IC carrier plate detection data processing method, system and equipment based on edge calculation

Abstract

The invention relates to the technical field of data processing, and particularly provides an IC carrier detection data processing method, system and equipment based on edge computing, and the method comprises the steps: obtaining multi-modal detection data, and carrying out the abnormality recognition of the detection data of each modal; generating structured data according to the modal data and the corresponding anomaly recognition results; adding field definition and attribute description information for the fields of the structured data by utilizing a generative large model; and uploading the structured data with the field definition and the attribute description information to a database. According to the method, the cross-modal unified structured abnormal feature table is automatically generated, information islands are eliminated, and the defect root cause analysis efficiency is improved.

Description

Technical Field

[0001] The present invention belongs to the field of data processing technology, and specifically relates to an IC substrate detection data processing method, system and equipment based on edge computing. Background Art

[0002] In the field of integrated circuit substrate (IC substrate) manufacturing, multimodal quality inspection has become a core link, involving multi-dimensional data such as physical structure, electrical performance, environmental parameters and process defects. Traditional inspection systems have significant bottlenecks: Data processing lag: Multi-source heterogeneous data (structured numerical data and unstructured images) needs to be transmitted back to the cloud for processing, resulting in high latency and making it difficult to meet real-time quality inspection requirements. Difficulty in information fusion: Abnormal results of each modality are stored separately and lack a unified structured expression, which restricts root cause analysis of defects; Metadata relies on manual effort: Database field definitions must be manually compiled by engineers, leading to inconsistent terminology and high maintenance costs. While existing edge computing solutions improve response speed, they fail to address the issues of automated integration and semantic standardization of multimodal data, limiting detection efficiency and data availability. Summary of the Invention

[0003] In view of the above-mentioned deficiencies in the prior art, the present invention provides an IC substrate detection data processing method, system and device based on edge computing to solve the above-mentioned technical problems.

[0004] In a first aspect, the present invention provides an IC substrate detection data processing method based on edge computing, comprising: Acquire multi-modal detection data and perform anomaly identification on the detection data of each modality; Generate structured data based on each modal data and corresponding abnormality recognition results; Using a generative big model to add field definitions and attribute description information to the fields of the structured data; Upload structured data with field definitions and attribute descriptions to the database.

[0005] In an optional embodiment, obtaining multimodal detection data includes: Obtain physical structure data, electrical performance data, environmental data, and process defect data of the carrier board; Among them, physical structure data includes structured specification and dimension data, X-ray detection images, and ultrasonic scanning images; electrical performance data and environmental data are both structured data; and process defect data includes automatic optical detection images.

[0006] In an optional embodiment, abnormality identification is performed on the detection data of each modality separately, including: Use pre-set threshold comparison rules to identify anomalies in structured data; An image recognition model is used to identify anomalies in unstructured data, including X-ray detection images, ultrasound scanning images, and automatic optical detection images.

[0007] In an optional embodiment, structured data is generated based on each modal data and the corresponding abnormality recognition result, including: Setting a structured field table template, wherein the table items of the structured field table template include field name, data source and anomaly recognition result; Assigning a field name to each parameter item in the detection data, and filling the field name, data source and anomaly recognition result corresponding to each parameter item into the corresponding table item of the structured field table template to obtain a structured field table; A cross-modal exception association field identifier is added to the structured field table.

[0008] In an optional embodiment, using a generative big model to add field definitions and attribute description information to the fields of the structured data includes: Constructing prompt words based on the structured field table; The prompt words are input into a generative big model so that the big model generates field definitions and attribute description information for the fields in the structured field table.

[0009] In an optional embodiment, the method further comprises: Use self-explanatory snake case naming to generate corresponding English field definitions for field definitions; Use pre-established domain term comparison tables to unify the terms in attribute description information.

[0010] In an optional embodiment, the method further comprises: Use regular expressions to validate unit symbols and range expressions.

[0011] In a second aspect, the present invention provides an IC substrate detection data processing system based on edge computing, comprising: The anomaly recognition module is used to obtain multi-modal detection data and perform anomaly recognition on the detection data of each modality; A data conversion module is used to generate structured data based on each modal data and the corresponding abnormality recognition results; A description generation module, configured to add field definitions and attribute description information to the fields of the structured data using a generative large model; The data upload module is used to upload structured data with field definitions and attribute description information to the database.

