Lithography machine configuration information processing method, electronic device, and storage medium

Abstract

The application provides a photolithography machine configuration information processing method, an electronic device and a storage medium. The method comprises the following steps: according to a plurality of binary configuration files of photolithography machines obtained, using a pre-trained identification model to analyze and process the binary configuration files, generating configuration parameter sequences and parameter association information of the binary configuration files of the photolithography machines, determining a process parameter difference evaluation score of a current parameter according to attribute information of at least one parameter of the same parameter type in the configuration parameter sequences of different versions, and determining a process stability report of the current parameter according to the attribute information of at least one parameter of the same parameter type in the configuration parameter sequences of different versions and the parameter association information. The application solves the problem that the prior art relies on fixed rules and the analysis is invalid after firmware upgrade, reduces the dependence on expert experience, and improves the identification efficiency.

Description

Technical Field

[0001] This application relates to the field of lithography machine information processing technology, and more specifically, to a lithography machine configuration information processing method, electronic device, and storage medium. Background Technology

[0002] In the semiconductor manufacturing field, lithography machines, as core equipment in the front-end process, directly determine the yield and performance of chip products through their operational stability and process precision. With the continuous shrinking of technology nodes and the increasing complexity of processes, various key parameters are stored and managed using proprietary binary formats. Typical formats include Structured Binary Configuration Format (SBCF) and Multi-modal Calibration Parameter Block (MCPB). These configuration files cover the operational settings of multiple subsystems, such as the lighting system, projection optics, stage control, and environmental compensation, constituting the core configuration data of the equipment. Accurate parsing and effective management of these configuration files have become the technical foundation for ensuring stable equipment operation, achieving process optimization, and rapidly locating faults.

[0003] Currently, the processing of lithography machine configuration files mainly relies on three methods: First, using dedicated software tools provided by the equipment manufacturer, which can read and display some field contents, suitable for basic viewing and simple modification operations; second, engineers conduct manual reverse engineering based on long-term experience, comparing configuration differences under different states and inferring the meaning of fields by combining process knowledge; and third, using scripted semi-automated analysis methods, encoding expert experience into fixed rules for batch identification of the location and type of known fields.

[0004] However, existing technologies have significant limitations. The first approach relies on closed-loop dedicated tools that only support manufacturer-preset operations, cannot deeply analyze unknown or modified formats, and are difficult to extend due to commercial licensing restrictions. The second approach, manual reverse engineering, heavily depends on individual experience, is inefficient, produces inconsistent results, and struggles to form a reusable knowledge system. The third approach, based on fixed-rule scripts, lacks flexibility; once firmware upgrades cause configuration changes, the original rules become invalid, requiring significant new manpower for maintenance. Summary of the Invention

[0005] The purpose of this application is to address the shortcomings of the prior art by providing a method for processing configuration information of a lithography machine, an electronic device, and a storage medium, so as to solve the problems of commercial licensing restrictions, poor flexibility, low efficiency, and low accuracy in the prior art.

[0006] To achieve the above objectives, the technical solution adopted in this application is as follows: In a first aspect, this application provides a method for processing configuration information of a lithography machine, the method comprising: Based on the acquired binary configuration files of multiple lithography machines, a pre-trained recognition model is used to parse the binary configuration files to generate a configuration parameter sequence and parameter association information for each lithography machine's binary configuration file. The configuration parameter sequence includes attribute information of the parameters corresponding to each byte segment in the binary configuration file. The attribute information includes: byte position, parameter type, parameter value, and system to which it belongs. The parameter association information represents the functional association relationship between the parameters corresponding to each byte segment in the configuration parameter sequence. Based on the attribute information of at least one parameter of the same parameter type in different versions of the configuration parameter sequence, the process parameter difference assessment score of the current parameter is determined. The process parameter difference assessment score is used to characterize the degree of influence of the difference of at least one parameter of the same parameter type in different versions on the corresponding process indicators. Based on the attribute information of at least one parameter of the same parameter type in different versions of the configuration parameter sequence and the parameter association information, the process stability report of the current parameter is determined, and the process stability includes a stability score and a warning level.

[0007] Optionally, the step of parsing the binary configuration files of the multiple lithography machines using a pre-trained recognition model to generate a sequence of configuration parameters and parameter association information for each lithography machine's binary configuration file includes: Based on the binary configuration files of the multiple lithography machines, a byte feature sequence is generated. The byte feature sequence includes multiple byte feature vectors arranged in order. The byte feature vectors include byte value feature vectors, position feature vectors, context feature vectors, and system feature vectors. Based on the byte feature sequence, the byte segment information in each binary configuration file is determined based on the recognition model. The byte segment information includes: byte information, context information, and the system to which it belongs. The configuration parameter sequence is determined based on the information of each byte segment; The parameter association information is determined based on the configuration parameter sequence.

[0008] Optionally, determining the information of each byte segment in each binary configuration file based on the recognition model according to the byte feature sequence includes: Based on the bottom attention layer, middle attention layer and high attention layer of the recognition model, feature extraction is performed on the byte feature sequence to determine the boundary score of each byte. Bytes whose boundary scores are greater than a preset boundary score threshold are designated as boundary bytes, and all bytes in two adjacent boundary bytes are treated as a single byte segment. The byte position, number of bytes, and byte value in the byte segment are used as the byte information of the byte segment. Based on the system characteristics of each byte in the byte segment, the system to which the byte segment belongs is determined. The byte information and system to which the adjacent byte segments belong are used as the context information of the byte segment.

[0009] Optionally, determining the configuration parameter sequence based on the byte segment information includes: Based on the byte information in each byte segment, the statistical distribution characteristics of each byte segment are extracted, including entropy, skewness, and kurtosis. Based on the byte information in each byte segment, the system to which it belongs, and the preset parameter ranges, determine the parameter matching range for each byte segment; The format matching range of the byte segment is determined based on the byte value of the byte information in each byte segment information; Based on the statistical distribution characteristics, parameter matching range, and format matching range of each byte segment, the initial parameter type of each byte segment is determined. Based on the context information in each byte segment and the initial parameter type of each byte segment, determine the parameter type and parameter value of the parameter corresponding to each byte segment in the binary configuration file.

[0010] Optionally, determining the parameter association information based on the configuration parameter sequence includes: Based on the configuration parameter sequence, the attention similarity, positional proximity, numerical correlation, and module similarity of the parameters corresponding to each byte segment are determined. The attention similarity is used to characterize the semantic association strength between parameters in the parameter group, the positional proximity is used to characterize the positional distance between parameters in the parameter group, the numerical correlation is used to characterize the synchronicity of changes between the current version and the historical version of the parameters in the parameter group, and the module similarity is used to characterize whether the parameters in the parameter group belong to the same system. The parameter group includes any two parameters. The parameter association information is determined based on the attention similarity, positional proximity, numerical relevance, module similarity, preset attention similarity weight, preset positional proximity weight, preset numerical relevance weight, and preset module similarity weight of the parameters corresponding to each byte segment.

[0011] Optionally, determining the process parameter difference evaluation score of the current parameter based on the attribute information of at least one parameter of the same parameter type in different versions of the configuration parameter sequence includes: The importance weight of the current parameter is determined based on the preset process sensitivity weight, preset system importance weight, preset parameter criticality level, and attribute information of each parameter. The difference value of the current parameter is determined according to the preset process influence distance function corresponding to each parameter and the attribute information of each parameter; Based on the importance weight and difference value of the current parameter, the process parameter difference evaluation score of the current parameter is determined.

