetwork. The intelligent smoke monitoring model is trained with historically recorded smoke data, equipment data and process data during the training process.

[0067] Embodiment 3

[0068] This embodiment also provides an electronic device, referring to Figure 5 , comprises a memory 404 and a processor 402, wherein the memory 404 stores a computer program, and the processor 402 is configured to run the computer program to execute the steps in any of the above method embodiments.

[0069] Specifically, the processor 402 may include a central processing unit (CPU), or an application specific integrated circuit (ASIC), or may be configured to implement one or more integrated circuits of the embodiments of the present application.

[0070] Among them, the memory 404 may include a large capacity memory 404 for data or instructions. For example, but not limitation, the memory 404 may include a hard disk drive (HDD), a floppy disk drive, a solid state drive (SSD), a flash memory, an optical disk, a magneto-optical disk, a tape, or a universal serial bus (USB) drive, or a combination of two or more of these. In appropriate cases, the memory 404 may include a removable or non-removable (or fixed) medium. In appropriate cases, the memory 404 may be inside or outside the data processing device. In a specific embodiment, the memory 404 is a non-volatile memory. In a specific embodiment, the memory 404 includes a read-only memory (ROM) and a random access memory (RAM). Where appropriate, the ROM may be a mask-programmed ROM, a programmable ROM (Programmable Read-Only Memory, PROM for short), an erasable PROM (Erasable Programmable Read-Only Memory, EPROM for short), an electrically erasable PROM (Electrically Erasable Programmable Read-Only Memory, EEPROM for short), an electrically alterable ROM (Electrically Alterable Read-Only Memory, EAROM for short) or a flash memory (FLASH) or a combination of two or more of these. In appropriate circumstances, the RAM may be a static random access memory (SRAM) or a dynamic random access memory (DRAM), wherein the DRAM may be a fast page mode dynamic random access memory 404 (FPMDRAM), an extended data output dynamic random access memory (EDODRAM), a synchronous dynamic random access memory (SDRAM), etc.

[0071] The memory 404 may be used to store or cache various data files required for processing and / or communication, as well as possible computer program instructions executed by the processor 402 .

[0072] The processor 402 reads and executes the computer program instructions stored in the memory 404 to implement any one of the methods for constructing an intelligent smoke monitoring model based on wafer processing in the above embodiments.

[0073] Optionally, the electronic device may further include a transmission device 406 and an input / output device 408 , wherein the transmission device 406 is connected to the processor 402 , and the input / output device 408 is connected to the processor 402 .

[0074] The transmission device 406 can be used to receive or send data via a network. Specific examples of the above-mentioned network may include a wired or wireless network provided by a communication provider of the electronic device. In one example, the transmission device includes a network adapter (Network Interface Controller, referred to as NIC), which can be connected to other network devices through a base station so as to communicate with the Internet. In one example, the transmission device 406 can be a radio frequency (Radio Frequency, referred to as RF) module, which is used to communicate with the Internet wirelessly.

[0075] The input and output device 408 is used to input or output information. In this embodiment, the input information may be smoke data, equipment data, process data, etc., and the output information may be fusion features, prediction results, etc.

[0076] Optionally, in this embodiment, the processor 402 may be configured to perform the following steps through a computer program:

[0077] Constructing a smoke prediction subnetwork, wherein the smoke prediction subnetwork is composed of a first number of LSTM blocks connected in series, and the smoke prediction subnetwork is trained to perform feature processing on smoke data acquired in real time to obtain a first feature;

[0078] Constructing a device prediction subnetwork, wherein the device prediction subnetwork is composed of a second number of LSTM blocks connected in series, and the device prediction subnetwork is trained to perform feature processing on a fusion result of the first feature and the device data acquired in real time to obtain a second feature;

[0079] Constructing a process prediction subnetwork, wherein the process prediction subnetwork is composed of a third number of LSTM blocks connected in series, and the process prediction subnetwork is trained to perform feature processing on a fusion result of the second feature and the process data acquired in real time to obtain a third feature;

[0080] Constructing an attention mechanism sub-network, wherein the attention mechanism fusion layer is trained to perform attention calculation on the first feature, the second feature, and the third feature and weightedly fuse them to obtain a fusion feature;

[0081] A prediction subnetwork is constructed, where the prediction subnetwork is trained to make predictions based on the fused features to obtain prediction results, where the prediction results are problems existing in the wafer processing process and the causes of the corresponding problems.

