Method and system for automatically processing AOI detection data in semiconductor industry

Abstract

The invention discloses a method and system for automatically processing AOI detection data in the semiconductor industry, and belongs to the technical field of intelligent manufacturing of semiconductor manufacturing enterprises, and the method comprises the steps: based on an equipment feature knowledge base, according to AOI equipment feature information, matching an equipment feature knowledge base, loading a corresponding analysis engine, and extracting measurement data from AOI equipment; analyzing the measurement data and cleaning abnormal values to generate data in a standardized format: transmitting the generated data in the standardized format to a statistical process control system through branches by a dynamic routing mechanism according to real-time requirements of detection parameters; the integrated statistical process control algorithm calculates the data in the standardized format, a multi-dimensional detection report is generated, and when the process capability index value is lower than a set threshold value, abnormal early warning is triggered. According to the method, the misjudgment rate of the system is reduced, the data processing efficiency and accuracy are improved, height overrun, yield fluctuation and the like of the solder balls are positioned in time, and the method has reliability and engineering applicability.

Description

Technical Field

[0001] This invention relates to the field of intelligent manufacturing technology for semiconductor manufacturing enterprises, specifically to a method and system for automated processing of AOI inspection data in the semiconductor industry. Background Technology

[0002] Currently, the data processing workflow for Automated Optical Inspection (AOI) equipment in the advanced semiconductor packaging testing process is highly reliant on manual labor. For example, critical inspection data such as solder ball height and yield must be manually exported from the AOI equipment and entered into the SPC system, taking over 20 minutes per batch and posing a risk of data transcription errors. Furthermore, due to compatibility barriers among multiple vendors, AOI equipment output formats vary significantly, including SEMF and CTF. Manual analysis and software configuration for file structure parsing and data mapping are required, potentially leading to system adaptation cycles of up to one week. Moreover, it only supports one or two brands of equipment, failing to meet the needs of semiconductor companies' multi-brand mixed-line production, thus lacking applicability. In addition, the current AOI data processing workflow lacks real-time control capabilities. Manually generated quality reports are generated on a shift-by-shift basis, with abnormal batches taking up to two hours to respond. This makes it difficult to promptly locate issues such as solder ball height exceeding limits and yield fluctuations, resulting in delayed process adjustments and impacting packaging yield. In summary, the current AOI data processing workflow suffers from long processing cycles and relatively low efficiency due to manual data handling, making it difficult to meet the real-time data synchronization and control requirements of large-scale production lines in advanced semiconductor packaging plants. Furthermore, the lack of a unified protocol parsing and standardized conversion mechanism addresses the heterogeneous data formats of AOI equipment from various manufacturers, resulting in relatively poor compatibility. The insufficient real-time performance of data processing and delayed quality data feedback make it difficult to support the real-time monitoring requirements of advanced semiconductor packaging plants for the Process Capability Kilometer (CPK), necessitating further optimization. Summary of the Invention

[0003] This invention provides a method and system for automated processing of AOI inspection data in the semiconductor industry. The purpose is to solve the problems of the current AOI data processing process, which has a long data processing cycle and is difficult to meet the needs of large-scale production lines in advanced semiconductor packaging plants; the heterogeneous data formats of AOI equipment from different manufacturers and relatively poor compatibility; insufficient real-time data processing and delayed quality data feedback, which make it difficult to support the real-time monitoring requirements of advanced semiconductor packaging plants for production process capability indices.

[0004] To achieve the above objectives, the present invention adopts the following technical solution: This invention provides a method for automated processing of AOI inspection data in the semiconductor industry, comprising the following steps: S1. Based on the device feature knowledge base, use the SECS / GEM protocol adapter to send a device query command to the AOI device, match the device feature knowledge base according to the AOI device feature information and load the corresponding parsing engine to extract measurement data from the AOI device. S2. The measurement data extracted using a predefined regular expression template library is parsed to obtain parsed data; outlier cleaning of the parsed data is performed using a combination of three Sigma statistical methods and an LSTM model, generating data in a standardized format: S3. Based on the real-time requirements of the detection parameters, the generated standardized format data is transmitted to the statistical process control system through a dynamic routing mechanism. S4 integrates statistical process control algorithms to calculate standardized data in statistical process control systems, generating multi-dimensional inspection reports containing process capability index values. When the process capability index value is lower than a set threshold, it triggers an anomaly warning, completing the automated processing of AOI inspection data in the semiconductor industry.