[0012] Optionally, determining the process stability report for the current parameter based on the attribute information of at least one parameter of the same parameter type in different versions of the configuration parameter sequence and the parameter association information includes: Based on the attribute information of at least one parameter of the same parameter type in different versions of the configuration parameter sequence, the preset parameter standard deviation, and the preset parameter standard value, determine the stability score of the current parameter; The weight of the trend duration of the current parameter is determined based on the duration of the change trend of each parameter. Based on the parameter correlation information of each parameter, determine the relevant parameters of the current parameter, and based on the changing trends of the relevant parameters and the current parameter, determine the cross-correlation weight of the current parameter; Based on the frequency of anomalies of each parameter, determine the anomaly frequency impact factor of the current parameter; The warning level of the current parameter is determined based on the changing trend of the current parameter, the weight of the trend duration, the weight of the cross-correlation, and the impact factor of the abnormal frequency.

[0013] Optionally, the method further includes: Based on the attribute information of the current parameter, the preset maximum value of each parameter, the preset minimum value of each parameter, and the preset standard value of the parameter, it is determined whether the current parameter exceeds the process window warning, and a health report is generated based on the judgment result.

[0014] Secondly, this application provides an electronic device, including: a processor, a storage medium, and a bus, wherein the storage medium stores machine-readable instructions executable by the processor, and when the electronic device is running, the processor communicates with the storage medium via the bus, and the processor executes the machine-readable instructions to perform the steps of the lithography machine configuration information processing method described above.

[0015] Thirdly, this application provides a computer-readable storage medium storing a computer program, which, when executed by a processor, performs the steps of the lithography machine configuration information processing method described above.

[0016] The beneficial effects of this application are as follows: A recognition model is used to parse and process the binary configuration files of multiple lithography machines, generating a sequence of configuration parameters and parameter association information for each machine's binary configuration file. This automatically identifies field boundaries and structural levels in binary files without requiring manual pre-setting of format rules or reverse engineering, solving the problems of existing technologies relying on fixed rules and parsing failures after firmware upgrades. Furthermore, it reduces reliance on expert experience and improves recognition efficiency. Based on the attribute information of at least one parameter of the same parameter type in different versions of the configuration parameter sequence, the process parameter difference evaluation score of the current parameter is determined, providing real-time monitoring and early warning for lithography machine process stability. This allows for prediction of the impact of parameter changes on key indicators based on the current parameter's process difference evaluation score, reducing yield losses caused by trial and error. Attached Figure Description

[0017] To more clearly illustrate the technical solutions of the embodiments of this application, the accompanying drawings used in the embodiments will be briefly introduced below. It should be understood that the following drawings only show some embodiments of this application and should not be regarded as a limitation of the scope. For those skilled in the art, other related drawings can be obtained based on these drawings without creative effort.

[0018] Figure 1 This is a flowchart illustrating a method for processing configuration information of a lithography machine according to an embodiment of this application; Figure 2 This is a schematic diagram of a process for generating a sequence of configuration parameters and parameter association information provided in an embodiment of this application; Figure 3 This is a flowchart illustrating a process for determining information for each byte segment, provided in an embodiment of this application. Figure 4 This is a flowchart illustrating a method for determining a sequence of configuration parameters provided in an embodiment of this application; Figure 5 This is a flowchart illustrating a method for determining the process parameter difference evaluation score of the current parameter, as provided in an embodiment of this application. Figure 6 This is a schematic flowchart illustrating a process stability report for determining current parameters, provided in an embodiment of this application. Figure 7 This is a schematic diagram of the structure of an electronic device provided in an embodiment of this application. Detailed Implementation

[0019] To make the objectives, technical solutions, and advantages of the embodiments of this application clearer, the technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. It should be understood that the accompanying drawings in this application are for illustrative and descriptive purposes only and are not intended to limit the scope of protection of this application. Furthermore, it should be understood that the schematic drawings are not drawn to scale. The flowcharts used in this application illustrate operations implemented according to some embodiments of this application. It should be understood that the operations in the flowcharts may not be implemented in sequence, and steps without logical contextual relationships may be reversed or implemented simultaneously. In addition, those skilled in the art, guided by the content of this application, may add one or more other operations to the flowcharts, or remove one or more operations from the flowcharts.

[0020] Furthermore, the described embodiments are merely some, not all, of the embodiments of this application. The components of the embodiments of this application described and illustrated herein can typically be arranged and designed in various different configurations. Therefore, the following detailed description of the embodiments of this application provided in the accompanying drawings is not intended to limit the scope of the claimed application, but merely to illustrate selected embodiments of the application. All other embodiments obtained by those skilled in the art based on the embodiments of this application without inventive effort are within the scope of protection of this application.

[0021] It should be noted that the term "comprising" will be used in the embodiments of this application to indicate the presence of the features declared thereafter, but does not exclude the addition of other features.

[0022] In existing technologies, dedicated tools have closed functions and are restricted by commercial licenses, making it impossible to parse unknown or changed formats. Manual reverse engineering relies on individual experience, resulting in low efficiency, poor consistency of results, and difficulty in forming a reusable knowledge system. Fixed rule scripts lack flexibility, and the original rules become invalid after firmware upgrades, requiring a large amount of manpower for maintenance.

[0023] Based on this, this application provides a method for processing configuration information of lithography machines. This method uses a pre-trained recognition model to parse and process the binary configuration files of multiple lithography machines, generating a sequence of configuration parameters and parameter association information for each machine. Based on the attribute information and parameter association information of at least one parameter of the same parameter type in different versions of the configuration parameter sequence, the method determines the process parameter difference assessment score and process stability report for the current parameter. By using a recognition model to parse and process the binary configuration files, it better handles binary configuration files of unknown formats, offering high flexibility. Furthermore, by using the process parameter difference assessment score and process stability report, it quantitatively assesses the process impact of configuration changes.

[0024] Next, refer to Figure 1The steps involved in processing configuration information for a lithography machine are described. Figure 1 This is a flowchart illustrating a method for processing configuration information of a lithography machine according to an embodiment of this application.

[0025] S101. Based on the acquired binary configuration files of multiple lithography machines, the binary configuration files are parsed using a pre-trained recognition model to generate configuration parameter sequences and parameter association information for each lithography machine's binary configuration file. The configuration parameter sequence includes attribute information of parameters corresponding to each byte segment in the binary configuration file. The attribute information includes: byte position, parameter type, parameter value, and system to which it belongs. The parameter association information represents the functional association relationship between parameters corresponding to each byte segment in the configuration parameter sequence.

[0026] Optionally, the lithography machine binary configuration file can be stored in the form of SBCF and / or MCPB. However, the stored raw byte stream lacks intuitive semantics and cannot be directly used for configuration management. Therefore, a recognition model is used to parse the binary configuration file, transforming the discrete byte data into structured, understandable parameter information, while simultaneously determining the functional relationships between parameters.

[0027] The binary configuration file can consist of multiple bytes, with byte values ​​between 0 and 255. Multiple bytes can form a byte segment, and a byte segment can correspond to a parameter in a system. Parameters can be, for example, exposure dose, illumination uniformity, polarization state, etc. in the illumination system; aberration coefficient, lens distortion, focal length compensation, etc. in the projection optics system; positioning accuracy, grid matching, interferometer compensation, etc. in the stage system; temperature compensation, air pressure correction, vibration suppression, etc. in the environmental control system; and mark recognition, position offset, overlay accuracy, etc. in the alignment system.