[0012] According to a third aspect, a device is provided, comprising: A memory, used to store an IC substrate detection data processing program based on edge computing; A processor is used to implement the steps of the IC substrate detection data processing method based on edge computing as provided in the first aspect when executing the IC substrate detection data processing program based on edge computing.

[0013] In a fourth aspect, a computer-readable storage medium is provided, on which an IC substrate detection data processing program based on edge computing is stored. When the IC substrate detection data processing program based on edge computing is executed by a processor, the steps of the IC substrate detection data processing method based on edge computing provided in the first aspect are implemented.

[0014] The beneficial effects of the present invention are that the IC substrate detection data processing method, system, and device based on edge computing provided by the present invention achieve three major breakthrough benefits through the integration of edge computing and generative AI: Improved real-time performance: Parallel processing of multimodal data (X-ray / ultrasound / AOI images + structured parameters) on the edge reduces detection latency and meets the millisecond-level response requirements of high-speed production lines. Intelligent data fusion: Automatically generates a unified, cross-modal structured anomaly feature table, eliminating information silos and improving the efficiency of defect root cause analysis. Semantic self-evolution: Generative big models are used to dynamically build standardized field definitions and attribute descriptions, avoiding manual maintenance deviations, improving database field compliance, and supporting seamless cross-platform migration. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, for ordinary technicians in this field, other drawings can be obtained based on these drawings without any creative work.

[0016] Figure 1 is a schematic flow chart of a method according to an embodiment of the present invention.

[0017] Figure 2 FIG. 4 is a schematic block diagram of a system according to an embodiment of the present invention.

[0018] Figure 3 A schematic structural diagram of a device provided in an embodiment of the present invention. DETAILED DESCRIPTION

[0019] In order to enable those skilled in the art to better understand the technical solutions of the present invention, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts should fall within the scope of protection of the present invention.

[0020] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as those commonly understood by those skilled in the art of the present invention. The terms used in this specification of the present invention are only for the purpose of describing specific embodiments and are not intended to limit the present invention.

[0021] The IC substrate detection data processing method based on edge computing provided in an embodiment of the present invention is executed by a computer device. Accordingly, the IC substrate detection data processing system based on edge computing runs in the computer device.

[0022] Figure 1 is a schematic flow chart of a method according to an embodiment of the present invention. Figure 1 The execution entity can be an IC substrate detection data processing system based on edge computing. According to different requirements, the order of the steps in the flowchart can be changed, and some steps can be omitted.

[0023] like Figure 1 As shown, the method includes: S1, obtain multi-modal detection data and perform abnormality identification on the detection data of each modality; S2. Generate structured data based on each modal data and the corresponding abnormality recognition results; S3. Add field definitions and attribute description information to the fields of the structured data using a generative large model; S4. Upload the structured data with field definition and attribute description information to the database.

[0024] In an embodiment of the present invention, based on step S1, a possible embodiment will be given below to illustrate its specific implementation scheme in a non-limiting manner.

[0025] S101. Obtain physical structure data, electrical performance data, environmental data, and process defect data of the carrier board; wherein the physical structure data includes structured specification and dimension data and X-ray inspection images and ultrasonic scanning images; the electrical performance data and environmental data are both structured data; and the process defect data includes automatic optical inspection images.