[0005] In some implementations, S1 specifically includes: S11: By analyzing the SECS / GEM protocol features of various AOI devices, a device feature knowledge base containing device model, communication protocol version, and data format template is constructed. S12: Match the device feature knowledge base based on the device model, load the parsing engine corresponding to various AOI devices, and extract the ball height, yield, 2D and 3D feature value parameters.

[0006] In some implementations, in S2, the construction of the device feature knowledge base includes: using the Measurement Data tag structure of the Semiconductor Device Manufacturing Format 3.0 protocol and the field encoding of the Camtek Transmission Format 2.0 protocol to parse the data, and dynamically updating the parsing results through a locally deployed online learning algorithm.

[0007] In some implementations, S2 specifically includes: S21: Use the predefined regular expression template library to parse the raw AOI data; S22: Use three sigma statistical methods and long short-term memory network models to dynamically clean outliers and output data in a standardized format.

[0008] Furthermore, in S21, the regular expression template library includes: predefined templates for parsing KLA semiconductor device manufacturing format files.

[0009] Furthermore, in S22, the three sigma statistical cleaning includes: calculating the mean and standard deviation of historical data, filtering outliers that exceed the range of mean minus three standard deviations and mean plus three standard deviations, and predicting process fluctuations using an LSTM model in order to dynamically adjust the cleaning threshold.

[0010] In some implementations, S3 specifically includes: S31: Configure the mapping relationship between detection parameters and statistical process control indicators through a visual interface, and generate a mapping table in Extensible Markup Language format; S32: The real-time monitoring module transmits real-time parameters to the statistical process control system via the SECS / GEM protocol, and transmits statistical parameters to the historical database of the statistical process control system via the MQTT protocol.

[0011] Furthermore, in S32, when transmitting solder ball height, surface height, or three-dimensional feature value parameters, the transmission delay is no greater than 200ms; when querying yield or two-dimensional feature value parameters, the query response time is no greater than 3s.

[0012] In some implementations, S4 specifically includes: S41: Call the algorithm library of the statistical process control system to calculate the process capability index value and draw the control chart; S42: Generate a dynamic report that includes the average height curve of the spheres and a heatmap of yield fluctuations; S43: When the process capability index value is less than the preset value, a red warning is marked, a portable document format report containing abnormal parameter information, impact range and process adjustment suggestions is generated, and the report is pushed through the office automation system.

[0013] This invention also provides a system for automated processing of AOI inspection data in the semiconductor industry. The system includes a multi-source data acquisition module, an intelligent format conversion module, a dynamic routing module, and an intelligent reporting module, wherein: Multi-source data acquisition module: Based on the device feature knowledge base, it sends device query commands to the AOI device using the SECS / GEM protocol adapter, matches the device feature knowledge base with the AOI device feature information and loads the corresponding parsing engine to extract measurement data from the AOI device; The intelligent format conversion module is used to parse the measurement data extracted using a predefined regular expression template library to obtain parsed data; it combines the three sigma statistical method and the LSTM model to clean outliers from the parsed data and generate data in a standardized format. Dynamic routing module: Used to transmit the generated standardized data to the statistical process control system through a dynamic routing mechanism based on the real-time requirements of the detection parameters. Intelligent Reporting Module: This module integrates statistical process control algorithms to calculate standardized data in statistical process control systems, generating multi-dimensional inspection reports that include process capability index values. It triggers anomaly warnings when the process capability index value falls below a set threshold, thus automating the processing of AOI inspection data in the semiconductor industry.