[0028] Specifically, binary configuration files from multiple lithography machines are acquired, and each file is serialized byte-wise to obtain a continuous byte stream. Multidimensional features of the bytes are constructed to form a byte feature sequence. This sequence is then input into a pre-trained recognition model. The model automatically identifies the boundaries of each sub-stage and delineates independent parameter fields through hierarchical attention calculations. Each parameter field is parsed to determine its byte position, parameter type, parameter value, and system affiliation. These fields are then combined according to the original file order to form a configuration parameter sequence. Finally, based on the configuration parameter sequence and considering factors such as byte position proximity, numerical change synchronization, and system consistency, the functional relationships between parameters are determined, and parameter association information is compiled.

[0029] The byte position in the attribute information includes the start and end offset positions of the corresponding parameter in the binary configuration file. The parameter type indicates the type of the parameter, such as the exposure dose, temperature compensation, and position offset mentioned above. The parameter value indicates the specific value of the parameter and its corresponding unit, for example, a parameter value of 50. The system to which this parameter belongs is the system to which it belongs; for example, the system to which it belongs could be a lighting system.

[0030] S102. Based on the attribute information of at least one parameter of the same parameter type in different versions of the configuration parameter sequence, determine the process parameter difference assessment score of the current parameter. The process parameter difference assessment score is used to characterize the degree of influence of the difference of at least one parameter of the same parameter type in different versions on the corresponding process indicators.

[0031] Among them, process indicators refer to the core photolithography process parameters that determine chip yield and performance, such as overlay accuracy, linewidth uniformity, and exposure uniformity.

[0032] Specifically, different versions of lithography machine configuration files may have parameter changes, and simply comparing parameter value differences cannot determine their actual impact on production. Therefore, by calculating a process parameter difference assessment score, parameter differences can be correlated with process indicators, quantifying the risk level of changes and helping engineers quickly determine whether changes are safe, avoiding blind modifications that could lead to a decrease in yield.

[0033] It is worth noting that different versions can be old or new versions, or versions corresponding to different lithography machines. As long as the parameter types are consistent, there are no restrictions on the version type.

[0034] For example, the process parameter difference evaluation score of the illumination parameters of lithography machine A can be determined based on the attribute information of the illumination parameters of lithography machine A and the attribute information of the illumination parameters of lithography machine B. Alternatively, the process parameter difference evaluation score of the current illumination parameters of lithography machine A can be determined based on the attribute information of the illumination parameters of historical versions of lithography machine A and the attribute information of the illumination parameters of the current version of lithography machine A.

[0035] Specifically, parameters of the same type are selected from the configuration parameter sequences of the two versions, and their attribute information is extracted. Combined with preset process sensitivity weights, system importance weights, and parameter criticality registration, the importance weight of each parameter is calculated comprehensively. Then, a specific preset difference measurement method is used for different types of parameters to calculate the parameter difference value. Finally, the importance weight of each parameter is multiplied by its corresponding difference, and then adjusted based on the frequency of the parameter's use in the current process formulation. The final result is a process parameter difference assessment score for each parameter; a higher score indicates a greater impact of the parameter change on the process.

[0036] S103. Based on the attribute information and parameter association information of at least one parameter of the same parameter type in different versions of the configuration parameter sequence, determine the process stability report of the current parameter. Process stability includes stability score and warning level.

[0037] The process stability report reflects the long-term operational stability of the lithography machine's configuration parameters. The stability score quantifies the operational stability of parameters based on their consistency across different versions; a higher score indicates greater stability. The warning level, based on parameter change trends and the impact of related parameters, classifies risk levels to alert engineers to potential problems.

[0038] Optionally, during long-term operation of a lithography machine, parameters may experience slow drift or abnormal fluctuations. If these are not detected in time, they may lead to process instability and decreased yield. Therefore, by generating a process stability report, the stability status of parameters can be monitored in real time, providing early warnings of potential risks.

[0039] Specifically, firstly, attribute information of the same parameter type is extracted from multiple versions of the configuration parameter sequence. Then, the parameter standard deviation and parameter standard value are referenced to statistically analyze the degree of deviation between the parameter value and the standard value within the time window. Combining the consistency of the deviation, the stability score of the parameter is calculated. Parameters with small deviation and regular fluctuations have higher scores.

[0040] Simultaneously, the number of versions of statistical parameters that maintain the same direction of change is used to determine the trend duration weight; the longer the duration, the higher the weight. Then, based on parameter correlation information, related parameters of the current parameter are identified, and it is determined whether there is a coordinated change between the related parameters and the current parameter, thus determining the cross-correlation weight; the more significant the coordinated change, the higher the weight. Next, based on historical data, the number of times the parameter's value exceeds the reasonable range is counted to determine the anomaly frequency impact factor; the higher the frequency, the larger the impact factor. Finally, by combining the trend duration weight, cross-correlation weight, and anomaly frequency impact factor, the warning level is determined.

[0041] As an optional implementation, stability scores and warning levels can be integrated to generate a process stability report.

[0042] In this embodiment, a recognition model is used to parse the binary configuration files of multiple lithography machines, generating a sequence of configuration parameters and parameter association information for each machine. This automatically identifies field boundaries and structural levels in the binary files without requiring manual pre-setting of format rules or reverse engineering, solving the problems of existing technologies relying on fixed rules and parsing failures after firmware upgrades. It also reduces reliance on expert experience and improves recognition efficiency. Based on the attribute information of at least one parameter of the same parameter type in different versions of the configuration parameter sequence, the process parameter difference evaluation score of the current parameter is determined. This provides real-time monitoring and early warning for lithography process stability, allowing for prediction of the impact of parameter changes on key indicators based on the current parameter's process difference evaluation score, thus reducing yield losses caused by trial and error.

[0043] Next, refer to Figure 2 The specific steps in step S101 above, which generate the configuration parameter sequence and parameter association information for the binary configuration files of each lithography machine, are described. Figure 2 This is a schematic diagram of a process for generating a sequence of configuration parameters and parameter association information provided in an embodiment of this application.

[0044] S201. Generate a byte feature sequence based on the binary configuration files of multiple lithography machines. The byte feature sequence includes multiple byte feature vectors arranged in order. The byte feature vectors include byte value feature vectors, position feature vectors, context feature vectors, and system feature vectors.

[0045] Specifically, each binary configuration file is first serialized byte-by-byte to obtain a continuous byte stream, such as... For each byte in the binary configuration file, four types of features are constructed separately, as shown in the following formula (1): (1) in, This represents the byte feature vector corresponding to the i-th byte. It is a preset byte embedding matrix used to map byte values ​​to a high-dimensional semantic space, so that different byte values ​​can obtain differentiated semantic representations, where d is the preset vector dimension. This represents the byte value of the i-th byte. That is, the feature vector of the byte value of the i-th byte. The preset position encoding matrix can be constructed using a rotation position encoding method, where L max Given the preset maximum byte sequence length, this positional encoding matrix is ​​used to inject sequence order information into the bytes, enabling the model to identify the storage location of the bytes. The rotational positional encoding method is, for example, the following formula (2): (2) in, This represents the position feature vector component of the i-th byte, k is the dimension index of the position encoding, and d is the preset total dimension of the position feature vector.

[0046] Let i be the position feature vector of the i-th byte. A predefined context feature embedding matrix is ​​used to map the local context pattern of a byte into a high-dimensional vector, reflecting the association pattern between the byte and its neighboring bytes. Let i represent the context features of the i-th byte, where... The local context features of the i-th byte can be obtained by performing convolution operations on the multimodal embedding vectors of adjacent bytes and then activating them with ReLU. (3) in, The kernel size is a hyperparameter preset based on prior knowledge of the field lengths configured in a typical lithography machine. For example, for integer parameters, the kernel size is 4, and for floating-point parameters, it is 16. It is a d×d matrix used for input features. Perform a linear transformation. This represents the intermediate feature vector of the (i+j)th byte. j is the offset within the convolution window, i.e., from - arrive , This represents the learnable bias vector in the convolution operation, and the vector has dimension d.

