Flexible circuit board defect tracing method and system based on time sequence knowledge graph
Through the method based on the timing knowledge graph, the problems of artificial dependence and low efficiency in the traceability of flexible circuit board defects are solved, automated traceability and intelligent decision-making support are realized, traceability accuracy and efficiency are improved, and diversified needs are adapted to.
Patent Information
- Application Number
- CN202411009985.4
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-07-26
- Publication Date
- 2024-11-15
AI Technical Summary
The existing flexible circuit board defect traceability methods rely on manual inspection efficiency, low detection accuracy, high cost and difficult to adapt to diversified needs. They cannot effectively manage and integrate unstructured and discrete defect traceability text data, resulting in cumbersome traceability process and Inefficient.
Using a time-series knowledge graph method, entity extraction is performed through BIO annotation and Bert-BiLSTM-CRF model, defect cause knowledge graph is constructed, intelligent question-and-answer module is designed, and the weight probability of event paths is calculated using hierarchical analysis method to realize automated defect traceability and decision support.
It improves the accuracy and efficiency of defect traceability, can more comprehensively analyze the defects of flexible circuit boards, reduce rework and scrap rates, reduce production costs, adapt to the needs of different application scenarios, and provide intelligent decision-making support.
Smart Images

Figure CN118966221A_ABST
Abstract
Description
Technical Field
[0001] The present invention belongs to the field of electronic engineering technology, and in particular relates to a flexible circuit board defect tracing method and system based on a time series knowledge graph. Background Art
[0002] Flexible circuit boards are lightweight, small, thin, and highly reliable flexible circuit boards. They are widely used in defense, aerospace, automotive, medical, electronic equipment, and other fields, and their proportion in circuit boards is increasing. According to Prismark data, by 2023, the global flexible circuit board market will exceed $18 billion, and it is expected to reach $28.7 billion by 2025, indicating that the market has great growth potential. Flexible circuit boards are in wide demand in the current market, mainly benefiting from the following trends and application needs:
[0003] Consumer electronics: Flexible circuit boards are widely used in smartphones, tablet computers, and wearable devices due to their small size, light weight, and flexible design.
[0004] Automotive electronics: The demand for flexible, high-reliability circuit connection solutions in this field is gradually increasing, and flexible circuit boards have become an ideal choice to meet these needs.
[0005] Medical devices and medical equipment: The demand for miniaturized and portable medical equipment in the medical field is increasing, and flexible circuit boards can meet the requirements of these devices for flexible layout and compact space. Therefore, their application in medical devices and medical equipment is also showing a growth trend.
[0006] Industrial control and aerospace fields: Industrial automation equipment, avionics equipment in the aerospace field, etc. have a stable demand for high-reliability, vibration-resistant, and impact-resistant circuit connection solutions, and flexible circuit boards also have a certain market share in these fields.
[0007] The market demand for flexible circuit boards will continue to grow steadily. At the same time, with the continuous emergence of emerging technologies and new applications, flexible circuit boards will further expand their application scope to meet diverse market needs.
[0008] The production process of flexible circuit boards includes dozens of steps such as punching, drilling, copper plating, and developing. It is inevitable that defects such as drilling deviation, pressure pits, and uneven copper thickness will occur, and even cause permanent damage to electronic equipment.
[0009] Therefore, the defect traceability of flexible circuit boards must be carried out in a timely manner after the defects occur. The defect traceability method of flexible circuit boards is to find out the problems in the manufacturing process and trace their causes. The following is an overview of the commonly used flexible circuit board defect traceability methods under general circumstances:
[0010] Surface inspection method: Surface inspection method refers to the inspection and analysis of the surface of flexible circuit boards to determine the defects and trace them. For example, machine vision technology can quickly detect scratches, cracks, deformations and other problems on the surface of circuit boards; optical microscopes can be used to observe and analyze surface defects of circuit boards.
[0011] Microstructure analysis method: Microstructure analysis method refers to the method of determining the source and nature of defects by analyzing the microstructure of the material. For example, scanning electron microscopy technology can analyze the microstructure of the circuit board to reveal the source of defects and provide accurate elimination solutions.
[0012] Data mining method: Data mining method refers to the analysis and mining of a large amount of production data to identify the existing correlations and regularities. For example, a large amount of production data can be analyzed using machine learning algorithms to find out the sources of defects and provide effective prediction and control measures. These methods play an important role in the traceability of flexible circuit board defects, but they have disadvantages such as cumbersome steps, low efficiency, and high labor demand. Therefore, there is an urgent need for a flexible circuit board defect traceability model framework that can effectively avoid the above shortcomings, help manufacturers and designers analyze product defects more comprehensively, and thus improve product quality and reliability. This is particularly important.
[0013] The development of equipment defect tracing technology is based on practical application needs. It has undergone a transformation from relying on on-site observation and experience of staff to locate the process where defects occur and provide maintenance plans, to identifying machine equipment defects based on machine information. At present, most methods based on mathematical models are used to trace the defects of industrial equipment.