[0014] Compared with existing technologies, the present invention provides a method and system for automated processing of AOI inspection data in the semiconductor industry, which has the following advantages: This invention provides an automated processing method for AOI (Automated Optical Inspection) data in the semiconductor industry. It employs an automatic extraction of AOI measurement data based on a device feature knowledge base, improving compatibility across multiple vendors. The automatic matching and parsing engine, powered by the device feature knowledge base, replaces manual identification of device protocols and data export, eliminating transcription errors. Standardized data is generated through regular expression parsing and 3σ / LSTM cleaning. A regular expression template library provides a unified parsing entry point for heterogeneous data, avoiding manual file format analysis and improving real-time performance. Dynamic routing is used, routing data to the SPC (System-Specific Processing) system based on real-time requirements. This dynamic routing mechanism intelligently allocates transmission paths according to parameter real-time needs, overcoming the latency of traditional reports. Through SPC algorithm integration, this invention automatically generates reports and anomaly warnings. Real-time calculation of process capability index values ​​(e.g., CPK < threshold) triggers warnings within seconds, reducing anomaly response time from hours to seconds, addressing the lack of real-time monitoring and improving system reliability. This invention reduces the system's false positive rate and improves data accuracy by filtering obvious anomalies using 3σ filtering and dynamically adjusting thresholds to predict process fluctuations through LSTM collaborative cleaning. It integrates trend charts and heatmaps through multi-dimensional reports to more intuitively identify problems such as ball over-limit and yield fluctuations. It replaces manual statistics and report preparation with automated report generation and early warning, improving efficiency and accuracy. This, in turn, optimizes the overall reliability and engineering applicability of the system. Attached Figure Description

[0015] The accompanying drawings are provided to further understand the invention and constitute a part of this invention. The illustrative embodiments of the invention and their descriptions are used to explain the invention and do not constitute an improper limitation of the invention.

[0016] Figure 1 This is a schematic diagram of the process framework for an automated processing method for AOI inspection data in the semiconductor industry according to the present invention. Figure 2 This is a schematic diagram of the architecture of an advanced packaging and bonding process wafer-level operation control system according to the present invention. Detailed Implementation

[0017] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. The components of the embodiments of the present invention described and shown in the accompanying drawings can generally be arranged and designed in various different configurations.

[0018] Therefore, the following detailed description of the embodiments of the invention provided in the accompanying drawings is not intended to limit the scope of the claimed invention, but merely to illustrate selected embodiments of the invention. All other embodiments obtained by those skilled in the art based on the embodiments of the invention without inventive effort are within the scope of protection of the invention.

[0019] It should be noted that, in this document, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Unless otherwise specified, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.

[0020] It should be noted that the apparatus and methods disclosed in the embodiments herein can also be implemented in other ways. The apparatus embodiments described above are merely illustrative; for example, the flowcharts and block diagrams in the accompanying drawings show the architecture, functionality, and operation of possible implementations of apparatus, methods, and computer program products according to various embodiments herein. In this regard, each block in a flowchart or block diagram may represent a module, program, or part of code containing one or more executable instructions for implementing a specified logical function. It should also be noted that in some alternative implementations, the functions marked in the blocks may occur in a different order than those marked in the drawings. For example, two consecutive blocks may actually be executed substantially in parallel, and they may sometimes be executed in reverse order, depending on the functions involved. It should also be noted that each block in a block diagram and / or flowchart, and combinations of blocks in block diagrams and / or flowcharts, can be implemented using a dedicated hardware-based system to perform the specified function or action, or can be implemented using a combination of dedicated hardware and computer instructions.

[0021] In addition, the functional modules in the various embodiments of this article can be integrated together to form an independent part, or each module can exist independently, or two or more modules can be integrated to form an independent part.

[0022] How to provide a standardized method for generating and collecting AOI inspection data from multiple vendors' equipment, intelligent format conversion, dynamic routing transmission, and statistical process control (SPC) report generation, thereby achieving full-process automated processing of AOI inspection data from the equipment end to the SPC system and the reporting system, and improving accuracy and safety.