[0047] This represents the context feature embedding vector of the i-th byte. The system feature embedding matrix is ​​a preset matrix, where N modules This represents the total number of functional subsystems in the lithography machine, used to map the functional system identifier of a byte to a high-dimensional vector. This is the system characteristic of the i-th byte, that is, the system identifier of the lithography machine to which this byte belongs. Let i be the system feature vector of the i-th byte.

[0048] S202. Based on the byte feature sequence, determine the byte segment information in each binary configuration file according to the identification model. The byte segment information includes: byte information, context information, and the system to which it belongs.

[0049] The byte information includes the start offset address, end offset address, byte length, and byte content of the byte segment. For example, starting at 0x100 and ending at 0x104, the length is 4 bytes, and the byte content is [0x42, 0xBE, 0x76, 0xC0]. The context information includes the byte information of the preceding and following byte segments of this sub-stage. The system to which it belongs is the system of the lithography machine corresponding to this sub-stage.

[0050] Specifically, a self-supervised learning model based on the Transformer architecture is used as the recognition model. The model can contain 12 Transformer layers, with 8 attention heads in each layer. Then, layered attention calculation is used to segment the field, and the structural boundaries are determined by analyzing the mutation patterns of the attention weights in each layer. Bytes between adjacent boundaries are divided into byte segments to determine the byte information.

[0051] Based on the sorting order of each byte in the byte feature sequence, the context byte of each byte segment is determined, and then the byte information of the context byte is used as the context information.

[0052] Based on the system characteristics of each byte in the byte segment information, determine the system to which the byte segment belongs.

[0053] S203. Determine the configuration parameter sequence based on the information of each byte segment.

[0054] As an optional implementation, the distribution characteristics of each byte segment can be statistically analyzed first to determine which parameter type corresponds to the characteristics of each byte segment. This parameter type is then used as the first parameter type set. Next, it is checked whether the decoded byte value falls within the parameter range of the corresponding system. The parameter types that fall within the range are used as the second parameter type set. Then, it is verified whether each byte segment conforms to the standard encoding format that matches each parameter type. The matching parameter types are used as the third parameter type set. Finally, based on the first parameter type set, the second parameter type set, the third parameter type set, and the context information, the parameter type and parameter value corresponding to each byte segment are determined.

[0055] S204. Determine the parameter association information based on the configuration parameter sequence.

[0056] Specifically, attention similarity, positional proximity, numerical correlation, and module similarity are calculated based on the configuration parameter sequence, and then the parameter association relationship is determined based on attention similarity, positional proximity, numerical correlation, and module similarity.

[0057] Among them, attention similarity is a cross-layer attention weight based on the recognition model, which is used to quantify the semantic association strength between parameters; positional proximity is the proximity association degree calculated by using an exponential decay function based on the physical distance of the parameters in the configuration file; numerical correlation is the synchronicity of parameter value changes calculated based on historical data; and module similarity is used to determine whether parameters belong to the same functional subsystem, with parameters in the same module being given higher association weights.

[0058] In this embodiment, a byte feature sequence is generated based on the binary configuration files of multiple lithography machines. The byte feature sequence includes multiple sequentially arranged byte feature vectors. Based on the information of each byte segment, a configuration parameter sequence is determined. Based on the configuration parameter sequence, parameter association information is determined, thereby refining the parsing range, making the system more accurate in identifying different parameters, and reducing cross-module parsing ambiguity.

[0059] Next, refer to Figure 3 The step S202 above, which involves determining the information of each byte segment in each binary configuration file based on the byte feature sequence and an identification model, will be described. Figure 3 This is a flowchart illustrating a process for determining information for each byte segment, as provided in an embodiment of this application.

[0060] S301. Based on the recognition model, the bottom attention layer, the middle attention layer, and the high attention layer extract features from the byte feature sequence and determine the boundary score of each byte.

[0061] In the lower attention layer, namely layers 1-4, a local attention window is used, for example, the first and last 32 bytes, focusing on the internal structure of the field, as shown in the following formula (4): (4) in, This represents the local attention weight matrix output by the h-th attention head in the l-th layer. This represents the query vector matrix of the h-th attention head in the l-th layer. This represents the transpose of the key vector matrix of the h-th attention head in the l-th layer. This represents the dimension of the query / key vector. This represents the local attention mask matrix.

[0062] In the middle attention layer, i.e., layers 5-8, the attention window is expanded, for example, to 256 bytes before and after, to identify the boundary of the parameter group, as shown in the following formula (5): (5) in, This represents the mid-level attention mask matrix, allowing each position to focus on the preceding and following 256 bytes, covering the entire range of bytes corresponding to the parameter.

[0063] At the higher-level attention layer, namely layers 9-12, global attention is used to focus on the overall layout of the document, as shown in the following formula (6): (6); Then, the boundary score of each byte is determined by analyzing the attention weights of each layer. Specifically, this analysis is based on the attention weights of each layer provided by the aforementioned formulas (4), (5), and (6). First, the weights of each layer... Derive the normalized attention distribution for each byte Next, the attention distribution between adjacent bytes i and i-1 is calculated. and The KL divergence between the two is used to measure the degree of structural difference between them; finally, the weighted sum of the difference values ​​of all Transformer layers is used to obtain the final boundary score of the byte. Specifically, as shown in the following formula (7): (7) Where BoundaryScore(i) is the boundary score of the i-th byte. These are the weighting coefficients for each layer, used to balance the importance of features at different depths. This represents the Kullback-Leibler divergence, used to measure the difference between two probability distributions. and These represent the normalized attention distribution vectors of the i-th and (i-1)-th bytes in the l-th layer, respectively. Specifically, It is a vector obtained by averaging the attention weights of all parallel analysis units in the layer for the i-th byte, reflecting the distribution of the overall association strength between the i-th byte and all other bytes in the file, as perceived by the model.

[0064] S302. Take the bytes in each byte whose boundary score is greater than the preset boundary score threshold as boundary bytes, and take all bytes in two adjacent boundary bytes as a byte segment.

[0065] The preset boundary score threshold can also be a dynamic threshold, which can be calculated based on the statistical characteristics of the scores within the window. These statistical characteristics can include the mean and standard deviation. A dynamic threshold can adapt to changes in structural density in different regions of the file.

[0066] Byte i is considered a boundary byte when its boundary score is greater than a preset boundary score threshold. Bytes with boundary scores greater than the preset threshold are then considered candidate boundary bytes. These candidate boundary bytes are then type-determined to distinguish their function: they can serve as metadata separators between parameters, such as fixed ASCII characters or specific byte sequences, or as data type boundaries for valid parameters, such as bytes marking the start position of floating-point numbers or integers. The determination criteria are: if a candidate byte or its neighboring byte combination conforms to a predefined separator pattern library, it is classified as a separator and removed from subsequent processing; if it conforms to the header characteristics of common data type encodings, it is determined as a valid data type boundary. Finally, the filtered valid data type boundary bytes are arranged in file order, and all bytes between two adjacent valid boundaries (excluding the boundary byte itself) are divided into an independent byte segment, each of which corresponds to a parameter field to be parsed.

[0067] S303. Take the byte position, byte quantity and byte value in the byte segment as the byte information of the byte segment, and determine the system to which the byte segment belongs based on the system characteristics of each byte in the byte segment. Take the byte information and system to which the adjacent byte segments belong as the context information of the byte segment.