[0026] Acquiring the carrier's physical structure data, electrical performance data, environmental data, and process defect data is fundamental to the entire technical solution. The structured dimensional data within this physical structure data is typically collected using high-precision measuring equipment, such as a three-dimensional coordinate measuring machine (CMM). A CMM can measure the carrier's length, width, height, and various dimensional parameters, such as hole diameters and slot depths, with micron-level accuracy. This measurement data is directly transmitted to the data processing system via a data interface and stored in a pre-defined structured format, such as XML or JSON, for subsequent processing and analysis. X-ray inspection images and ultrasonic scanning images are unstructured data, and their acquisition relies on specialized inspection equipment. X-ray inspection equipment transmits X-rays through the substrate, exploiting the differences in X-ray absorption by different materials to produce a grayscale image of the substrate's internal structure on the detector. Ultrasonic scanning utilizes the principle that ultrasonic waves propagating through the substrate are reflected at the interfaces between different media. By receiving and processing the reflected signals, an image is generated that reflects the substrate's internal defects and structure. During the image acquisition process, precise equipment parameters, such as X-ray tube voltage and current, and ultrasonic frequency and probe angle, must be set to ensure high-quality image data. Furthermore, to facilitate subsequent storage and transmission, the captured image data is typically compressed and stored in formats such as JPEG2000 or PNG. Both electrical performance data and environmental data are structured data. Electrical performance data primarily includes electrical parameters such as the substrate's resistance, capacitance, and inductance, as well as performance indicators such as signal transmission delay and attenuation. This data is measured using professional electrical testing equipment, such as network analyzers and impedance testers. The testing equipment tests the substrate according to standard testing procedures and transmits the measurement results to the data processing system using a specific protocol, where they are also stored in a structured format. Environmental data covers environmental parameters such as temperature, humidity, and air pressure during the substrate production and testing process. This data is collected in real time by various sensors and stored in a structured time series format. The automated optical inspection images included in the process defect data are captured by automated optical inspection (AOI) equipment. Utilizing optical imaging principles, AOI equipment scans and images the substrate surface, comparing the scanned image with a standard image to detect surface defects such as scratches, stains, and poor solder joints. During image acquisition, parameters such as the AOI light source type, illumination intensity, and camera resolution must be adjusted based on the substrate's material and surface characteristics to improve the accuracy and reliability of defect detection. The captured image data is also compressed and stored.

[0027] S102. Use pre-set threshold comparison rules to identify anomalies in structured data.

[0028] Use pre-set threshold comparison rules to identify anomalies in structured data. The establishment of threshold comparison rules needs to be based on a large amount of historical data and industry standards. First, perform statistical analysis on historical test data to calculate statistical quantities such as the mean and standard deviation of each parameter. For example, for a key dimensional parameter of a carrier board, by analyzing the past 1,000 test data, its mean is X and the standard deviation is σ. Then, based on the product quality requirements and process stability, determine a reasonable threshold range. In general, the threshold can be set to the mean plus or minus k times the standard deviation, where the value of k is adjusted according to the specific situation and is usually between 2 and 3. When the test data exceeds this threshold range, it is determined to be abnormal data. In practical applications, a multi-level threshold strategy can be employed to improve the accuracy of anomaly identification. For example, a warning threshold and a severe anomaly threshold can be set. When data exceeds the warning threshold but falls below the severe anomaly threshold, the system issues a warning signal, prompting the operator to conduct further inspection. When data exceeds the severe anomaly threshold, the system automatically determines that the substrate has quality issues and removes it from the production process. Furthermore, to accommodate product batches and process variations, the threshold comparison rules need to be dynamically adjustable. Machine learning algorithms, such as adaptive threshold algorithms, can be used to continuously optimize threshold settings based on real-time detection data to improve the accuracy and adaptability of anomaly identification.

[0029] S103. Use an image recognition model to perform anomaly recognition on unstructured data, where the unstructured data includes X-ray detection images, ultrasonic scanning images, and automatic optical detection images.

[0030] Image recognition models are used to identify anomalies in unstructured data. Deep learning algorithms, such as convolutional neural networks (CNNs), are often used when building image recognition models. For X-ray and ultrasonic scan images, since they reflect the internal structure of the substrate, the defect features within them are often complex and hidden. Therefore, a suitable CNN architecture is needed to extract these features. A combination of multiple convolutional and pooling layers can be used to gradually extract high-level semantic features from the image. For example, in the first convolutional layer, a small convolution kernel (such as 3×3) is used to extract local, detailed features of the image. As the number of network layers increases, the kernel size is gradually increased to extract more macroscopic structural features.