[0023] This invention provides a method for automated processing of AOI inspection data in the semiconductor industry, comprising the following steps: S1. Based on the device feature knowledge base, use the SECS / GEM protocol adapter to send a device query command to the AOI device, match the device feature knowledge base according to the AOI device feature information and load the corresponding parsing engine to extract measurement data from the AOI device. S2. The measurement data extracted using a predefined regular expression template library is parsed to obtain parsed data; outlier cleaning of the parsed data is performed using a combination of three Sigma statistical methods and an LSTM model, generating data in a standardized format: S3. Based on the real-time requirements of the detection parameters, the generated standardized format data is transmitted to the statistical process control system through a dynamic routing mechanism. S4 integrates statistical process control algorithms to calculate standardized data in statistical process control systems, generating multi-dimensional inspection reports containing process capability index values. When the process capability index value is lower than a set threshold, it triggers an anomaly warning, completing the automated processing of AOI inspection data in the semiconductor industry.

[0024] This invention provides a method for automating AOI inspection data processing in the semiconductor industry. It offers a comprehensive solution integrating innovative methods and system architecture, enabling automated acquisition, intelligent processing, and SPC system integration of inspection data from multiple vendors' AOI equipment. This improves upon the low efficiency, poor compatibility, and insufficient real-time performance of traditional AOI inspection data processing workflows, thereby enhancing the intelligent level of quality control in advanced semiconductor packaging plants. This invention focuses on the full-process automated processing of semiconductor AOI inspection data, sharing equipment feature knowledge bases, dynamic routing mechanisms, and 3σ data cleaning to form a technology encompassing AOI inspection data acquisition and quality control. The method reduces single-batch data processing time from 30 minutes to 2 minutes, decreases manual operation by 90%, and reduces the data entry error rate from 3% to 0.3%. Equipment adaptation time is shortened from one week to two hours, supporting major mainstream AOI equipment brands, potentially covering over 90% of production lines' AOI equipment. The method of this invention achieves a synchronization delay of ≤100ms for ball height and yield parameters, reduces the abnormal batch identification time from 3 hours to 10 minutes, and improves the yield fluctuation response speed by 10 times. 3σ+LSTM cleaning reduces the outlier false positive rate to 0.02%, and the CPK calculation accuracy is above 99.7%. The device feature knowledge base and regular expression template library used in this invention construct a multi-vendor universal parsing platform with a compatibility rate of 98%. Dynamic routing efficiency is improved by 40%, and the use of FPGA acceleration hardware enables the system data throughput to reach over 10MB / s, supporting concurrent access from thousands of devices, demonstrating excellent applicability.

[0025] In some embodiments, the method for automated processing of AOI inspection data in the semiconductor industry according to the present invention includes a knowledge base containing device model, protocol version and data template, supports dynamic loading of multi-brand engines such as KLA or Camtek, and has an adaptation cycle of 1 week to 2 hours; extracts key parameters such as ball height, yield, 2D or 3D feature values, covering core packaging quality indicators.

[0026] Furthermore, the knowledge base of this invention, by parsing the SEMF3.0 / CTF2.0 protocol structure and pre-setting KLASEMF parsing templates, achieves an initial matching accuracy of over 97%. It dynamically updates rules through online learning algorithms to adapt to device firmware upgrades, avoiding repetitive manual configuration. The method of this invention is plug-and-play compatible with devices from multiple vendors.

[0027] This invention provides a method for automated processing of AOI inspection data in the semiconductor industry. It employs a regular expression library to parse the raw data, solving the problem of manual conversion between multiple formats. Standardized data is output using 3σ and LSTM, unifying CSV, JSON, or TXT formats for subsequent processing and analysis. Outliers are filtered based on μ±3σ to remove obviously erroneous data. Thresholds are dynamically adjusted using LSTM to predict process fluctuations and adaptively adjust the range, reducing the false negative rate to 0.02%. This invention's anomaly cleaning method considers both statistical rules and process fluctuation prediction, and the standardized data seamlessly integrates with SPC systems, making it highly applicable.