[0068] Specifically, if the probability that the system characteristics of each byte in a byte segment belong to the same system is greater than a preset probability threshold, then the system to which the system characteristic belongs is the system to which the byte segment belongs.

[0069] In this embodiment, multiple byte segments are determined based on the boundary scores of the boundary bytes, thereby automatically splitting the fields of the binary file without the need for manual pre-setting of the field format. At the same time, the basic information of the byte segments is obtained, the context and the system to which they belong are associated, and the format changes of different solid-state versions are adapted.

[0070] Figure 4 This is a flowchart illustrating a method for determining a sequence of configuration parameters according to an embodiment of this application. For example... Figure 4 As shown, the specific method for determining the configuration parameter sequence based on the information of each byte segment in step S203 above is as follows: S401. Based on the byte information in each byte segment, extract the statistical distribution characteristics of each byte segment. The statistical distribution characteristics include entropy, skewness, and kurtosis.

[0071] Among them, the entropy value is used to measure the dispersion of byte values ​​within a byte segment. The lower the entropy value, the more concentrated the byte value distribution, and the more likely it is to be a control parameter with clear physical meaning. The higher the entropy value, the more random the distribution, and the more likely it is to be noise or meaningless data. Specifically, the formula for calculating the entropy value is shown in the following formula (8): (8) in, The entropy value of the current byte segment. This indicates the probability that the byte value v will appear in the current byte segment.

[0072] Skewness describes the degree of symmetry in the distribution of byte values, reflecting the skewness of the data distribution. For example, the distribution of control parameters is usually more symmetrical, while noisy data may exhibit a skewed distribution. Specifically, the formula for calculating skewness is shown in the following formula (9): (9) in, Indicates the skewness of the current byte segment. A random variable representing the byte value in a byte segment. This represents the average value of all bytes in the byte segment. This represents the standard deviation of byte values ​​within the current byte segment.

[0073] Kurtosis describes the steepness of the distribution of byte values, reflecting whether there are extreme values ​​in the data. For example, the kurtosis of the distribution of process control parameters is usually moderate, while fields containing outlier correction values ​​may have high kurtosis. Specifically, the formula for calculating peak value is shown in the following formula (10): (10) in, This indicates the peak value of the current byte segment.

[0074] As an alternative implementation, cluster analysis, such as K-means, can be performed on the statistical distribution characteristics of all byte segments. The cluster centers are then compared with the features of a small number of known parsed parameter types, or byte segments with similar features are grouped together to form a candidate set of the first parameter type.

[0075] S402. Determine the parameter matching range for each byte segment based on the byte information, the system to which it belongs, and the preset parameter ranges in each byte segment information.

[0076] The parameter matching range refers to the preset physical range of the corresponding system to which the value of the decoded byte segment belongs. The preset parameter range is a reasonable range of values ​​for each parameter that is pre-set based on the physical characteristics of the photolithography process and equipment requirements, such as the exposure dose range of the illumination system and the temperature adjustment range of the temperature compensation system.

[0077] Specifically, the byte information of the byte segment is first decoded: based on the byte length and preliminary format, the byte sequence is converted into decimal values. Based on the system to which the byte segment belongs, the corresponding preset parameter range library is located. Reasonable ranges for various parameters under that system are retrieved from the preset range library. For example, if the system is a lighting system, preset exposure dose ranges [0.1, 100.0] mJ / cm² and lighting uniformity ranges [0.95, 1.05] are matched; if the system is a workbench system, preset positioning accuracy ranges [-100, 100] μm and grid matching ranges [0.99, 1.01] are matched; if the system is an environmental control system, preset temperature compensation ranges [15, 25] ℃ and air pressure correction ranges [99, 101] kPa are matched. Then, it is determined whether the byte segment value falls within the corresponding preset range and the specific parameter type range that is matched.

[0078] Optionally, a second parameter type set can be generated by matching the decoded value of the byte segment with a preset physical range. For example, if the value falls within a preset reasonable range for parameter Q, then parameter Q is added to the candidate set of the second parameter type.

[0079] S403. Determine the format matching range of the byte segment based on the byte values ​​in the byte information of each byte segment.

[0080] Specifically, parameters of different data types have fixed encoding formats. For example, floating-point numbers follow the IEEE 754 standard, integers have explicit byte order rules, and ASCII strings are composed of printable characters. By detecting whether the byte values ​​conform to these standard formats, the storage type of the parameter can be determined, thereby determining the set of third parameter types.

[0081] Specifically, the IEEE 754 floating-point matching test involves determining whether the byte sequence conforms to the single-precision (4 bytes) or double-precision (8 bytes) floating-point encoding rules, such as checking whether the arrangement of the sign bit, exponent bit, and mantissa bit conforms to the standard.

[0082] Integer byte order consistency detection specifically involves determining whether a byte sequence conforms to the big-endian or little-endian byte order integer encoding rules, such as whether the high and low byte arrangement of a multi-byte field conforms to common integer storage logic.

[0083] The ASCII readable character ratio detection specifically involves: counting the percentage of printable ASCII characters in a byte segment; if the percentage exceeds a preset threshold, it is determined to be an ASCII string format.

[0084] Then, the parameter type corresponding to the data storage format category matched by the byte segment is taken as the format matching range of the byte segment, that is, the third parameter type set.

[0085] S404. Determine the initial parameter type of each byte segment based on the statistical distribution characteristics, parameter matching range, and format matching range of each byte segment.

[0086] Specifically, based on the statistical distribution characteristics of each byte segment, the parameter matching range, and the format matching range, the overlapping parameter types in the parameter type set are used as the initial parameter types.

[0087] S405. Based on the context information in each byte segment and the initial parameter type of each byte segment, determine the parameter type and parameter value of the corresponding parameter in each byte segment of the binary configuration file.

[0088] Specifically, the context similarity of each byte segment can be determined based on the following formula (11): (11) in: Indicates the relationship with the first The set of indices of other byte segments that are directly adjacent to each other. This indicates the number of elements in the set, i.e., the number of adjacent byte segments. and They are the first and the The segment-level feature vector of a byte segment. This vector can be obtained by analyzing the intermediate feature vectors of all bytes within that byte segment. This is obtained by performing pooling (e.g., averaging). ,in Indicates the first The set of byte indices contained in a byte segment. This represents the length of the byte segment. This provides a comprehensive representation of the semantic information of the byte segment.

[0089] The context similarity of each byte segment reflects the degree of consistency between the current byte segment and its context. When determining the parameter type, if a candidate type of the current segment is the same as the determined (or high-confidence candidate) types of several highly similar neighboring segments, the confidence of that candidate type will be significantly enhanced. Finally, the system will integrate statistical, range, format, and context similarity evidence to apply a weighted score to each candidate type, thereby determining the most likely parameter type.

[0090] Next, the step of determining the parameter association information based on the configuration parameter sequence in step S204 above will be described: Optionally, based on the configuration parameter sequence, the attention similarity, positional proximity, numerical correlation, and module similarity of the parameters corresponding to each byte segment are determined. Attention similarity is used to characterize the semantic association strength between parameters in the parameter group, positional proximity is used to characterize the positional distance between parameters in the parameter group, numerical correlation is used to characterize the synchronicity of changes between the current version and the historical version of the parameters in the parameter group, and module similarity is used to characterize whether the parameters in the parameter group belong to the same system. The parameter group includes any two parameters.

[0091] Attention similarity can be determined according to the following formula (12): (12) in, Let L represent the attention similarity between byte segment i and byte segment j, L be the total number of layers in the Transformer model, and H be the number of attention heads in each layer. This represents the normalized attention weight of byte segment i to byte segment j in the i-th layer and h-th attention head. This value comes directly from the forward propagation calculation of the recognition model and reflects the degree of attention the model pays to byte segment j when processing byte segment i in this specific head.