[0014] (1) Defect tracing method based on mathematical model
[0015] The defect traceability method based on mathematical models is to use mathematical models to model the expected behavior of equipment or systems, compare the actual observed data with the model prediction results, and identify defects or abnormal behaviors in equipment or systems by analyzing the differences between them. Li Lijuan et al. (2010) used Bayesian networks to improve the interpretation and reasoning ability of defect cause tracing methods for uncertain information. Song Zhihuan et al. (2019) proposed a defect traceability method based on distributed Bayesian networks and applied it in Baosteel's high-speed wire production line. Chen Xinyu et al. (2022) proposed a defect traceability method based on Bayesian networks that does not rely on prior knowledge to solve the problem of mismatch between the amount of steel defect samples and normal samples. Some scholars combine decision trees and case-based reasoning (CBR) to complete quality traceability. Selvamani et al. (2005) proposed a method based on case reasoning and cause induction in discrimination tree (CID) to analyze the causes of quality defects in the steel strip manufacturing field. However, this work is based on an existing case library and cannot cope with the situation where there are no similar cases in the case library. Lyu et al. (2020) proposed a solution using decision trees and association rules to trace defects, reducing the product defect rate from 20% to 5%. Liu Huan et al. (2007) designed a continuous casting billet quality traceability analysis system to find the causes of billet defects. Feng Xiaolei et al. (2020) designed a blockchain-based steel quality traceability system for the construction of a steel quality system. Wang et al. (2020) proposed a fault tracing model that integrates historical cases with process data information to achieve fault location of hot rolling lines with surface defects of strip steel. For historical cases, the defect cause labels are calculated by text similarity, and the probability distribution of defect causes is obtained by fuzzy semantic reasoning. For process data, the L1 regularization method is used for feature selection, and the XGBoost (2016) integration method is used to train the association model between process data and defects to determine the contribution of each feature in the data source. Finally, based on the DS evidence theory (2008), different rules are set to combine the two judgments to determine the probability of each fault source in the hot rolling production line. The model is then applied to the real steel production environment, which can effectively assist expert decision-making and improve the efficiency of hot rolling production line fault tracing.
[0016] (2) Knowledge-based defect tracing method
[0017] The knowledge-based defect traceability method uses knowledge engineering and other disciplines to complete the defect traceability task, among which the most representative is the knowledge graph. Liu Di (2023) proposed a steel quality defect traceability method based on data mining and knowledge graphs. He built a steel quality traceability knowledge graph based on the Neo4j graph database, mined potential laws and patterns from a large amount of data, and summarized the causes and rules of quality defects in steel products to make the traceability process more reliable and explainable. Compared with traditional defect traceability methods, knowledge graphs can make full use of a large number of flexible circuit board defect traceability cases accumulated over the years, so that the flexible circuit board defect traceability information exists in the same knowledge graph, and the flexible circuit board defect traceability knowledge is effectively managed through the graph database, and the complex hidden relationship between the traceability knowledge is intuitive, so it can effectively improve the efficiency of defect traceability. At present, the research on flexible circuit board defect traceability based on knowledge graphs is still in its initial stage, which is a direction that needs to be studied urgently.
[0018] Through the above analysis, the problems and defects of the prior art are as follows:
[0019] (1) Reliance on manual inspection: Many existing methods still rely on manual inspection to identify defects in flexible circuit boards, which leads to inefficient, costly and inconsistent results.
[0020] (2) Limitations: Some methods can only detect specific types of defects, such as open circuits, short circuits, etc., but are not sensitive enough or cannot detect other types of defects.
[0021] (3) Low detection accuracy: Some methods have false positives or false negatives when detecting defects in flexible circuit boards, which affects their reliability and accuracy.
[0022] (4) High cost: Some high-precision and high-efficiency detection methods require expensive equipment or technical support, which increases production costs.
[0023] (5) Poor real-time performance: Some methods require samples to be sent to the laboratory for analysis, resulting in poor real-time performance of defect detection results, which is not conducive to timely adjustment of the production process.
[0024] (6) Difficult to adapt to diverse needs: With the continuous expansion of the application areas of flexible circuit boards and the diversification of needs, existing methods cannot flexibly adapt to the needs of different application scenarios. Summary of the invention
[0025] In view of the problems existing in the prior art, the present invention provides a method and system for tracing the defects of flexible circuit boards based on a time-series knowledge graph. In order to solve the problems that the unstructured and discrete text data of flexible circuit board defect tracing cannot be effectively managed and integrated, and the flexible circuit board defect tracing process is cumbersome and inefficient, the present invention, based on the research of predecessors, designs a flexible circuit board defect tracing model framework based on a time-series knowledge graph, simplifies the defect tracing process, and improves the tracing accuracy.
[0026] The present invention is implemented as follows: a method for tracing the defect of a flexible circuit board based on a time series knowledge graph, comprising:
[0027] Step 1: Use the BIO annotation method to annotate the text data of the process defect relationship of the flexible circuit board, and extract entities based on the Bert-BiLSTM-CRF model;
[0028] Step 2: Use the entity relationship data extracted by the named entity recognition method to construct a knowledge graph of defect causes;
[0029] Step 3: Extracting timing information from defect traceability record documents to form a timing knowledge graph of the causes of defects in flexible circuit boards;
[0030] Step 4: Design a defect traceability model framework based on the time series knowledge graph, use the hierarchical analysis method to assign weights to each process parameter, calculate the weight probability of the event path, and verify the effectiveness of the model framework;
[0031] Step 5: Design an intelligent question-answering module based on the defect cause time series knowledge graph, intent recognition, and slot filling;
[0032] Step 6: Establish a general knowledge graph integration platform and integrate the intelligent question and answer module.
[0033] Further, the step 1 comprises:
[0034] The text data was sorted based on the relationship between the process defects of the flexible circuit board, and the data set was annotated using the BIO annotation method, which was divided into five entity types: process, process work content, process parameters, defects caused by the process, and equipment used in the process.
[0035] The entity extraction method based on the Bert-BiLSTM-CRF model is applied to the defect tracing of flexible circuit boards, and the results of entity extraction are stored as csv files as the basis for subsequent knowledge graph visualization.
[0036] Further, the step 2 includes:
[0037] According to the flexible circuit board defect traceability knowledge extracted by named entity recognition method, a defect cause knowledge graph was constructed using a top-down approach for the defect traceability field. The defect cause ontology model in the pattern layer was independently designed, and the flexible circuit board defect cause knowledge data in the data layer was filled. Finally, the csv file was imported into the Neo4j graph database and the entity relationship was manually established according to the ontology to complete the storage and visualization of the flexible circuit board defect cause knowledge graph.
[0038] Further, the step three includes:
[0039] The above-constructed knowledge graph is updated using the flexible circuit board defect traceability record document data, and the timing information is manually extracted to form a flexible circuit board defect cause timing knowledge graph to provide a knowledge background for defect tracing.