[0028] Furthermore, the method of this invention allows for the visualization and configuration of SPC metric mapping, enabling users to define monitoring logic such as ball height to CPK; real-time parameters are transmitted via SECS or GEM with a latency of no more than 200ms, while traditional MQTT parameter calculations have a response time of no more than 3 seconds, with bandwidth allocated on demand; and the transmission time for ball height or 3D feature values ​​is no more than 200ms, and yield queries are no more than 3 seconds, meeting the real-time calculation requirements of SPC. This invention achieves millisecond-level access to the monitoring module for high real-time parameters and efficient archiving and querying of statistical parameters.

[0029] Furthermore, the method of this invention combines CPK calculation and control chart plotting to quantify process stability, achieving an accuracy rate of over 99.7%. Dynamic reports support drill-down to 2D or 3D batch images for rapid identification of anomaly root causes. When CPK is set < 1.33, PDF reports and OA push notifications are generated with a response time of ≤ 1 second. This invention achieves a second-level closed loop from anomaly detection to process adjustment, and mobile-adaptive reports improve on-site response speed.

[0030] This invention also provides a system for automated processing of AOI inspection data in the semiconductor industry. The system includes a multi-source data acquisition module, an intelligent format conversion module, a dynamic routing module, and an intelligent reporting module, wherein: Multi-source data acquisition module: Based on the device feature knowledge base, it sends device query commands to the AOI device using the SECS / GEM protocol adapter, matches the device feature knowledge base with the AOI device feature information and loads the corresponding parsing engine to extract measurement data from the AOI device; The intelligent format conversion module is used to parse the measurement data extracted using a predefined regular expression template library to obtain parsed data; it combines the three sigma statistical method and the LSTM model to clean outliers from the parsed data and generate data in a standardized format. Dynamic routing module: Used to transmit the generated standardized data to the statistical process control system through a dynamic routing mechanism based on the real-time requirements of the detection parameters. Intelligent Reporting Module: This module integrates statistical process control algorithms to calculate standardized data in statistical process control systems, generating multi-dimensional inspection reports that include process capability index values. It triggers anomaly warnings when the process capability index value falls below a set threshold, thus automating the processing of AOI inspection data in the semiconductor industry.

[0031] The following detailed description of a method and system for automated processing of AOI inspection data in the semiconductor industry, based on specific embodiments, further illustrates the present invention.

[0032] like Figure 1 As shown, this embodiment achieves automatic adaptation and intelligent processing of AOI equipment data from multiple semiconductor packaging plants through a standardized process, solving the efficiency, compatibility, and real-time bottlenecks of manual on-site operations. The method includes: Step S1: Multi-source data standardization acquisition method, which involves constructing a device feature knowledge base, establishing device protocol features to determine the location data acquisition and extraction method, and automatically capturing measurement data generated by AOI equipment; specifically including S11, S12 and S13: Step S11: In the construction of the equipment feature knowledge base, the AOI equipment in the production workshop is connected to the production network. Locally deployed web crawler technology is used to parse the firmware information of AOI equipment from manufacturers such as KLA and Camtek, extracting the SEMF 3.0 protocol's Measurement Data tag structure and the CTF 2.0 protocol's field encoding rules to identify the equipment's SECS / GEM protocol features. A device feature knowledge base containing equipment model, communication protocol version, and data format templates is constructed, and a parsing engine is built. This library supports automatic identification of equipment from multiple mainstream manufacturers, with an initial matching accuracy of over 97%. The parsing rules are dynamically updated through a locally deployed online learning algorithm to improve the parsing engine.

[0033] Step S12: In the automatic parameter extraction, the EAP system sends a device query command to the AOI device through the SECS / GEM protocol adapter, obtains the device feature information, matches it with the device feature knowledge base, and then automatically loads the corresponding parsing engine. The KLA device loads the SEMF parsing engine to extract parameters such as sphere height (unit: μm), yield (percentage), and 2D / 3D feature values.