[0092] The proximity can be determined according to the following formula (13): (13) in, The proximity of byte segment i and byte segment j. This is a preset distance scaling factor used to control the rate at which proximity decays as distance increases.

[0093] The numerical correlation can be determined according to the following formula (14): (14) in, The numerical correlation between byte segment i and byte segment j. These are the values ​​of the i-th byte segment and the j-th byte segment, respectively. These are the standard deviations of the values ​​in the i-th byte segment and the j-th byte segment, respectively.

[0094] Module similarity can be determined according to the following formula (15): (15) in, Let i be the module similarity between byte segment i and byte segment j. This is the system identifier of byte segment i. It is worth noting that formula (15) is a hard matching rule. The module similarity of two byte segments is 1 if and only if they belong to the same functional subsystem. If they belong to different subsystems, the similarity is 0, which ensures that the parameters within the same module are more correlated.

[0095] Optionally, parameter association information is determined based on the attention similarity, positional proximity, numerical relevance, module similarity, preset attention similarity weight, preset positional proximity weight, preset numerical relevance weight, and preset module similarity weight of the parameters corresponding to each byte segment.

[0096] Specifically, the parameter association information can be determined according to the following formula (16): (16) in, This is the association information between byte segment i and byte segment j. It is the attention similarity weight. It is the proximity weight. It is the numerical correlation weight. It is the module similarity weight.

[0097] Optionally, the association information of all byte segments constitutes the parameter association information.

[0098] In this embodiment, parameter association information is determined from multiple dimensions such as attention similarity, positional proximity, numerical correlation, and module similarity, thereby avoiding misjudgment based on a single dimension.

[0099] Next, refer to Figure 5 The specific steps in step S102 above, which involve determining the process parameter difference evaluation score of the current parameter based on the attribute information of at least one parameter of the same parameter type from different versions in the configuration parameter sequence, are described below: S501. Determine the importance weight of the current parameter based on the preset process sensitivity weight, preset system importance weight, preset parameter criticality level, and attribute information of each parameter.

[0100] The process sensitivity weight represents the preset weight of the parameter's impact on key lithography process indicators. This weight can be calibrated based on a large amount of process experimental data; the higher the sensitivity, the greater the weight. The system importance weight can be preset based on the core status of the system to which the parameter belongs. For example, parameters of core systems such as lighting systems and stage systems have higher weights than those of auxiliary systems. The parameter criticality level is a dynamically adjusted level indicator based on the impact of parameter changes on chip yield; the greater the impact on yield, the higher the level.

[0101] Specifically, the process sensitivity weight of each parameter can be obtained based on the attribute information of each parameter and the preset process sensitivity weight mapping. Similarly, the system importance weight of each parameter can be obtained based on the attribute information of each parameter and the preset system importance weight. The parameter criticality level of each parameter can be obtained based on the attribute information of each parameter and the preset parameter criticality level. Then, the importance weight of each parameter can be determined based on the process sensitivity weight, system importance weight, and parameter criticality level of each parameter, as shown in the following formula (17): (17) Here, p is an identifier that specifically refers to the particular parameter parsed from byte segment i. It is the importance weight of parameter p corresponding to byte segment i. It is the process sensitivity weight of parameter p corresponding to byte segment i. This is the system importance weight of parameter m corresponding to byte segment i. It is the criticality level of parameter c corresponding to byte segment i.

[0102] S502. Determine the difference value of the current parameter according to the preset process influence distance function corresponding to each parameter and the attribute information of each parameter.

[0103] The process influence distance function is a difference measurement function customized for parameters of different types of lithography machines. It fully combines the physical characteristics of the lithography process to quantify the actual impact difference between two versions of parameters. The difference value of the current parameter is a quantitative value that reflects the degree of actual impact of the parameter changes of the two versions on the process, calculated by the process influence distance function. The larger the value, the more significant the impact.

[0104] Optionally, the impact logic of changes in different types of parameters on the process varies. For example, changes in exposure dosage affect photosensitivity, while changes in stage positioning parameters affect overlay accuracy. If a uniform numerical difference calculation method is used, the physical characteristics of the process will be ignored, leading to distorted evaluation. Therefore, it is necessary to customize specific difference measurement methods for each type of parameter to ensure that the difference values ​​can truly reflect the impact of the process.

[0105] The following section uses exposure metering parameters and stage positioning parameters as examples to introduce the method for obtaining difference values: If the current parameter is an exposure measurement parameter, then the formula for calculating the difference value of the current parameter is shown in (18): (18) in, For the difference in exposure measurement parameters, Let i be the value of the i-th parameter in version A. Let i be the value of the i-th parameter in version B. The sensitivity coefficient of this parameter to critical dimensions, DU is the preset dose uniformity influence factor, and RT is the preset photoresist type correction coefficient.

[0106] If the current parameter is a workbench positioning parameter, then the formula for calculating the difference value of the current parameter is shown in (19): (19) in, The table positioning parameters are: OS is the preset overlay accuracy specification value, which can be dynamically adjusted according to the requirements of different technical nodes; M is the preset system magnification factor, including the difference between scanning direction and stepping direction; GD is the preset grid distortion compensation factor, which can be determined based on historical grid correction data; and ST is the table type correction factor, used to distinguish between immersive and non-immersive tables.

[0107] S503. Determine the process parameter difference assessment score for the current parameter based on the importance weight and difference value of the current parameter.

[0108] Among them, the process parameter difference assessment score is the final score that combines the importance weight of the parameters with the difference value, and quantifies the degree of impact of configuration changes on the overall photolithography process. The higher the score, the greater the risk of the change to the process.

[0109] Optionally, the process impact of parameter changes depends on both the magnitude of the difference and the importance of the parameter itself. Therefore, by combining importance weights with difference values, a comprehensive risk quantification of configuration changes can be achieved, avoiding one-sided assessments that prioritize numerical values ​​over importance or vice versa.

[0110] Specifically, the process parameter difference assessment score for the current parameter can be determined using formula (20): (20) in, Assess the difference in process parameters for the current parameters with a score. For the number of all parameters, This represents the difference value of the current parameter. It is a parameter activity indicator function based on the i-th parameter, ensuring that only the parameters actually used in the current process formulation are evaluated. It is the formulation relevance coefficient based on the i-th parameter, which can be adjusted according to the frequency of use of the parameter in the current process formulation.

[0111] In this embodiment, by determining the importance weight of the current parameter and the difference value of the current parameter, the process parameter difference assessment score of the current parameter is determined, so as to realize the accurate, efficient and standardized assessment of configuration change risk, provide support for the stable operation and optimization decision of the lithography process, and significantly reduce the production risk and trial and error cost caused by parameter changes.

[0112] Next, refer to Figure 6 The specific steps in step S103 above, which involve determining the process stability report for the current parameter based on the attribute information and parameter association information of at least one parameter of the same parameter type from different versions of the configuration parameter sequence, are described below. Figure 6 This is a schematic diagram of a process stability report for determining current parameters provided in an embodiment of this application.

[0113] S601. Determine the stability score of the current parameter based on the attribute information of at least one parameter of the same parameter type in different versions of the configuration parameter sequence, the preset parameter standard deviation, and the preset parameter standard value.

[0114] Specifically, the stability score of the current parameter is used to evaluate the long-term stability of the parameter across different versions. By comparing the parameter values ​​of different versions with preset standard values, and considering the reasonableness of the parameter within the process window, the impact of the environment and maintenance cycle, the average stability within the time window is calculated. The higher the score, the more stable the parameter.