[0040] Further, the step 4 includes:
[0041] A defect traceability model framework for flexible circuit boards based on a time-series knowledge graph is designed to find the historical traceability records of defects and the related event paths that lead to the defects, and to count the number of traceability locations. Then, the weight probability of each event path is quantitatively analyzed by assigning weights to each process parameter using the analytic hierarchy process (AHP). Combined with the above statistical times, the effectiveness of the AHP is proved, the traceability results are visualized, and the defect traceability records and results are updated to the knowledge graph. A traceability result evaluation method is designed to verify the effectiveness of the defect traceability model framework.
[0042] Furthermore, the step six of constructing the knowledge graph integrated platform includes manual construction and automatic construction methods.
[0043] Another object of the present invention is to provide a flexible circuit board defect tracing system based on a time series knowledge graph, comprising:
[0044] The entity extraction module is used to annotate the text data of the process defect relationship of the flexible circuit board using the BIO annotation method, and to extract entities based on the Bert-BiLSTM-CRF model;
[0045] The knowledge graph construction module is used to construct a defect cause knowledge graph using entity relationship data extracted by the named entity recognition method, and extract timing information from the defect traceability record document to form a flexible circuit board defect cause timing knowledge graph;
[0046] The defect traceability model module is used to design a defect traceability model framework based on the time series knowledge graph, use the hierarchical analysis method to assign weights to each process parameter, calculate the weight probability of the event path, and verify the effectiveness of the model framework;
[0047] Intelligent question-answering module, which is used to design intelligent question-answering functions based on the defect cause time series knowledge graph, intent recognition, and slot filling;
[0048] The knowledge graph integrated construction platform is used to integrate the above modules and provide methods for manually and automatically constructing knowledge graphs.
[0049] Furthermore, it also includes:
[0050] The data annotation module is used to perform BIO annotation on the text data of the process defect relationship of the flexible circuit board, which is divided into five entity types: process, process work content, process parameters, defects caused by the process, and equipment used in the process;
[0051] The entity relationship establishment module is used to import the csv file into the Neo4j graph database and manually establish the entity relationship according to the ontology, so as to complete the storage and visualization of the knowledge graph of the causes of defects in flexible circuit boards;
[0052] The timing information extraction module is used to manually extract timing information from the flexible circuit board defect traceability record document data and update the knowledge graph;
[0053] The traceability result visualization module is used to visualize the traceability results and update the defect traceability records and results into the knowledge graph.
[0054] Another object of the present invention is to provide a computer device, comprising a memory and a processor, wherein the memory stores a computer program, and when the computer program is executed by the processor, the processor executes the steps of the flexible circuit board defect tracing method based on the time series knowledge graph.
[0055] Another object of the present invention is to provide a computer-readable storage medium storing a computer program, which, when executed by a processor, enables the processor to perform the steps of a flexible circuit board defect tracing method based on a time series knowledge graph.
[0056] In combination with the above technical solutions and the technical problems solved, the advantages and positive effects of the technical solutions to be protected by the present invention are as follows:
[0057] First, the problems existing in the prior art solved by the present invention are:
[0058] Difficulty in tracing defects in flexible printed circuits (FPCs): Traditional FPC defect tracing mainly relies on manual experience and simple statistical analysis, lacking systematic and intelligent analysis methods, resulting in a time-consuming and inefficient tracing process, and the accuracy of the tracing results is difficult to guarantee.
[0059] Insufficient knowledge representation of defect causes: The existing technology does not represent the knowledge of FPC defect causes comprehensively and structuredly enough, making it difficult to effectively use domain knowledge to assist defect tracing.
[0060] Timing information is not fully utilized: Timing information in the FPC production process is of great significance for defect tracing, but existing technologies fail to fully utilize timing information to analyze the evolution process of defects.
[0061] Lack of intelligent question-and-answer support: The existing technology lacks an intelligent question-and-answer system for FPC defect tracing, and cannot provide engineers with convenient knowledge query and auxiliary decision-making support.
[0062] Detailed and specific technological progress is reflected in the following aspects:
[0063] Building a knowledge graph: By building a knowledge graph of the causes of flexible circuit board defects, a structured knowledge base can be created to organize and represent key factors, process flows, and other relevant information related to flexible circuit board production. Various data and knowledge can be linked to each other to provide a more comprehensive and scalable information resource. Using this resource, the defect traceability of flexible circuit boards can be more effectively carried out to further improve production processes and quality control measures.
[0064] The present invention integrates multi-source information: In later research, the time-series knowledge graph can integrate information from multiple data sources in the flexible circuit board production process, including sensor data, production process parameters, equipment status, etc., to comprehensively and multi-angle describe the production and quality of flexible circuit boards.
[0065] Dynamic modeling capabilities: The time series knowledge graph can capture relationships and patterns that change over time, model the dynamic changes in the flexible circuit board production process, update the graph in real time, and reflect the time series information of defect occurrence.
[0066] Multi-factor analysis: Through the time series knowledge graph, the changes of various factors in the time series can be taken into account, and the time series relationship between different factors can be analyzed, including process parameters, material properties, equipment status and other factors, to help determine the causes of flexible circuit board defects.
[0067] Path inference capability: The time series knowledge graph can automatically discover and infer potential defect traceability paths through graph algorithms and path inference technology. This path inference capability makes the traceability results more interpretable and explainable, helping engineers to deeply understand the causes of defects. However, the path inference of Bayesian networks usually relies on known prior probabilities and conditional probabilities, which are affected by the network structure and parameter settings, resulting in the inability to fully discover potential defect traceability paths.
[0068] Dealing with uncertainty: Time series knowledge graphs can effectively deal with data uncertainty and noise interference. By analyzing historical data and trends, they can identify truly influential factors and reduce the rate of misjudgment. In contrast, Bayesian networks have higher requirements for data quality and prior knowledge, are more susceptible to noise interference, and have difficulty accurately identifying the true defect tracing factors.