[0034] Step S2: Intelligent format conversion and data cleaning. This step uses a regular expression template library to parse the original AOI detection data from multiple vendors, inputs the difference between the predicted and measured values ​​from the LSTM model, and combines 3σ statistics to identify outliers and generate standardized data. Specifically, this includes steps S21 and S22: Step S21: In regular expression parsing: For different vendor data formats, a predefined regular expression template library is called. For KLA SEMF files, Pad Height is used (\d+\.\d+) to extract the high value of pad, thus achieving efficient parsing of the original data.

[0035] Step S22: In the 3σ and LSTM cleaning process, the 3σ statistical method is combined with LSTM for dynamic cleaning. The mean (μ) and standard deviation (σ) of historical data are calculated to filter out outliers that exceed the range of (μ-3σ, μ+3σ). At the same time, the LSTM model can be used to predict production line process fluctuations and dynamically adjust the cleaning threshold. Finally, standardized data containing equipment number, batch number and detection time are generated in csv / json / txt format.

[0036] Step S3: Dynamic routing and protocol conversion strategy. A mapping table is generated by configuring the mapping relationship between detection parameters and SPC metrics. The intelligent routing engine performs data splitting based on real-time data requirements according to the device characteristic knowledge base. This includes steps S31 and S32.

[0037] In step S31: SPC index mapping configuration: the mapping relationship between detection parameters and SPC indexes is defined through a visual interface, the calculation of Pad height and CPK value, yield and P value statistics are defined, and an editable XML format mapping table is generated, which supports user-defined Pad height specifications of 50μm-150μm.

[0038] In the intelligent routing transmission in step S32, parameters with high real-time requirements for Pad, Surface, Height, and 3D feature values ​​are transmitted to the real-time monitoring module of the SPC system via the SECS / GEM protocol, triggering real-time calculation of CPK values ​​at a frequency of no less than 10 times / minute and a transmission delay of no more than 200ms. For yield and 2D feature value statistical parameters, they are transmitted to the SPC historical database for archiving and storage via the MQTT protocol, supporting efficient querying by batch / device model, with a query response time set to less than 3 seconds.

[0039] Step S4: Multi-dimensional intelligent report generation, integrating SPC algorithm to calculate CP / CPK values, generating ball height trend charts and yield comparison heatmaps, real-time monitoring of CPK values, and achieving second-level response to quality anomalies. This includes steps S41, S42, and S43.

[0040] Step S41: In SPC algorithm integration, the algorithm library of the SPC system is called to calculate the CP / CPK value, draw the control charts of X-bar and P-charts, and support user-defined process capability evaluation criteria.

[0041] Step S42: Report type generation in progress. Display the average height curve of the ball in 5-minute increments. Anomalies are marked with equipment number and inspection time. Generate yield fluctuation heatmaps by equipment model, batch, and shift. Supports drilling down to 2D / 3D inspection images of specific batches. Step S43: In the anomaly warning mechanism, the SPC system automatically calculates the CPK value. When the CPK is less than 1.33, a red warning is marked, and a PDF report containing abnormal parameter information, the scope of impact, and process adjustment suggestions is generated. Anomaly report notifications are pushed through the OA system, with the notification title customizable to include the identifier "CPK Anomaly Warning Batch Number LotX - CPK Value DataY," and the response time is less than 10 seconds.

[0042] like Figure 2 As shown, the system for automated processing of AOI inspection data in the semiconductor industry of this invention provides an efficient execution platform for the above-mentioned method process through deep hardware and software integration, meeting the high-concurrency and low-latency data processing needs of advanced semiconductor packaging plants. This system includes: Multi-source data acquisition module; This module corresponds to step S1, which enables connection to the AOI device and parameter extraction, as detailed below: SECS / GEM Protocol Adapter: At the hardware level, it supports network protocols such as TCP / IP and UDP to achieve physical connection with various AOI devices (such as KLA, Camtek, etc.); at the software level, it encapsulates the SECS / GEM semiconductor-specific protocol and supports interactive commands such as device status query and data request.

[0043] Equipment Feature Knowledge Base Database: After the equipment feature knowledge base is generated, it uses a relational database to store information such as equipment protocol features, data format templates, and equipment models. It supports SQL queries and online updates, and the built-in equipment identification algorithm ensures that the initial matching accuracy is above 97%.