[0082] Among them, the intelligent smoke monitoring model includes a smoke prediction subnetwork, an equipment prediction subnetwork, a process prediction subnetwork, an attention mechanism subnetwork and a prediction subnetwork. The intelligent smoke monitoring model is trained with historically recorded smoke data, equipment data and process data during the training process.

[0083] It should be noted that the specific examples in this embodiment can refer to the examples described in the above embodiments and optional implementation modes, and this embodiment will not be described in detail here.

[0084] In general, various embodiments may be implemented in hardware or dedicated circuits, software, logic, or any combination thereof. Some aspects of the invention may be implemented in hardware, while other aspects may be implemented in firmware or software that may be executed by a controller, microprocessor, or other computing device, but the invention is not limited thereto. Although various aspects of the invention may be shown and described as block diagrams, flow charts, or using some other graphical representation, it should be understood that, as non-limiting examples, the boxes, devices, systems, techniques, or methods described herein may be implemented in hardware, software, firmware, dedicated circuits or logic, general-purpose hardware or controllers or other computing devices, or some combination thereof.

[0085] Embodiments of the present invention may be implemented by computer software that is executable by a data processor of a mobile device, such as in a processor entity, or by hardware, or by a combination of software and hardware. Computer software or programs (also referred to as program products) including software routines, applets and / or macros may be stored in any device-readable data storage medium, and they include program instructions for performing specific tasks. A computer program product may include one or more computer executable components configured to perform an embodiment when the program is run. One or more computer executable components may be at least one software code or a portion thereof. In addition, at this point, it should be noted that, for example, Figure 5 Any block of the logic flow in the program may represent program steps, or interconnected logic circuits, blocks and functions, or a combination of program steps and logic circuits, blocks and functions. The software may be stored on physical media such as memory chips or storage blocks implemented within a processor, magnetic media such as hard disks or floppy disks, and optical media such as, for example, DVDs and their data variants, CDs, etc. Physical media are non-transitory media.

[0086] Those skilled in the art should understand that the technical features of the above embodiments may be arbitrarily combined. To make the description concise, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0087] The above embodiments only express several implementation methods of the present application, and the descriptions thereof are relatively specific and detailed, but they cannot be understood as limiting the scope of the present application. It should be pointed out that, for a person of ordinary skill in the art, several variations and improvements can be made without departing from the concept of the present application, and these all belong to the protection scope of the present application. Therefore, the protection scope of the present application shall be subject to the attached claims.

Claims

1. A method for constructing an intelligent smoke monitoring model based on wafer processing, characterized in that: The following steps are involved: Constructing a smoke prediction subnetwork, wherein the smoke prediction subnetwork is composed of a first number of LSTM blocks connected in series, and the smoke prediction subnetwork is trained to perform feature processing on smoke data acquired in real time to obtain a first feature; Constructing a device prediction subnetwork, wherein the device prediction subnetwork is composed of a second number of LSTM blocks connected in series, and the device prediction subnetwork is trained to perform feature processing on a fusion result of the first feature and the device data acquired in real time to obtain a second feature; Constructing a process prediction subnetwork, wherein the process prediction subnetwork is composed of a third number of LSTM blocks connected in series, and the process prediction subnetwork is trained to perform feature processing on a fusion result of the second feature and the process data acquired in real time to obtain a third feature; Constructing an attention mechanism sub-network, wherein the attention mechanism sub-network is trained to perform attention calculation on the first feature, the second feature, and the third feature and weightedly fuse them to obtain a fused feature; Constructing a prediction subnetwork, wherein the prediction subnetwork is trained to predict according to the fusion features to obtain prediction results, wherein the prediction results are problems existing in the wafer processing process and causes of the corresponding problems; Among them, the intelligent smoke monitoring model includes a smoke prediction subnetwork, an equipment prediction subnetwork, a process prediction subnetwork, an attention mechanism subnetwork and a prediction subnetwork. The intelligent smoke monitoring model is trained with historically recorded smoke data, equipment data and process data during the training process.