[0069] Decision support: The time series knowledge graph can not only be used for defect tracing, but also provide support for production decisions, such as providing references for optimizing production processes and improving production efficiency.
[0070] To sum up, the advantages of the flexible circuit board defect tracing method based on time series knowledge graph are reflected in the construction of knowledge graph, integration of multi-source information, dynamic modeling capabilities, multi-factor analysis, path inference capabilities, uncertainty processing, and decision support. It helps to trace the defects of flexible circuit boards more comprehensively and accurately, and helps companies improve production processes and improve product quality.
[0071] Second, the expected benefits and commercial value of the technical solution of the present invention after transformation are:
[0072] The present invention provides an innovative method for tracing the defects of flexible circuit boards based on a time-series knowledge graph, and this technical solution has significant practical application value and broad commercial prospects. By accurately tracking and tracing various defects in the production process, this method can significantly improve the accuracy and reliability of defect tracing of flexible circuit boards and related electronic products, and help manufacturers and designers to more comprehensively analyze product defects, thereby improving product quality and reliability. In addition, the implementation of this technology can effectively reduce the rework and scrap rate caused by production defects, thereby directly reducing production costs and enhancing the market competitiveness of products. Furthermore, the method optimizes the fault response process through in-depth analysis of production data, significantly shortens the time for problem diagnosis and resolution, and greatly improves the overall production efficiency. The application of this technology is not limited to the production of flexible circuit boards, and its principle is also applicable to the manufacture of other high-precision electronic components, providing a highly scalable solution. The data-driven decision support system of the present invention provides a powerful analysis tool for enterprise management to help them better understand and control the production process and optimize product design and production processes. By implementing this technology, enterprises will be able to produce more reliable products with higher efficiency and lower costs, thereby gaining a clear advantage in the fierce market competition.
[0073] The technical solution of the present invention fills the technical gap in the industry at home and abroad: The technical solution of the present invention effectively fills the technical gap in this field at home and abroad by introducing a flexible circuit board defect tracing method based on time series knowledge graph. The core innovation of this technology lies in its ability to form a new dynamic defect tracking and analysis method by integrating time series data analysis and knowledge graph technology. This method can trace the development history and causes of defects, thereby achieving the prevention and control of product quality problems from the source.
[0074] In the current manufacturing industry, although there are a variety of fault diagnosis and quality control technologies, most of them focus on data analysis at a single time point and fail to fully utilize the time series data generated during the entire production process. This invention can reveal the inherent mechanism and influencing factors of defect generation by conducting an in-depth analysis of the time series data in the production process and combining it with the construction of a knowledge graph.
[0075] Time series data analysis can capture and analyze changes in data over time, which is particularly critical for the highly dynamic and complex environment of electronic manufacturing. This technology enables real-time monitoring and prediction of equipment performance and its failures from a continuous production process, greatly improving the accuracy and timeliness of failure prediction. Through in-depth analysis of time series data, the present invention can identify abnormal patterns and trends in the production process, so as to intervene and correct problems before they become significant quality defects.
[0076] The construction of the knowledge graph also plays a vital role in the present invention. By organizing complex production data and process information into an interconnected graph database, the knowledge graph not only improves the visualization of the data, but also enhances the understanding of the correlation between the data. This structured data representation method supports complex queries and analysis, allowing companies to understand the defect generation mechanism in the production process at a deeper level. In addition, the knowledge graph also supports the automation of the decision-making process, and helps companies formulate more accurate and effective improvement measures by providing evidence-based reasoning.
[0077] In addition, the implementation of this technical solution is not limited to the production of flexible circuit boards. Its theory and method are also applicable to a wider range of electronic manufacturing fields, including but not limited to the production of semiconductors, integrated circuit boards and other high-precision components. Through intelligent data integration and automated analysis, the present invention promotes the systematization and automation of defect management in the production process, and realizes real-time monitoring and rapid response to production line defects. BRIEF DESCRIPTION OF THE DRAWINGS
[0078] Figure 1 It is a flow chart of a method for tracing the defect of a flexible circuit board based on a time series knowledge graph provided by an embodiment of the present invention;
[0079] Figure 2 The flexible circuit board defect tracing knowledge graph construction process provided by the embodiment of the present invention;
[0080] Figure 3 It is an ontology model diagram of the flexible circuit board defect tracing field provided by an embodiment of the present invention;
[0081] Figure 4 It is the BiLSTM model structure provided by the embodiment of the present invention;
[0082] Figure 5 It is the LSTM unit network structure provided by the embodiment of the present invention;
[0083] Figure 6 This is the Bert-BiLSTM-CRF model structure provided by the embodiment of the present invention;
[0084] Figure 7 This is the named entity recognition experimental process provided by the embodiment of the present invention;
[0085] Figure 8 It is a knowledge graph for tracing defects of flexible circuit boards provided by an embodiment of the present invention;
[0086] Fig. 9 It is a partial flexible circuit board defect tracing knowledge graph provided by an embodiment of the present invention;
[0087] Fig.10 It is a time series knowledge graph for tracing the source of unclean development defects provided by an embodiment of the present invention;
[0088] Fig.11 These are all the paths that lead to the text falling-off defect provided by the embodiment of the present invention;
[0089] Fig.12 It is a historical traceability record of text loss defects provided by an embodiment of the present invention;
[0090] Fig.13 This is an example of a historical traceability record of a text loss defect provided by an embodiment of the present invention;
[0091] Fig.14 This is an example of a historical traceability record of a text loss defect provided by an embodiment of the present invention;
[0092] Fig.15 This is the text falling-off defect tracing path (a) provided by the embodiment of the present invention;
[0093] Fig.16 This is the text falling-off defect tracing path (b) provided by the embodiment of the present invention;
[0094] Fig.17 This is the text falling-off defect tracing path (c) provided by the embodiment of the present invention;
[0095] Fig.18 This is an updated text loss defect tracing path provided by an embodiment of the present invention;
[0096] Fig.19 This is the intelligent question-answering process provided by an embodiment of the present invention;
[0097] Fig. 20 is a question and answer screenshot (1) provided in an embodiment of the present invention;
[0098] Fig.21 is a question and answer screenshot (2) provided in an embodiment of the present invention;
[0099] Fig. 22 It is a diagram of the defect tracing model architecture provided by an embodiment of the present invention. DETAILED DESCRIPTION
[0100] In order to make the purpose, technical solution and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with the embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not used to limit the present invention.