[0044] Parameter extraction engine group: Based on the matching results of the device feature knowledge base, the corresponding manufacturer's parsing engine (such as KLA parser, Camtek parser) is dynamically loaded, and the device detection parameters are extracted in parallel through multi-threading technology, thereby improving the processing speed.

[0045] Intelligent format conversion module; This module corresponds to method step S2, which implements data parsing and cleaning; Regular expression parsing engine: Developed based on a regular expression engine, it has built-in parsing template libraries from multiple vendors, supports users to customize parsing rules through a web interface, and provides a RESTful API interface for upper-layer application modules to call.

[0046] Data standardization middleware: Converts data parsed by the parsing engine into a data format that conforms to the ISA-95 standard, ensuring compatibility with various analysis systems such as SPC, and supports multiple output formats such as CSV / JSON / TXT.

[0047] 3σ data cleaning hardware unit: The 3σ statistical cleaning algorithm is implemented using FPGA acceleration technology, which improves the processing speed by 5 times compared with traditional CPU processing; it integrates an abnormal data log processing module to record the original data of AOI equipment for process traceability.

[0048] Dynamic routing module; This module corresponds to method step S3, which implements data routing and protocol conversion, specifically: SPC Metric Mapping Configuration Interface: A web-based visual interactive interface that supports drag-and-drop configuration of the mapping relationship between detection parameters and SPC metrics, and generates XML format configuration files in real time.

[0049] Intelligent routing engine: The hardware is deployed using high-performance servers, supporting bidirectional conversion between SECS / GEM and MQTT protocols; the software implements a QoS policy engine, which dynamically adjusts the transmission queue according to data priority (level 1-5).

[0050] Local cache database: It uses an SQLite cluster to store failed data transmissions and uses an exponential backoff algorithm to retransmit abnormal data. The success rate of retransmission is over 99%, and it supports breakpoint resume function.

[0051] Intelligent reporting module; This module corresponds to step S4, which generates analysis reports and provides anomaly alerts: The SPC algorithm calculation server is equipped with a statistical analysis library and SPC engine, supporting distributed computing with a CPK value calculation latency of no more than 200ms and supporting parallel processing of data from multiple production lines. The dynamic report generation system develops a web-based visualization component based on a general reporting platform, supporting interactive operations such as real-time trend chart zooming and batch data drill-down, with adaptive display on both PC and mobile (phone / tablet) terminals. The anomaly warning terminal integrates an audible and visual alarm device at the hardware level and can connect to enterprise email systems and OA platforms at the software level, enabling multi-channel warnings via email or internal messages, with an anomaly response time of no more than 10 seconds.

[0052] Finally, it should be noted that the above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention in any way. Anyone skilled in the art can readily implement the present invention according to the description and above. Any modifications, alterations, or equivalent variations made using the technical content disclosed above are equivalent embodiments of the present invention. Furthermore, any modifications, alterations, or variations made to the above embodiments based on the essential technology of the present invention are still within the protection scope of the present invention.

Claims

1. A method for automated processing of AOI inspection data in the semiconductor industry, characterized in that, Includes the following steps: S1. Based on the device feature knowledge base, use the SECS / GEM protocol adapter to send a device query command to the AOI device, match the device feature knowledge base according to the AOI device feature information and load the corresponding parsing engine to extract measurement data from the AOI device. S2. The measurement data extracted using a predefined regular expression template library is parsed to obtain parsed data; outlier cleaning of the parsed data is performed using a combination of three Sigma statistical methods and an LSTM model, generating data in a standardized format: S3. Based on the real-time requirements of the detection parameters, the generated standardized format data is transmitted to the statistical process control system through a dynamic routing mechanism. S4. The integrated statistical process control algorithm calculates standardized data in the statistical process control system, generates multi-dimensional inspection reports containing process capability index values, triggers anomaly warnings when the process capability index value is lower than a set threshold, and completes automated processing of AOI inspection data in the semiconductor industry.