[0031] Training image recognition models requires a large amount of labeled data. For X-ray and ultrasound scan images, the labeling process requires specialized technicians to label defects based on image characteristics and product standards, including information such as defect type, location, and size. To improve the model's generalization capabilities, data augmentation techniques such as rotation, translation, scaling, and noise addition can be used to augment the original images and increase the diversity of the training data. During training, an appropriate loss function, such as the cross-entropy loss function, is used to measure the difference between the model's predictions and the true labels. The model's parameters are then optimized through a backpropagation algorithm to continuously reduce the loss function and improve the model's recognition accuracy.

[0032] Since automated optical inspection (AOI) images primarily detect defects on the substrate surface, defect features are relatively distinct. Therefore, classic CNN models, such as AlexNet and VGGNet, can be used as the underlying architecture, with appropriate adjustments and optimizations tailored to specific needs. Data annotation and data augmentation are also required during the training process to improve model performance. Furthermore, to detect defects on the substrate surface in real time, trained image recognition models can be deployed on edge computing devices for rapid image recognition and anomaly detection, reducing data transmission latency and costs.

[0033] In an embodiment of the present invention, based on step S2, a possible embodiment will be given below to illustrate its specific implementation scheme in a non-limiting manner.

[0034] S201. Set a structured field table template, where the table items of the structured field table template include field names, data sources, and exception recognition results.

[0035] The purpose of setting up a structured field table template is to standardize the management and organization of test data. The table entries in the structured field table template contain key information such as the field name, data source, and anomaly identification results. Field names must follow specific naming conventions to ensure uniqueness, readability, and scalability. For example, a naming scheme of "module name_parameter name" (such as "electrical performance_resistance") can be used to clearly indicate the module and specific parameter to which the data belongs.

[0036] The data source field is used to record the data collection device or system so that it can be traced and investigated when problems arise with the data. When designing the data source field, it can be stored in a coded manner. For example, a unique code can be assigned to each detection device and the code recorded in the data source field. At the same time, a mapping table between device codes and device detailed information is established to facilitate query and management. The abnormality identification result field is used to store the results of the abnormality identification of the data in step S1, usually using Boolean values (0 for normal, 1 for abnormal) or enumeration types (such as "normal", "warning", and "serious abnormality").

[0037] To make structured field table templates more adaptable and flexible, you can define them using Extensible Markup Language (XML) or JavaScript Object Notation (JSON). By defining the template's schema, you can constrain field types, lengths, and required fields within the field table, ensuring data standardization and consistency. For example, in XML schema, you can use XSD (XML Schema Definition) to define the field table template; in JSON schema, you can use JSON Schema.

[0038] S202. Assign a field name to each parameter item in the detection data, and fill the field name, data source and anomaly recognition result corresponding to each parameter item into the corresponding table item of the structured field table template to obtain a structured field table.

[0039] Assign a field name to each parameter item in the test data, and then enter the corresponding field name, data source, and anomaly identification results of each parameter item into the corresponding table item of the structured field table template to obtain a structured field table. When assigning field names, it is necessary to name them according to the meaning of the data and the module to which they belong, following pre-set naming conventions. At the same time, to improve work efficiency, an automated field name assignment tool can be developed. This tool can automatically generate appropriate field names by reading data metadata information, such as data description and unit, and then match and verify them with the structured field table template.

[0040] When entering data into a structured field table, ensure its accuracy and completeness. Data validation mechanisms can be used to perform format checks and logical verification on the entered data. For example, for numeric data, check whether it is within a reasonable range; for date data, check whether it conforms to the date format specification. Furthermore, to ensure real-time data availability, a real-time data transmission channel can be established between the data acquisition system and the structured field table. Once test data collection is completed, the data is immediately transferred to the structured field table for storage and processing.

[0041] S203. Add a cross-modal exception association field identifier to the structured field table.

[0042] The purpose of adding a cross-modal anomaly association field identifier to the structured field table is to establish an association between data of different modalities, facilitating comprehensive analysis and tracing of carrier quality issues. The design of the cross-modal anomaly association field identifier needs to consider the characteristics and association methods of data of different modalities. For example, for X-ray inspection images in physical structure data and automatic optical inspection images in process defect data, an association can be established through the carrier's unique identifier (such as a serial number). Add a "Carrier Serial Number" field to the structured field table as a cross-modal anomaly association field identifier to associate different modal data with the same serial number.