[0101] like Figure 1 The figure shows a flow chart of a method for tracing the defect of a flexible circuit board based on a time series knowledge graph provided by an embodiment of the present invention, which specifically includes:
[0102] Step 1: BIO text sequence annotation
[0103] The present invention uses the relationship between process defects of flexible circuit boards to organize text data, and uses BIO sequence annotation. The entity starting position is marked with "B-entity name", and the position after the starting position is marked with "I-entity name", and the rest of the irrelevant text is marked with "O". The present invention needs to mark 5 entity categories including process, work content, equipment, defects, and process parameters, 11 label types, and the entity label definition and corresponding label description are shown in the table.
[0104] Table Flexible PCB defect traceability text labels and instructions
[0105]
[0106] Step 2: Entity Extraction
[0107] Entity extraction in the text data of flexible circuit board process defect relationship arrangement is a named entity recognition task in a specific field. The present invention uses the Bert-BiLSTM-CRF model for entity extraction. In addition, in order to highlight the effectiveness of Bert-BiLSTM-CRF in extracting entities using text data of flexible circuit board process defect relationship arrangement, a comparative experiment was conducted using the Word2Vec-BiLSTM-CRF model and the Bert-CRF model. The Bert-BiLSTM-CRF model can directly use text character vectors as input, and no longer requires manual definition of features. BiLSTM is a bidirectional long short-term memory neural network, which can process words that have not been seen in the training set by using its contextual association characteristics. CRF is a conditional random field probability model that can learn the contextual association of the output sequence of the BiLSTM model and improve the recognition accuracy. Experiments have shown that the Bert-BiLSTM-CRF model has the highest accuracy in named entity recognition of the text of flexible circuit board process defect relationship arrangement.
[0108] Entity extraction model architecture:
[0109] The Bert-BiLSTM-CRF model contains a Bert layer, a Bi-LSTM layer, and a CRF layer. The structure of Bert includes multiple layers of Transformer Encoder. The model is divided into two main parts: the Embedding layer, which includes word embedding, position embedding, and sentence embedding; the Transformer Encoder layer, which processes the text sequence and generates the final representation. Word2Vec is a natural language processing technology that aims to convert words in a vocabulary into vector representations. The main advantage of Word2Vec is that it can capture the semantic relationship between words, which makes Word2Vec widely used in natural language processing tasks such as word clustering, sentiment analysis, and semantic similarity calculation. BiLSTM consists of two LSTMs in different directions, and its structure is as follows: Figure 4 shown.
[0110] Each LSTM in the BiLSTM model consists of three gated units: input gate, forget gate, and output gate. The input gate controls the input of the network; the forget gate determines which knowledge needs to be forgotten and is the core of the LSTM unit; the output gate controls the output of the network. The LSTM network structure is as follows: Figure 5 shown.
[0111] During the recognition process, the two LSTMs recognize the sentences from front to back and from back to front respectively, output the label prediction values in the two directions respectively, and splice the outputs in the two directions to obtain the final result. Although BiLSTM takes into account the influence of contextual content on the probability of each word label, it does not take into account the correlation between the previous and next labels. For example, in the BIO sequence annotation used in the present invention, if the correct annotation sequence of a sentence is "O B- process I- process I- process OO", but the prediction result of the BiLSTM model is "O I- process I- process I- process OO", obviously, this prediction result does not conform to the rules. The label of the first word of each word should start with "B-", which shows that BiLSTM did not consider the correlation between the previous and next labels during the prediction process. At this time, adding a conditional random field model (CRF) after the output layer of BiLSTM can solve this problem, so the BiLSTM-CRF model can obtain better prediction results. Therefore, the present invention adopts the Bert-BiLSTM-CRF model for entity extraction, and the model structure is as follows: Figure 6 As shown. Based on the above-mentioned flexible circuit board defect process defect relationship, the text data is sorted to train the model.
[0112] The named entity recognition process of the flexible circuit board process defect relationship collation text data is as follows Figure 2 shown.
[0113] Step 3: Storage and visualization of knowledge graph of causes of defects in flexible circuit boards
[0114] This paper selects Neo4j as the graph database to store the knowledge graph, imports the csv file saved in the previous section into Neo4j, and some Cypher statements are shown in the table:
[0115] Import table CSV file into Neo4j statement
[0116]
[0117] After importing the csv file to form entities, the relationships between them are manually added according to the relationship definition of the ontology layer to finally form a knowledge graph.
[0118] The knowledge graph data layer of the causes of defects in flexible circuit boards includes processes such as cutting, drilling, electroplating, dry film, equipment such as cutting machines, die-cutting machines, and laminating machines, defects such as drilling deviation, dents, and ink shedding, and process parameters such as cover film materials, number of layers, development time, and curing temperature.
[0119] Step 4: Construction of the time series knowledge graph of the causes of defects in flexible circuit boards
[0120] The above-constructed knowledge graph is updated using the document data of flexible circuit board defect traceability records to form a time series knowledge graph of the causes of flexible circuit board defects. Based on the above data, three types of data are mainly sorted out: defects, process traceability records, and parameter traceability records. Defects mainly include flexible circuit board defects such as offset, ink shedding, text shedding, and cover film overlap; process traceability records mainly describe the process to which the defect is traced, and conduct process-level traceability; parameter traceability records are used to describe which parameter is incorrectly caused by the defect, and conduct process parameter-level traceability. The above data are processed to form triplets of <defect, traceability record, process traceability record> and <defect, traceability record, parameter traceability record>. Taking into account the existence of time information in the traceability record of the defect and the timing of the manufacturing process of the flexible circuit board, the present invention innovatively adds time information to the defect cause knowledge graph to form a defect cause timing knowledge graph, which is composed of multiple quadruplets of the form of <entity, relationship, entity, time>, for example, <defect, traceability record, process traceability record, time> and <defect, traceability record, parameter traceability record, time>.