2. The method for constructing an intelligent smoke monitoring model based on wafer processing according to claim 1, characterized in that: The first number is greater than the second number, and the second number is greater than the third number.

3. The method for constructing an intelligent smoke monitoring model based on wafer processing according to claim 1, characterized in that: A selective attention module is constructed between the smoke prediction subnetwork and the equipment prediction subnetwork, and between the equipment prediction subnetwork and the process prediction subnetwork. The selective attention module is used to fuse the first feature with the equipment data acquired in real time, and to fuse the second feature with the process data acquired in real time.

4. The method for constructing an intelligent smoke monitoring model based on wafer processing according to claim 3 is characterized in that: The selective attention module is composed of a self-attention calculation layer, a linear calculation layer, a screening layer and a splicing layer. The first feature or the second feature is self-attention calculated in the self-attention calculation layer, and the linear calculation layer maps the self-attention calculation result to the [0, 1] screening interval. The screening layer removes feature vectors less than a set threshold in the screening interval to obtain screening features, and the screening features are fused and spliced ​​with equipment data or process data in the splicing layer.

5. The method for constructing an intelligent smoke monitoring model based on wafer processing according to claim 1, characterized in that: The smoke data, equipment data and process data are time series data. The smoke data is the smoke change during the wafer processing process, the equipment data is the equipment operation status information, and the process data is the process parameters during the processing.

6. The method for constructing an intelligent smoke monitoring model based on wafer processing according to claim 1, characterized in that: The smoke data, equipment data and process data are cleaned, and the cleaned smoke data, equipment data and process data are mapped to the [0, 1] interval and then input into the intelligent smoke monitoring model.

7. The method for constructing an intelligent smoke monitoring model based on wafer processing according to claim 1, characterized in that: The first feature indicates abnormal conditions revealed by smoke during the processing, the second feature indicates abnormal conditions related to the equipment during the processing, and the third feature indicates abnormal conditions related to the process during the processing.

8. A device for constructing an intelligent smoke monitoring model based on wafer processing, characterized in that: include: smoke A prediction subnetwork construction module, used to construct a smoke prediction subnetwork, wherein the smoke prediction subnetwork is composed of a first number of LSTM blocks connected in series, and the smoke prediction subnetwork is trained to perform feature processing on smoke data acquired in real time to obtain a first feature; A device prediction subnetwork construction module, used to construct a device prediction subnetwork, wherein the device prediction subnetwork is composed of a second number of LSTM blocks connected in series, and the device prediction subnetwork is trained to perform feature processing on a fusion result of the first feature and the device data acquired in real time to obtain a second feature; A process prediction subnetwork construction module, used to construct a process prediction subnetwork, wherein the process prediction subnetwork is composed of a third number of LSTM blocks connected in series, and the process prediction subnetwork is trained to perform feature processing on a fusion result of the second feature and the process data acquired in real time to obtain a third feature; An attention mechanism sub-network construction module, used to construct an attention mechanism sub-network, wherein the attention mechanism sub-network is trained to perform attention calculation on the first feature, the second feature, and the third feature and weightedly fuse them to obtain a fused feature; A prediction subnetwork construction module is used to construct a prediction subnetwork, which is trained to make predictions based on fusion features to obtain prediction results, wherein the prediction results are problems existing in the wafer processing process and the causes of the corresponding problems. The intelligent smoke monitoring model includes a smoke prediction subnetwork, an equipment prediction subnetwork, a process prediction subnetwork, an attention mechanism subnetwork and a prediction subnetwork. The intelligent smoke monitoring model is trained with historically recorded smoke data, equipment data and process data during the training process.

9. An electronic device comprising a memory and a processor, characterized in that: The memory stores a computer program, and the processor is configured to run the computer program to execute the method for constructing an intelligent smoke monitoring model based on wafer processing as described in any one of claims 1-7.

10. A readable storage medium, characterized in that: The readable storage medium stores a computer program, which includes a program code for controlling a process to execute a process, and the process includes a method for constructing an intelligent smoke monitoring model based on wafer processing according to any one of claims 1-7.

Images