[0043] To automate the addition of cross-modal anomaly association field identifiers, corresponding algorithms and programs can be developed. This program can automatically match and associate related data by reading key information from different modal data, such as the carrier serial number and detection time, and add corresponding identifiers to the structured field table. Furthermore, to improve the accuracy of association, a fuzzy matching algorithm can be used to address matching issues caused by data errors or inconsistencies. For example, when there are slight character differences in the carrier serial number, the fuzzy matching algorithm can identify that the data belongs to the same carrier and correctly associate them.

[0044] In an embodiment of the present invention, based on step S3, a possible embodiment will be given below to illustrate its specific implementation scheme in a non-limiting manner.

[0045] S301. Construct prompt words based on the structured field table; input the prompt words into a generative big model, so that the big model generates field definitions and attribute description information for the fields in the structured field table.

[0046] Based on the structured field table, prompt words are constructed and input into the generative large model so that the large model generates field definitions and attribute descriptions for the fields in the structured field table. The construction of prompt words needs to clearly and accurately express the requirements for the large model and guide the large model to generate results that meet the requirements. The content of the prompt words usually includes an overall description of the structured field table, the names of each field, and detailed requirements for the field definitions and attribute descriptions to be generated. For example, the prompt words can be constructed as follows: "The following is a structured field table for substrate inspection data, including table items such as field name, data source, and anomaly identification results. Please generate an accurate field definition for each field to explain the meaning of the data represented by the field; at the same time, generate detailed attribute description information, including data type, value range, unit, etc. The field names are as follows: [list all field names]".

[0047] When selecting a large generative model, it's important to consider its performance, accuracy, and applicability. Currently, some advanced pre-trained language models, such as the GPT series and BERT, have demonstrated excellent performance in natural language processing tasks. You can select an appropriate model based on your specific needs and fine-tune it to accommodate the specialized terminology and data characteristics of substrate inspection. After inputting the prompt words into the large model, the results generated by the large model need to be evaluated and screened. A combination of manual review and automated evaluation can be used to check the generated field definitions and attribute descriptions to ensure their accuracy and completeness. For results that do not meet your requirements, adjust the prompt words or retrain the model until satisfactory results are achieved.

[0048] S302. Generate a corresponding English field definition for the field definition using a self-explanatory snake case naming method.

[0049] Generate corresponding English field definitions for field definitions using the self-explanatory snake case naming convention. Snake case is a common naming convention that uses underscores to create more readable and self-explanatory names. When generating English field definitions, first understand the Chinese meaning of the field definition and then translate it into an appropriate English word or phrase. For example, for the "Boardboard Length" field definition, the English definition could use the snake case naming convention of "board_length."

[0050] To improve the accuracy and consistency of generated English field definitions, a professional terminology translation library can be established, containing commonly used Chinese terms in the field of substrate testing and their corresponding English translations. When generating English field definitions, query the translation library to ensure the use of standard English terminology. Furthermore, for complex field definitions, a step-by-step translation and combination approach can be employed. For example, for a field definition like "Resistance value in substrate electrical performance test," you can first translate "substrate," "electrical performance test," and "resistance value" separately, then combine them using the snake case nomenclature to obtain "board_electrical_performance_test_resistance_value."

[0051] S303. Use a pre-established domain term comparison table to unify the terms in the attribute description information.

[0052] Use a pre-established domain terminology comparison table to unify the terminology used in attribute descriptions. This table requires collecting and organizing specialized terminology for substrate inspection, including synonyms and antonyms used across different manufacturers and standards. For example, when describing the "size" of a substrate, different terms such as "specification dimensions" and "external dimensions" may appear. These terms need to be unified into a standard "size" description.

[0053] When standardizing terminology, an automated term replacement program can be developed. This program reads attribute description information, matches it against a domain terminology table, and replaces non-standard terms with standard terms. To ensure the accuracy of term replacement, manual review of the replaced content is required to avoid incorrect replacements. Furthermore, as the industry develops and technology advances, the domain terminology table needs to be continuously updated and improved to promptly incorporate emerging professional terms and eliminate outdated terms, ensuring the standardization and accuracy of attribute description information.