[0121] The advantages of the defect tracing method based on the time series knowledge graph compared with other methods are shown in the table and described as follows:
[0122] Table Summary of the advantages of defect tracing based on time series knowledge graph
[0123]
[0124] The present invention proposes a flexible circuit board defect tracing model framework based on a temporal knowledge graph, comprising the following steps:
[0125] This experiment is based on python 3.7 and Neo4j 3.5.35. It uses Python to interact with the Neo4j database, executes Cypher query statements to query and operate data on the Neo4j graph database, and outputs the corresponding results for subsequent analysis and visualization.
[0126] By taking the names of defects generated in the production of flexible circuit boards as input, the defect traceability model can analyze the causes of the defects. First, the historical traceability results are queried through the defect cause time series knowledge graph. The query statement is shown in the table.
[0127] Table history tracing record query statement
[0128]
[0129] To find each causal path of the defect, the Cypher statements used for query are shown in the table. These query statements provide an effective method to analyze and explore the relationship path of flexible circuit board defect tracing.
[0130] Table defect occurrence path query statement
[0131]
[0132] Finally, the analytic hierarchy process is used to quantitatively analyze the weight probability of each causal path and visualize it, and combined with the historical traceability results, the event path with the highest probability of causing defects is recommended. This provides flexible circuit board manufacturers with stronger defect traceability capabilities, enabling them to discover and solve quality problems in a timely manner and achieve more sustainable development in the flexible circuit board industry.
[0133] Through the query method of Cypher query statement, all the path information leading to the defect can be quickly obtained. Now, the defect tracing model framework proposed in this paper is used to conduct a text detachment defect tracing experiment. When the defect type of the input flexible circuit board is text detachment, all the paths leading to the defect can be obtained, including font size, text content, curing temperature, and curing time. Fig.11 shown.
[0134] In addition, the historical traceability records of the defect were obtained, such as Fig.12 shown.
[0135] The traceability record of a certain text loss defect is as follows: Fig.13 and Fig.14 As shown, the source of the text falling off defect was located in the font size process parameters of the silk-screen text process, because the font size was not selected appropriately.
[0136] There are a total of 121 text peeling defect traceability records, and 5 records are listed in equal proportion, of which 4 records are related to the text size process parameters, and the remaining 1 record is related to the text content process parameters, as shown in the table.
[0137] Table text loss defect history traceability record
[0138]
[0139] From the historical traceability records of text peeling defects, we can see that this defect occurs in the silk-screen text (printing-curing) process. In most cases, it is caused by inappropriate text size. In very few cases, it is caused by insufficient clarity and continuity of the text content. Based on this, it is assumed that the text peeling defect is still caused by inappropriate text size. The hierarchical analysis method is used to verify this result.
[0140] exist Fig.11There are four causal paths that cause ink shedding defects, which occur in the silk-screen text (printing text-curing) process. The process parameters that cause this defect are font size, text content, curing time, and curing temperature. Now we use the hierarchical analysis method to calculate the weighted probabilities of the four process parameters. Finally, we sort the text shedding defect traceability paths according to the weighted probabilities, and visualize the top three probable paths.
[0141] As shown in the table, a judgment matrix is formed after scoring through expert experience. AD represents the four process parameters of font size, text content, curing time, and curing temperature. This judgment matrix has 4 rows of data. The meaning of the first row of data is explained. The first row represents that text size is slightly more important than text content, slightly more important than curing time selection, and more important than curing temperature.
[0142] Table Analytical Hierarchy Process Judgment Matrix
[0143]
[0144]
[0145] According to the judgment matrix of text falling defect traceability, the maximum eigenvalue is calculated to be 4.07, and the corresponding eigenvector is [0.708026930.267266540.626802690.14160564]. The index of the inconsistency of the judgment matrix calculated by the maximum eigenvalue is 0.023. The average random consistency index value in the table is read, and the consistency ratio of the judgment matrix is calculated to be 0.033, which is less than 0.1, that is, the judgment matrix has satisfactory consistency at this time. The weighted probabilities of the four process parameters are calculated by the maximum eigenvalue and eigenvector, which are 0.40604817, 0.35946668, 0.15327537, and 0.08120978, respectively. That is, the font size is the main reason for the text falling defect in the silk-screen text (printing text-curing) process of the flexible circuit board. The second and third process parameters are text content and curing time, respectively, and the curing temperature has the lowest probability. Fig.15 , 16 , 17 respectively visualize the paths of text-falling-off defects caused by these three process parameters.
[0146] The experimental verification shows that the tracing results of the text peeling defect are the same as the historical tracing records, both of which occurred in the silk-screen text (printing-curing) process, and the reason is that the text size is not appropriate.
[0147] After the defect is successfully traced, the probability of each process parameter that causes the text falling off defect is updated to the knowledge graph of the causes of the defect of the flexible circuit board. The final knowledge graph of the causes of the text falling off defect is as follows: Fig.18 shown.
[0148] The method for tracing the defect of a flexible circuit board based on a time series knowledge graph provided by an embodiment of the present invention comprises the following steps:
[0149] 1. Use the BIO annotation method to annotate the text data of the process defect relationship of the flexible circuit board
[0150] The BIO (Begin, Inside, Outside) annotation method is used to annotate the sorted flexible circuit board process defect relationship text data.
[0151] Entity extraction is performed based on the Bert-BiLSTM-CRF model to extract relevant entities from the text.