2. The method for automated processing of AOI inspection data in the semiconductor industry according to claim 1, characterized in that, S1 specifically includes: S11: By analyzing the SECS / GEM protocol features of various AOI devices, a device feature knowledge base containing device model, communication protocol version, and data format template is constructed. S12: Match the device feature knowledge base based on the device model, load the parsing engine corresponding to various AOI devices, and extract the ball height, yield, 2D and 3D feature value parameters.

3. The method for automated processing of AOI inspection data in the semiconductor industry according to claim 1, characterized in that, In S2, the construction of the device feature knowledge base includes: using the Measurement Data tag structure of the Semiconductor Device Manufacturing Format 3.0 protocol and the field encoding of the Camtek Transmission Format 2.0 protocol to parse the results dynamically through a locally deployed online learning algorithm.

4. The method for automated processing of AOI inspection data in the semiconductor industry according to claim 1, characterized in that, S2 specifically includes: S21: Use the predefined regular expression template library to parse the raw AOI data; S22: Use three sigma statistical methods and long short-term memory network models to dynamically clean outliers and output data in a standardized format.

5. The method for automated processing of AOI inspection data in the semiconductor industry according to claim 4, characterized in that, In step S21, the regular expression template library includes: predefined templates for parsing KLA semiconductor device manufacturing format files.

6. The method for automated processing of AOI inspection data in the semiconductor industry according to claim 4, characterized in that, In S22, the three sigma statistical cleaning includes: calculating the mean and standard deviation of historical data, filtering outliers that exceed the range of mean minus three standard deviations and mean plus three standard deviations, and predicting process fluctuations using an LSTM model in order to dynamically adjust the cleaning threshold.

7. The method for automated processing of AOI inspection data in the semiconductor industry according to claim 4, characterized in that, S3 specifically includes: S31: Configure the mapping relationship between detection parameters and statistical process control indicators through a visual interface, and generate a mapping table in Extensible Markup Language format; S32: The real-time monitoring module transmits real-time parameters to the statistical process control system via the SECS / GEM protocol, and transmits statistical parameters to the historical database of the statistical process control system via the MQTT protocol.

8. The method for automated processing of AOI inspection data in the semiconductor industry according to claim 7, characterized in that, In step S32, when transmitting solder ball height, surface height, or three-dimensional feature value parameters, the transmission delay is no greater than 200ms; when querying yield or two-dimensional feature value parameters, the query response time is no greater than 3s.

9. The method for automated processing of AOI inspection data in the semiconductor industry according to claim 1, characterized in that, S4 specifically includes: S41: Call the algorithm library of the statistical process control system to calculate the process capability index value and draw the control chart; S42: Generate a dynamic report that includes the average height curve of the spheres and a heatmap of yield fluctuations; S43: When the process capability index value is less than the preset value, a red warning is marked, a portable document format report containing abnormal parameter information, impact range and process adjustment suggestions is generated, and the report is pushed through the office automation system.

10. The system upon which the method for automated processing of AOI inspection data in the semiconductor industry according to any one of claims 1-9 is based, characterized in that, The system includes a multi-source data acquisition module, an intelligent format conversion module, a dynamic routing module, and an intelligent reporting module, wherein: Multi-source data acquisition module: Based on the device feature knowledge base, it sends device query commands to the AOI device using the SECS / GEM protocol adapter, matches the device feature knowledge base with the AOI device feature information and loads the corresponding parsing engine to extract measurement data from the AOI device; The intelligent format conversion module is used to parse the measurement data extracted using a predefined regular expression template library to obtain parsed data; it combines the three sigma statistical method and the LSTM model to clean outliers from the parsed data and generate data in a standardized format. Dynamic routing module: Used to transmit the generated standardized data to the statistical process control system through a dynamic routing mechanism based on the real-time requirements of the detection parameters. Intelligent Reporting Module: This module integrates statistical process control algorithms to calculate standardized data in statistical process control systems, generating multi-dimensional inspection reports that include process capability index values. It triggers anomaly warnings when the process capability index value falls below a set threshold, thus automating the processing of AOI inspection data in the semiconductor industry.

Images