[0054] S304. Use regular rules to check unit symbols and range expressions.

[0055] Use regular expressions to validate unit symbols and range expressions. Regular expressions are a powerful text matching tool that can be used to verify that strings conform to specific formatting requirements. To validate unit symbols, define a valid regular expression for unit symbols based on the International System of Units (SI) and industry standards. For example, a length unit can be defined as "^m|cm|mm$," indicating that the unit symbol can only be "m" (meter), "cm" (centimeter), or "mm" (millimeter). If the unit symbol in the attribute description does not conform to the regular expression, the system will display an error and require correction.

[0056] Validating range expressions also requires defining corresponding regular expressions. For example, for the numeric range expression "[lower limit value, upper limit value]," a regular expression can be defined to ensure the correct format of the range expression. Furthermore, logical validation of the lower and upper limits in the range expression is required to ensure that the lower limit is less than or equal to the upper limit. In practical applications, regular rule validation can be integrated into data processing systems. Automatic validation is performed when attribute description information is entered or modified, promptly identifying and correcting errors and ensuring data quality.

[0057] In an embodiment of the present invention, based on step S4, a possible embodiment will be given below to illustrate its specific implementation scheme in a non-limiting manner.

[0058] Upload structured data with field definitions and attribute descriptions to a database. When selecting a database, consider factors such as data size, access performance, and data security. For large-scale substrate inspection data, you can typically choose a relational database such as MySQL or Oracle, or a non-relational database such as MongoDB or Redis. Relational databases are suitable for storing structured data and have comprehensive data integrity constraints and transaction processing mechanisms. Non-relational databases offer greater scalability and flexibility, making them suitable for processing massive amounts of unstructured and semi-structured data.

[0059] Before uploading data, you need to design an appropriate data table structure based on the database's characteristics and requirements. Map the fields in the structured field table to database table columns and set appropriate constraints, such as data types, primary keys, and foreign keys. For example, the "Carrier Serial Number" field can be set as a primary key to ensure data uniqueness. Fields related to other tables can be set as foreign keys to establish relationships between tables. During the data upload process, batch inserts can be used to improve data transmission efficiency and reduce database load. To ensure data accuracy and completeness, the uploaded data must be verified and validated, such as for correct formatting and the presence of duplicate data.

[0060] Because joint indexes offer higher performance than single indexes of equal length, their design is crucial. When designing an index, you first need to perform data mining on the index fields to analyze potential relationships. Data mining algorithms, such as association rule mining algorithms (Apriori and FP-Growth), can be used to analyze structured data and identify frequently occurring field combinations. These combinations are likely to have strong correlations and are therefore suitable candidates for joint indexes.

[0061] After identifying candidate fields, they are screened and optimized based on database development specifications and index design principles. Factors to consider include field selectivity (higher selectivity leads to higher index efficiency), field update frequency (frequently updated fields are not suitable for indexing), and index size. Through comprehensive evaluation, the optimal field combination is selected as the joint index. After determining the target joint index as the field combination with the highest confidence level in the target item set, a data table is generated based on the English definitions of the fields and the target joint index. The data table design should consider data storage efficiency and query performance, and appropriately arrange the field order and data types.

[0062] Import the data corresponding to the English field definitions into the database using the Spark framework. Spark is a powerful big data processing framework with efficient data processing capabilities and excellent scalability. When importing data using Spark, first convert the data to a format supported by Spark, such as Parquet or CSV. Then, use Spark's DataFrame API to read the data and perform necessary data cleaning and transformation operations, such as removing duplicate data and filling missing values.

[0063] In some embodiments, the IC substrate detection data processing system based on edge computing may include multiple functional modules composed of computer program segments. The computer program of each program segment in the IC substrate detection data processing system based on edge computing may be stored in a memory of a computer device and executed by at least one processor to perform (see Figure 1 (Description) IC substrate detection data processing function based on edge computing.

[0064] In this embodiment, the IC substrate detection data processing system based on edge computing can be divided into multiple functional modules according to the functions it performs, such as Figure 2 As shown. The module referred to in the present invention refers to a series of computer program segments that can be executed by at least one processor and can perform fixed functions, which are stored in a memory. In this embodiment, the functions of each module will be described in detail in subsequent embodiments.