[0152] 2. Use the entity relationship data extracted by the named entity recognition method to build a knowledge graph of defect causes
[0153] The knowledge graph of the causes of defects in flexible circuit boards is constructed using the entity relationship data extracted from the named entity recognition (NER) method.
[0154] 3. Extract timing information from defect traceability record documents to form a knowledge graph of the timing of the causes of defects in flexible circuit boards
[0155] Extract timing information from the defect traceability record document, and combine this information with the previously constructed knowledge graph to form a timing knowledge graph of the causes of flexible circuit board defects containing time information.
[0156] 4. Design a defect tracing model framework based on temporal knowledge graph
[0157] The analytic hierarchy process (AHP) was used to assign weights to each process parameter.
[0158] Calculate the weighted probability of the event path and verify the effectiveness of the model framework.
[0159] 5. Design an intelligent question-answering module based on the defect cause time series knowledge graph, intent recognition, and slot filling
[0160] Based on the defect cause time series knowledge graph, intent recognition and slot filling technology, an intelligent question-answering module is designed to provide accurate defect tracing and answering services.
[0161] 6. Establish a general knowledge graph integration platform and integrate intelligent question-answering modules
[0162] Establish a general knowledge graph integration construction platform and integrate the intelligent question and answer module to achieve all-round traceability and intelligent question and answer functions for flexible circuit board defects.
[0163] Application Example 1: Quality Control in Electronic Manufacturing
[0164] In the electronics manufacturing industry, flexible circuit boards are widely used in various devices. Through this method, various process defects that occur during the production process can be accurately traced to find the root cause of the defects, thereby improving the quality control level of the product. The following are the specific implementation steps:
[0165] Data collection and annotation: BIO annotation of process data and defect records recorded during the production process.
[0166] Knowledge graph construction: Use the extracted entities and relationships to build a knowledge graph of defect causes.
[0167] Time series information integration: Extract time series information from defect records to form a time series knowledge graph.
[0168] Defect tracing model: Design a tracing model, calculate the weighted probability of the event path, and verify the effectiveness of the model.
[0169] Intelligent question and answer module: provides intelligent question and answer function based on knowledge graph and intent recognition.
[0170] Platform integration: Establish an integrated knowledge graph construction platform and integrate the intelligent question and answer module.
[0171] Through the above steps, defects in the production process can be effectively identified and prevented, and the product qualification rate and reliability can be improved.
[0172] Application Example 2: Optimization of Flexible Circuit Board Manufacturing Process
[0173] The manufacturing process of flexible circuit boards is complex and involves a variety of process parameters. Through this method, the impact of different process parameters on defects can be systematically analyzed and optimized. The specific implementation steps are as follows:
[0174] Data collection and annotation: Collect process data and defect records in the flexible circuit board manufacturing process and perform BIO annotation.
[0175] Knowledge graph construction: Use the entity and relationship data extracted by the NER method to build a process defect knowledge graph.
[0176] Timing information integration: Combine the timing information in the process with the defect knowledge graph to form a timing knowledge graph.
[0177] Traceability model design: Use the hierarchical analysis method to assign weights to each process parameter, calculate the weight probability of the event path, and verify the effectiveness of the model.
[0178] Intelligent question and answer module: Design an intelligent question and answer module to provide suggestions for process optimization.
[0179] Platform integration: Establish an integrated knowledge graph construction platform and integrate the intelligent question and answer module.
[0180] Through the above steps, the manufacturing process of flexible circuit boards can be comprehensively analyzed and optimized, the generation of defects can be reduced, and production efficiency and product quality can be improved.
[0181] The above two application embodiments demonstrate the application of the flexible circuit board defect tracing method based on time series knowledge graph in different fields, which helps to improve the intelligence level of the manufacturing process.
[0182] A specific embodiment of the present invention comprises the following steps:
[0183] 1) Data collection: Collect various data on the flexible circuit board production line, including but not limited to raw material batches, machine setting parameters, environmental conditions, and real-time data at each stage of production.
[0184] 2) Knowledge graph construction: Based on the collected data, a knowledge graph is constructed, which includes entities such as materials, equipment, environment and process parameters, as well as the relationships between them. For example, the relationship between raw material batches and initial test results, the relationship between machine parameter adjustments and product defect types, etc.
[0185] 3) Time series data analysis: Use time series data analysis technology to conduct in-depth analysis of data in the production process and identify patterns that lead to defects. For example, by analyzing the relationship between temperature and humidity changes and material behavior, predict the performance of materials under specific conditions.
[0186] 4) Defect prediction and traceability: Combine knowledge graphs and time series data analysis results to achieve defect prediction and traceability. When the system predicts a potential defect, it can quickly track down the specific production links and causes, such as poor raw materials, abnormal equipment parameters, etc.
[0187] 5) Real-time feedback and adjustment: The production system adjusts production parameters or replaces problematic materials in real time based on prediction and traceability results to prevent further defects from occurring.
[0188] It should be noted that the embodiments of the present invention can be implemented by hardware, software, or a combination of software and hardware. The hardware part can be implemented using dedicated logic; the software part can be stored in a memory and executed by an appropriate instruction execution system, such as a microprocessor or dedicated design hardware. It can be understood by a person of ordinary skill in the art that the above-mentioned devices and methods can be implemented using computer executable instructions and / or contained in a processor control code, such as a carrier medium such as a disk, CD or DVD-ROM, a programmable memory such as a read-only memory (firmware), or a data carrier such as an optical or electronic signal carrier. Such code is provided on the carrier medium. The device and its modules of the present invention can be implemented by hardware circuits such as very large-scale integrated circuits or gate arrays, semiconductors such as logic chips, transistors, etc., or programmable hardware devices such as field programmable gate arrays, programmable logic devices, etc., can also be implemented by software executed by various types of processors, and can also be implemented by a combination of the above-mentioned hardware circuits and software, such as firmware.