[0115] Alternatively, the stability score of the current parameter can be determined according to the following formula (21): (twenty one) in, It is a process consistency measurement function, and its calculation formula is as follows (22): (twenty two) This represents the stability score of the current parameter at time t. This indicates the length of the sliding time window used for the evaluation, i.e., the number of historical versions used to calculate the average. This indicates that the parameter is in historical time. The actual parameter value at that time, The preset seasonal weighting function, This is the preset standard deviation of the parameter, which can be set according to equipment capabilities and product requirements. The preset parameter standard value, This is the preset overlay accuracy margin factor, and LH is the preset lens thermal effect correction coefficient. It is a preset maintenance cycle weight, which can be adjusted based on the equipment maintenance history.

[0116] S602. Determine the trend duration weight of the current parameter based on the duration of the change trend of each parameter.

[0117] The trend duration weight of the current parameter is used to quantify the risk of continuous parameter drift. The weight is determined based on the number of consecutive versions or days of the same trend in the parameter; the longer the duration, the greater the weight.

[0118] S603. Based on the parameter correlation information of each parameter, determine the relevant parameters of the current parameter, and based on the changing trends of the relevant parameters and the changing trend of the current parameter, determine the cross-correlation weight of the current parameter.

[0119] Among them, the cross-correlation weight of the current parameter is used to assess the risk of coordinated change between the parameter and related parameters. Specifically, the related parameters are first found through parameter association information, and then the similarity of the change trends of the two is calculated as the cross-correlation weight.

[0120] S604. Determine the anomaly frequency impact factor of the current parameter based on the frequency of anomalies of each parameter.

[0121] The frequency of anomalies in the current parameter is used to quantify the risk of frequent parameter anomalies. It counts the number of times the parameter exceeds a reasonable range within a unit of time; the more times this occurs, the larger the impact factor.

[0122] S605. Determine the warning level of the current parameter based on the changing trend of the current parameter, the weight of the trend duration, the weight of the cross-correlation, and the impact factor of the abnormal frequency.

[0123] Specifically, the warning level of the current parameter can be determined according to formula (23): (twenty three) in, The score indicates the warning level. This shows the current trend of parameter changes. Here, TD represents the impact factor of anomaly frequency, CC represents the weight of trend duration, and CC represents the weight of cross-correlation between parameters. The weights are the preset weights for the changing trends of the current parameters. The weights of the second-order rate of change of the preset trend. The weights of the pre-defined anomaly frequency impact factors, The weight is the preset trend duration weight. The weights are the preset cross-correlation weights.

[0124] In this embodiment, the warning level of the current parameter is determined based on the changing trend of the current parameter, the weight of the trend duration, the weight of the cross-correlation, and the impact factor of the abnormal frequency. This allows engineers to quickly prioritize and process high-warning parameters without having to deeply analyze the original data, thereby improving operation and maintenance efficiency and reducing the production risks caused by process fluctuations.

[0125] As an optional implementation method, the lithography machine configuration information processing method further includes the following steps: determining whether the current parameter exceeds the process window warning based on the attribute information of the current parameter, the preset maximum value of each parameter, the preset minimum value of each parameter, and the preset standard value of the parameter, and generating a health report based on the determination result.

[0126] Among them, the preset maximum / minimum values ​​of each parameter are the safety boundaries of the process window, that is, the limit range allowed by the parameters.

[0127] The process window margin for the current parameters is determined according to the following formula (24): (twenty four) in, It represents the process window margin of the current parameters at time t. It is the parameter value at time t. It is the preset minimum value of each parameter. These are the preset maximum values ​​for each parameter. θ(t) represents the preset standard parameter value, while CL represents the confidence level, which reflects the reliability assessment of the current parameter value θ(t) or the model. When the confidence level is low, the calculated process window margin will decrease accordingly. This causes the system to adopt a more conservative strategy when the confidence level is low, thus being more inclined to trigger a higher-level warning to prompt operators to intervene and verify. Conversely, when the confidence level is high, the warning will be triggered normally based on the calculated margin. It is a preset device age factor.

[0128] Optionally, the system has preset multi-level alarm thresholds, such as a level 1 alarm threshold and a level 2 alarm threshold. The calculated current parameter process window margin (WindowMargin(t)) is compared with these thresholds. Based on the different threshold ranges within which WindowMargin(t) falls, the system automatically triggers the corresponding level of process window warning and provides specific parameter adjustment suggestions for that level, thereby generating a health report. In this embodiment, by determining whether the current parameter exceeds the process window warning and generating a health report based on the determination result, the system can detect parameter deviation trends in advance, avoiding yield decline or process failure due to parameter exceeding limits.

[0129] As an optional implementation, the above-mentioned recognition model can update its own parameters through an incremental learning strategy to adapt to the new configuration file format. The specific process is as follows: When a new version of the binary configuration file is available, the current recognition model is first used to parse the new version of the binary configuration file to obtain the configuration parameter sequence of the new version. Additionally, feature parsing can be performed on the configuration parameter sequence to obtain the feature parsing results.

[0130] Then, calculate the distributional similarity between the new version of the binary configuration file and the historical version of the binary configuration file, or calculate the distributional similarity between the feature parsing results of the new version of the binary configuration file and the feature parsing results of the historical version of the binary configuration file. Specifically, the distributional similarity is calculated as follows: (25) (25) in, It is distribution similarity. It is a historical version of the binary configuration file or the result of parsing the characteristics of the binary configuration file. It is a new version of the binary configuration file or the result of the binary configuration file's feature parsing.

[0131] Based on the distribution similarity mentioned above, the appropriate regularization strength for updating the current recognition model is determined. Lower similarity indicates greater data variation, requiring stronger regularization to prevent the model from forgetting old knowledge as it adapts to new data.

[0132] For example, the regularization strength can be determined based on the following formula (26): (26) in, This represents the regularization strength. 0.3 and 0.7 are both preset thresholds.

[0133] Incremental training and parameter updates are performed: The new version of the binary configuration file is used as the training sample, and one round of training is executed starting with the current model parameters. In the training objective function, a regularization constraint centered on the current model parameters is introduced, satisfying the regularization strength. The parameter update amount is calculated, thus obtaining candidate updated model parameters. For example, the formula for calculating the updated model parameters is as follows: (27) in, It is the updated model parameter i. It is the historical version model parameter i. It is a loss function. It is the optimal value of parameter i obtained after training the historical version of the model. The importance score is calculated using the following formula (28): (28) Where N is the number of training samples, Is the model in parameters Output generated below The probability of.

[0134] As an optional implementation, the recognition accuracy of the updated model parameters is evaluated on a separate validation set. If the accuracy is not lower than a preset threshold, the update is accepted, and the old parameters are replaced with the new ones. If the accuracy drops significantly, the update is rejected, the original model parameters are retained, and an alarm may be triggered or manual intervention may be required.

[0135] Specifically, if the data distribution similarity is extremely low (e.g., less than a preset threshold), and the validation accuracy after model update remains high (e.g., greater than the preset threshold), it indicates that the new data is of extremely high quality or represents a completely new but effective pattern, and the system can safely complete this strong update. Conversely, if both similarity and accuracy are low, it indicates that the update may be unfavorable, and the update should be rejected with an anomaly alert issued.

[0136] This application also provides an electronic device, such as... Figure 7 The diagram shown is a structural schematic of an electronic device provided in an embodiment of this application, including a processor 701, a memory 702, and a bus. The memory 702 stores machine-readable instructions executable by the processor 701. When the computer device is running, the processor 701 and the memory 702 communicate via the bus. The processor 701 executes the machine-readable instructions to perform the processing of the above-described lithography machine configuration information processing method.