[0065] The anomaly recognition module is used to obtain multi-modal detection data and perform anomaly recognition on the detection data of each modality; A data conversion module is used to generate structured data based on each modal data and the corresponding abnormality recognition results; A description generation module, configured to add field definitions and attribute description information to the fields of the structured data using a generative large model; The data upload module is used to upload structured data with field definitions and attribute description information to the database.

[0066] Figure 3 The IC substrate detection data processing method based on edge computing provided for the embodiment of the present application can be applied to the device. Those skilled in the art will understand that the device structure involved in the embodiment of the present invention does not constitute a limitation on the device, and the device may include more or fewer components than shown in the figure, or combine certain components, or arrange the components differently. In an embodiment of the present invention, the device includes but is not limited to a laptop computer, a desktop computer, a workbench, a personal digital assistant, a server, a blade server, a mainframe computer, and other suitable computers. The device can also represent various forms of mobile devices, such as personal digital processing, cellular phones, smart phones, wearable devices and other similar computing devices. The components shown herein, their connections and relationships, and their functions are merely examples and are not intended to limit the implementation of the embodiments of the present application described and / or required herein.

[0067] The device 300 may include a processor 310, a memory 320, and a communication unit 330. These components communicate via one or more buses. Those skilled in the art will appreciate that the server structure shown in the figure does not limit the present invention. The server structure may be a bus structure or a star structure, and may include more or fewer components than shown, or combine certain components, or arrange the components differently.

[0068] The memory 320 can be used to store execution instructions of the processor 310. The memory 320 can be implemented by any type of volatile or non-volatile memory device, or a combination thereof, such as static random access memory (SRAM), electrically erasable programmable read-only memory (EEPROM), erasable programmable read-only memory (EPROM), programmable read-only memory (PROM), read-only memory (ROM), magnetic memory, flash memory, magnetic disk, or optical disk. When the execution instructions in the memory 320 are executed by the processor 310, the device 300 can perform some or all of the steps in the above-described method embodiments.

[0069] The processor 310 is the control center of the storage device, which uses various interfaces and lines to connect various parts of the entire electronic device. It executes various functions of the electronic device and / or processes data by running or executing software programs and / or modules stored in the memory 320, and calling data stored in the memory. The processor can be composed of an integrated circuit (IC), for example, it can be composed of a single packaged IC, or it can be composed of multiple packaged ICs with the same or different functions. For example, the processor 310 can only include a central processing unit (CPU). In an embodiment of the present invention, the CPU can be a single computing core or multiple computing cores.

[0070] The communication unit 330 is configured to establish a communication channel so that the storage device can communicate with other devices, receive user data sent by other devices, or send user data to other devices.

[0071] The present invention also provides a computer storage medium, wherein the computer storage medium may store a program that, when executed, may include some or all of the steps of each embodiment provided herein. The storage medium may be a magnetic disk, an optical disk, a read-only memory (ROM), or a random access memory (RAM).

[0072] Those skilled in the art will clearly understand that the techniques in the embodiments of the present invention can be implemented using software and a necessary general-purpose hardware platform. Based on this understanding, the technical solutions in the embodiments of the present invention, or the portion that contributes to the prior art, can be embodied in the form of a software product. This computer software product is stored in a storage medium such as a USB flash drive, a mobile hard drive, a read-only memory (ROM), a random access memory (RAM), a magnetic disk, or an optical disk, among other media capable of storing program code, and includes instructions for causing a computer device (which can be a personal computer, a server, or a second device, a network device, etc.) to execute all or part of the steps of the methods described in various embodiments of the present invention.

[0073] In this specification, the same or similar parts between the various embodiments can be referred to each other. In particular, for the device embodiment, since it is basically similar to the method embodiment, the description is relatively simple, and the relevant parts can be referred to the description of the method embodiment.

[0074] In the several embodiments provided by the present invention, it should be understood that the disclosed systems and methods can be implemented in other ways. For example, the system embodiments described above are merely illustrative. For example, the division of the modules is merely a logical function division. In actual implementation, there may be other division methods, such as multiple modules or components can be combined or integrated into another system, or some features can be ignored or not executed. In addition, the mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, indirect coupling or communication connection of systems or modules, and can be electrical, mechanical or other forms.