[0189] The above description is only a specific implementation mode of the present invention, but the protection scope of the present invention is not limited thereto. Any modifications, equivalent substitutions and improvements made by any technician familiar with the technical field within the technical scope disclosed by the present invention and within the spirit and principle of the present invention should be covered by the protection scope of the present invention.
Claims
1. A defect tracing method for flexible circuit boards based on a temporal knowledge graph, characterized in that: include: Step 1: Use the BIO annotation method to annotate the text data of the process defect relationship of the flexible circuit board, and extract entities based on the Bert-BiLSTM-CRF model; Step 2: Use the entity relationship data extracted by the named entity recognition method to construct a knowledge graph of defect causes; Step 3: Extracting timing information from defect traceability record documents to form a timing knowledge graph of the causes of defects in flexible circuit boards; Step 4: Design a defect traceability model framework based on the time series knowledge graph, use the hierarchical analysis method to assign weights to each process parameter, calculate the weight probability of the event path, and verify the effectiveness of the model framework; Step 5: Design an intelligent question-answering module based on the defect cause time series knowledge graph, intent recognition, and slot filling; Step 6: Establish a general knowledge graph integration platform and integrate the intelligent question and answer module.
2. The method for tracing the defect of a flexible circuit board based on a time series knowledge graph according to claim 1, characterized in that: The step one comprises: The text data was sorted based on the relationship between the process defects of the flexible circuit board, and the data set was annotated using the BIO annotation method, which was divided into five entity types: process, process work content, process parameters, defects caused by the process, and equipment used in the process. The entity extraction method based on the Bert-BiLSTM-CRF model is applied to the defect tracing of flexible circuit boards, and the results of entity extraction are stored as csv files as the basis for subsequent knowledge graph visualization.
3. The method for tracing the defect of a flexible circuit board based on a time series knowledge graph according to claim 1, characterized in that: The second step comprises: According to the flexible circuit board defect traceability knowledge extracted by named entity recognition method, a defect cause knowledge graph was constructed using a top-down approach for the defect traceability field. The defect cause ontology model in the pattern layer was independently designed, and the flexible circuit board defect cause knowledge data in the data layer was filled. Finally, the csv file was imported into the Neo4j graph database and the entity relationship was manually established according to the ontology to complete the storage and visualization of the flexible circuit board defect cause knowledge graph.
4. The method for tracing the defect of a flexible circuit board based on a time series knowledge graph according to claim 1, characterized in that: The step three comprises: The above-constructed knowledge graph is updated using the flexible circuit board defect traceability record document data, and the timing information is manually extracted to form a flexible circuit board defect cause timing knowledge graph to provide a knowledge background for defect tracing.
5. The method for tracing the defect of a flexible circuit board based on a time series knowledge graph according to claim 1, characterized in that: The fourth step comprises: A defect traceability model framework for flexible circuit boards based on a time-series knowledge graph is designed to find the historical traceability records of defects and the related event paths that lead to the defects, and to count the number of traceability locations. Then, the weight probability of each event path is quantitatively analyzed by assigning weights to each process parameter using the analytic hierarchy process (AHP). Combined with the above statistical times, the effectiveness of the AHP is proved, the traceability results are visualized, and the defect traceability records and results are updated to the knowledge graph. A traceability result evaluation method is designed to verify the effectiveness of the defect traceability model framework.
6. The method for tracing the defect of a flexible circuit board based on a time series knowledge graph according to claim 1, characterized in that: The step six of building a knowledge graph integrated platform includes manual and automatic building methods.
7. A flexible circuit board defect tracing system based on a time series knowledge graph according to the method of claim 1, characterized in that: include: The entity extraction module is used to annotate the text data of the process defect relationship of the flexible circuit board using the BIO annotation method, and to extract entities based on the Bert-BiLSTM-CRF model; The knowledge graph construction module is used to construct a defect cause knowledge graph using entity relationship data extracted by the named entity recognition method, and extract timing information from the defect traceability record document to form a flexible circuit board defect cause timing knowledge graph; The defect traceability model module is used to design a defect traceability model framework based on the time series knowledge graph, use the hierarchical analysis method to assign weights to each process parameter, calculate the weight probability of the event path, and verify the effectiveness of the model framework; Intelligent question-answering module, which is used to design intelligent question-answering functions based on the defect cause time series knowledge graph, intent recognition, and slot filling; The knowledge graph integrated construction platform is used to integrate the above modules and provide methods for manually and automatically constructing knowledge graphs.
8. The flexible circuit board defect tracing system based on time series knowledge graph according to claim 7 is characterized in that: Also includes: The data annotation module is used to perform BIO annotation on the text data of the process defect relationship of the flexible circuit board, which is divided into five entity types: process, process work content, process parameters, defects caused by the process, and equipment used in the process; The entity relationship establishment module is used to import the csv file into the Neo4j graph database and manually establish the entity relationship according to the ontology, so as to complete the storage and visualization of the knowledge graph of the causes of defects in flexible circuit boards; The timing information extraction module is used to manually extract timing information from the flexible circuit board defect traceability record document data and update the knowledge graph; The traceability result visualization module is used to visualize the traceability results and update the defect traceability records and results into the knowledge graph.
9. A computer device, characterized in that: The computer device includes a memory and a processor, the memory stores a computer program, and when the computer program is executed by the processor, the processor executes the steps of the flexible circuit board defect tracing method based on a time series knowledge graph as claimed in claim 1.
10. A computer-readable storage medium storing a computer program, wherein when the computer program is executed by a processor, the processor executes the steps of the flexible circuit board defect tracing method based on a time series knowledge graph as claimed in claim 1.
Citation Information
Cited By
Garment fabric process tracing method and system based on block chain, medium and equipment
CN119904251A
Nuclear power unit quality defect source analysis and total value chain tracing method
CN120258632A
Pharmaceutical quality traceability decision-making method and system based on dynamic knowledge graph
CN120317535A
Quality defect root cause tracing method based on linkage of standard data and supply chain data
CN120912058A
Product defect tracing method based on knowledge graph reasoning
CN121212916A