[0137] This application also provides a computer-readable storage medium storing a computer program, which, when executed by a processor, performs the steps of the above-described lithography machine configuration information processing method.

[0138] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the specific working processes of the systems and devices described above can be referred to the corresponding processes in the method embodiments, and will not be repeated here. In the several embodiments provided in this application, it should be understood that the disclosed systems, devices, and methods can be implemented in other ways. The device embodiments described above are merely illustrative. For example, the division of modules is only a logical functional division, and in actual implementation, there may be other division methods. Furthermore, multiple modules or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the displayed or discussed mutual coupling or direct coupling or communication connection can be through some communication interfaces; the indirect coupling or communication connection of devices or modules can be electrical, mechanical, or other forms.

[0139] Furthermore, the functional units in the various embodiments of this application can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. If the functions are implemented as software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this invention, or the part that contributes to the prior art, or a part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of this invention. The aforementioned storage medium includes: USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, optical disks, and other media capable of storing program code.

[0140] The above are merely specific embodiments of this application, but the scope of protection of this application is not limited thereto. Any changes or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application.

Claims

1. A method for processing configuration information of a lithography machine, characterized in that, The method includes: Based on the acquired binary configuration files of multiple lithography machines, a pre-trained recognition model is used to parse the binary configuration files to generate a configuration parameter sequence and parameter association information for each lithography machine's binary configuration file. The configuration parameter sequence includes attribute information of the parameters corresponding to each byte segment in the binary configuration file. The attribute information includes: byte position, parameter type, parameter value, and system to which it belongs. The parameter association information represents the functional association relationship between the parameters corresponding to each byte segment in the configuration parameter sequence. Based on the attribute information of at least one parameter of the same parameter type in different versions of the configuration parameter sequence, the process parameter difference assessment score of the current parameter is determined. The process parameter difference assessment score is used to characterize the degree of influence of the difference of at least one parameter of the same parameter type in different versions on the corresponding process indicators. Based on the attribute information of at least one parameter of the same parameter type in different versions of the configuration parameter sequence and the parameter association information, the process stability report of the current parameter is determined, and the process stability includes a stability score and a warning level.

2. The lithography machine configuration information processing method according to claim 1, characterized in that, The step involves using a pre-trained recognition model to parse the binary configuration files of multiple lithography machines, based on the acquired binary configuration files, to generate a sequence of configuration parameters and parameter association information for each lithography machine's binary configuration file, including: Based on the binary configuration files of the multiple lithography machines, a byte feature sequence is generated. The byte feature sequence includes multiple byte feature vectors arranged in order. The byte feature vectors include byte value feature vectors, position feature vectors, context feature vectors, and system feature vectors. Based on the byte feature sequence, the byte segment information in each binary configuration file is determined based on the recognition model. The byte segment information includes: byte information, context information, and the system to which it belongs. The configuration parameter sequence is determined based on the information of each byte segment; The parameter association information is determined based on the configuration parameter sequence.

3. The lithography machine configuration information processing method according to claim 2, characterized in that, The step of determining the information of each byte segment in each binary configuration file based on the byte feature sequence and the recognition model includes: Based on the bottom attention layer, middle attention layer and high attention layer of the recognition model, feature extraction is performed on the byte feature sequence to determine the boundary score of each byte. Bytes whose boundary scores are greater than a preset boundary score threshold are designated as boundary bytes, and all bytes in two adjacent boundary bytes are treated as a single byte segment. The byte position, number of bytes, and byte value in the byte segment are used as the byte information of the byte segment. Based on the system characteristics of each byte in the byte segment, the system to which the byte segment belongs is determined. The byte information and system to which the adjacent byte segments belong are used as the context information of the byte segment.

4. The lithography machine configuration information processing method according to claim 2, characterized in that, Determining the configuration parameter sequence based on the byte segment information includes: Based on the byte information in each byte segment, the statistical distribution characteristics of each byte segment are extracted, including entropy, skewness, and kurtosis. Based on the byte information in each byte segment, the system to which it belongs, and the preset parameter ranges, determine the parameter matching range for each byte segment; The format matching range of the byte segment is determined based on the byte value of the byte information in each byte segment information; Based on the statistical distribution characteristics, parameter matching range, and format matching range of each byte segment, the initial parameter type of each byte segment is determined. Based on the context information in each byte segment and the initial parameter type of each byte segment, determine the parameter type and parameter value of the parameter corresponding to each byte segment in the binary configuration file.

5. The lithography machine configuration information processing method according to claim 2, characterized in that, Determining the parameter association information based on the configuration parameter sequence includes: Based on the configuration parameter sequence, the attention similarity, positional proximity, numerical correlation, and module similarity of the parameters corresponding to each byte segment are determined. The attention similarity is used to characterize the semantic association strength between parameters in the parameter group, the positional proximity is used to characterize the positional distance between parameters in the parameter group, the numerical correlation is used to characterize the synchronicity of changes between the current version and the historical version of the parameters in the parameter group, and the module similarity is used to characterize whether the parameters in the parameter group belong to the same system. The parameter group includes any two parameters. The parameter association information is determined based on the attention similarity, positional proximity, numerical relevance, module similarity, preset attention similarity weight, preset positional proximity weight, preset numerical relevance weight, and preset module similarity weight of the parameters corresponding to each byte segment.

6. The lithography machine configuration information processing method according to claim 1, characterized in that, The step of determining the process parameter difference evaluation score of the current parameter based on the attribute information of at least one parameter of the same parameter type in different versions of the configuration parameter sequence includes: The importance weight of the current parameter is determined based on the preset process sensitivity weight, preset system importance weight, preset parameter criticality level, and attribute information of each parameter. The difference value of the current parameter is determined according to the preset process influence distance function corresponding to each parameter and the attribute information of each parameter; Based on the importance weight and difference value of the current parameter, the process parameter difference evaluation score of the current parameter is determined.

7. The lithography machine configuration information processing method according to claim 1, characterized in that, The step of determining the process stability report for the current parameter based on the attribute information of at least one parameter of the same parameter type in different versions of the configuration parameter sequence and the parameter association information includes: Based on the attribute information of at least one parameter of the same parameter type in different versions of the configuration parameter sequence, the preset parameter standard deviation, and the preset parameter standard value, determine the stability score of the current parameter; The weight of the trend duration of the current parameter is determined based on the duration of the change trend of each parameter. Based on the parameter correlation information of each parameter, determine the relevant parameters of the current parameter, and based on the changing trends of the relevant parameters and the current parameter, determine the cross-correlation weight of the current parameter; Based on the frequency of anomalies of each parameter, determine the anomaly frequency impact factor of the current parameter; The warning level of the current parameter is determined based on the changing trend of the current parameter, the weight of the trend duration, the weight of the cross-correlation, and the impact factor of the abnormal frequency.

8. The lithography machine configuration information processing method according to claim 1, characterized in that, The method further includes: Based on the attribute information of the current parameter, the preset maximum value of each parameter, the preset minimum value of each parameter, and the preset standard value of the parameter, it is determined whether the current parameter exceeds the process window warning, and a health report is generated based on the judgment result.

9. An electronic device, characterized in that, include: The device includes a processor and a memory, the memory storing machine-readable instructions executable by the processor, which, when the electronic device is running, are executed by the processor to perform the steps of the lithography machine configuration information processing method as described in any one of claims 1 to 8.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program that, when executed by a processor, performs the steps of the lithography machine configuration information processing method as described in any one of claims 1 to 8.

Images