[0075] The modules described as separate components may or may not be physically separate, and the components shown as modules may or may not be physical modules, that is, they may be located in one place or distributed across multiple network modules. Some or all of the modules may be selected to achieve the purpose of the present embodiment according to actual needs.

[0076] In addition, each functional module in each embodiment of the present invention may be integrated into one processing module, or each module may exist physically separately, or two or more modules may be integrated into one module.

[0077] Although the present invention has been described in detail with reference to the accompanying drawings and in conjunction with preferred embodiments, the present invention is not limited thereto. Without departing from the spirit and essence of the present invention, persons of ordinary skill in the art may make various equivalent modifications or substitutions to the embodiments of the present invention, and such modifications or substitutions shall be within the scope of the present invention. Any changes or substitutions that can be easily conceived by persons skilled in the art within the technical scope disclosed in the present invention shall be within the scope of protection of the present invention.

Claims

1. A method for processing IC substrate detection data based on edge computing, characterized in that: include: Acquire multi-modal detection data and perform anomaly identification on the detection data of each modality; Generate structured data based on each modal data and corresponding abnormality recognition results; Using a generative big model to add field definitions and attribute description information to the fields of the structured data; Upload structured data with field definitions and attribute descriptions to the database.

2. The method according to claim 1, characterized in that Acquire multimodal detection data, including: Obtain physical structure data, electrical performance data, environmental data, and process defect data of the carrier board; Among them, physical structure data includes structured specification and dimension data, X-ray detection images, and ultrasonic scanning images; electrical performance data and environmental data are both structured data; and process defect data includes automatic optical detection images.

3. The method according to claim 2, characterized in that Perform abnormality identification on the detection data of each modality separately, including: Use pre-set threshold comparison rules to identify anomalies in structured data; An image recognition model is used to identify anomalies in unstructured data, including X-ray detection images, ultrasound scanning images, and automatic optical detection images.

4. The method according to claim 1, wherein Generate structured data based on each modal data and the corresponding abnormality recognition results, including: Setting a structured field table template, wherein the table items of the structured field table template include field name, data source and anomaly recognition result; Assigning a field name to each parameter item in the detection data, and filling the field name, data source and anomaly recognition result corresponding to each parameter item into the corresponding table item of the structured field table template to obtain a structured field table; A cross-modal exception association field identifier is added to the structured field table.

5. The method according to claim 4, characterized in that A generative big model is used to add field definitions and attribute description information to the fields of the structured data, including: Constructing prompt words based on the structured field table; The prompt words are input into a generative big model so that the big model generates field definitions and attribute description information for the fields in the structured field table.

6. The method according to claim 5, characterized in that The method further comprises: Use self-explanatory snake case naming to generate corresponding English field definitions for field definitions; Use pre-established domain term comparison tables to unify the terms in attribute description information.

7. The method according to claim 5, characterized in that The method further comprises: Use regular expressions to validate unit symbols and range expressions.

8. An IC substrate detection data processing system based on edge computing, characterized in that: include: The anomaly recognition module is used to obtain multi-modal detection data and perform anomaly recognition on the detection data of each modality; A data conversion module is used to generate structured data based on each modal data and the corresponding abnormality recognition results; A description generation module, configured to add field definitions and attribute description information to the fields of the structured data using a generative large model; The data upload module is used to upload structured data with field definitions and attribute description information to the database.

9. The system according to claim 8, characterized in that Acquire multimodal detection data, including: Obtain physical structure data, electrical performance data, environmental data, and process defect data of the carrier board; Among them, physical structure data includes structured specification and dimension data, X-ray detection images, and ultrasonic scanning images; electrical performance data and environmental data are both structured data; and process defect data includes automatic optical detection images.

10. An IC substrate detection data processing device based on edge computing, characterized in that: include: A memory, used to store an IC substrate detection data processing program based on edge computing; A processor is used to implement the steps of the IC substrate detection data processing method based on edge computing as described in any one of claims 1 to 7 when executing the IC substrate detection data processing program based on edge computing.

Images