Metal fine wire detection method and system based on end-cloud collaboration and storage medium
The metal micro-wire inspection system, built on an edge-cloud collaborative architecture, integrates sensor groups, rule engines, and machine learning models to achieve efficient and reliable quality inspection and prediction. This solves the problems of long inspection cycles and insufficient security in existing technologies, and improves the consistency and reliability of inspection results.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-12-22
- Publication Date
- 2026-04-07
AI Technical Summary
Existing metal micro-wire testing technologies suffer from problems such as long testing cycles, scattered data, lack of self-learning capabilities, and insufficient security for remote upgrades, making it difficult to achieve efficient and reliable quality testing and prediction.
Adopting an edge-cloud collaborative architecture, the system integrates sensor groups for real-time data acquisition and processing through the collaborative work of detection terminals, edge nodes, and cloud servers. It combines rule engines and machine learning models for evaluation, and introduces large language models and security upgrade mechanisms to achieve intelligent decision-making and quality prediction.
It improves the consistency and reliability of test results, optimizes the test response speed, reduces operation and maintenance risks, enhances quality control capabilities, and supports natural language operation and remote security upgrades.
Smart Images
Figure CN121808623A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of industrial intelligent inspection technology, and in particular to a method, system and storage medium for detecting fine metal wires based on edge-cloud collaboration. Background Technology
[0002] In the production of fine metal wires (such as precision wires for electronics and aerospace), quality inspection is a crucial step. Traditional inspection technologies mainly suffer from the following technical problems: Relying on multiple independent devices to detect parameters such as diameter and conductivity separately, the lack of data collaboration between devices results in long testing cycles and scattered data storage, making it impossible to conduct unified quality trend analysis.
[0003] It relies mainly on manual readings and post-event judgments, lacks the ability to predict quality based on historical data, and is difficult to achieve "prevention in advance".
[0004] Existing systems typically rely on a single fixed threshold (rule) for judgment, which cannot adapt to complex quality fluctuations; or they lack self-learning capabilities, making it difficult to maintain high detection accuracy across different scenarios.
[0005] Over-the-air (OTA) upgrades of industrial equipment often lack robust security mechanisms, resulting in high upgrade risks and making it difficult to adapt to the needs of agile manufacturing.
[0006] Therefore, there is an urgent need for an intelligent detection system that can integrate multi-dimensional data, possess highly reliable intelligent decision-making capabilities, and support remote security evolution. Summary of the Invention
[0007] This invention provides a method, system, terminal, and storage medium for detecting fine metal wires based on edge-cloud collaboration, aiming to solve the technical problems of low detection efficiency and insufficient accuracy in existing methods.
[0008] To achieve the above objectives, this invention provides a method for detecting fine metal wires based on edge-cloud collaboration. The method is applied to an industrial inspection system comprising a detection terminal, edge nodes, and a cloud server. The method includes the following steps: The detection terminal collects physical parameter data of the metal wire in real time through an integrated sensor group, and performs standardization processing and anomaly marking on the physical parameter data at the edge node to obtain the detection data to be evaluated. The cloud server receives the detection data to be evaluated and triggers the rule engine evaluation process and the machine learning model evaluation process in parallel. The rule engine evaluation process generates a first evaluation result based on a preset industry standard threshold. The machine learning model evaluation process uses a pre-trained classification model to infer the detection data to be evaluated and generates a second evaluation result and the corresponding model confidence. The cloud server generates the final quality judgment instruction by: automatically generating a determined quality judgment result when the first evaluation result is consistent with the second evaluation result and the model confidence is greater than or equal to a preset confidence threshold; and generating a manual review request instruction when the first evaluation result is inconsistent with the second evaluation result or the model confidence is less than the preset confidence threshold. The final quality judgment command is sent to the detection terminal or user interface for visualization or alarm control.
[0009] The following are preferred technical solutions of the present invention, but are not intended to limit the technical solutions provided by the present invention. The purpose and beneficial effects of the present invention can be better achieved and realized through the following preferred technical solutions.
[0010] As a preferred technical solution, the method for detecting fine metal wires further includes: real-time monitoring of the service status of the machine learning model evaluation process; when the machine learning model service is detected to be unavailable, automatically switching to a single rule engine mode, and generating the final quality judgment instruction based solely on the first evaluation result.
[0011] As a preferred technical solution, the aforementioned method for detecting metal micro-wires further includes an industrial detection system that integrates a large language model interaction module, and the method also includes: Receive a natural language query request input by the user; parse the query request using a large language model, and match the target tool function from a pre-built tool function library; the tool function library includes device status query functions, wire data query functions, and quality diagnosis functions for accessing the database; The large language model autonomously calls the target tool function to access the database and obtain real-time detection information. The real-time detection information is integrated into an analysis report in natural language format and fed back to the user.
[0012] As a preferred technical solution, the aforementioned method for detecting fine metal wires, wherein the detection terminal is based on an embedded microcontroller and real-time operating system architecture, performs lightweight data analysis, calculates parameter trends in real time before uploading to the cloud, and presents them on the local display terminal.
[0013] As a preferred technical solution, the method for detecting metal micro-wires further includes: Obtain the historical time-series characteristics of the detection data to be evaluated; Use a quality trend prediction model to predict the quality trend of this batch of wires within a preset future time period; The final quality assessment instruction, the model confidence level, and the predicted quality trend data are encoded to generate a product QR code containing a dynamic risk identifier.
[0014] As a preferred technical solution, the method for detecting fine metal wires further includes: a cloud server constructing an upgrade package containing a version verification code and distributing it to the detection terminal; the detection terminal verifying the source reliability and file integrity of the upgrade package; and executing a firmware update after successful verification. If an abnormality occurs during the update process, the update is automatically rolled back to the previous stable version.
[0015] Secondly, a metal micro-wire detection system, comprising: The data acquisition and edge processing module is configured at the testing site to acquire wire diameter, conductivity and weight parameters through sensors, and to perform preliminary data cleaning and anomaly marking. The cloud communication module is used to realize the data streaming transmission between the collection terminal and the cloud through a message queue; the dual-core evaluation module is configured on the cloud server and includes a rule engine unit and a machine learning inference unit that run in parallel, used to output the first evaluation result and the second evaluation result with confidence level respectively. The intelligent decision control module is used to compare the consistency between the first evaluation result and the second evaluation result and to determine whether the confidence level meets the preset threshold, and select the automatic judgment path or the manual review path accordingly. The data storage and management module is used to store detection data and judgment results, and supports full lifecycle traceability.
[0016] Thirdly, a computer-readable storage medium stores an intelligent detection program for fine metal wires based on edge-cloud collaboration, wherein the intelligent detection program for fine metal wires based on edge-cloud collaboration, when executed by a processor, implements the steps of the metal fine wire detection method based on edge-cloud collaboration as described above.
[0017] Beneficial effects: Compared to existing technologies, this invention employs a mechanism of collaborative work between a rule engine and a machine learning model, supplemented by a confidence threshold judgment strategy. This retains the determinism and interpretability of rule-based judgment while leveraging the model's adaptability in complex scenarios, effectively overcoming the shortcomings of single methods in terms of flexibility or interpretability, and significantly improving the consistency and reliability of detection results. It breaks through the limitations of traditional detection architectures in response latency, deploying a lightweight analysis module on resource-constrained embedded devices (such as STM32) while performing deep inference in the cloud, balancing real-time performance and computing power, and optimizing the overall detection response speed. It introduces a natural language operation interface, supporting querying and managing industrial databases through natural language-like commands; simultaneously, it constructs an industrial-grade OTA upgrade mechanism with security verification and rollback functions, effectively reducing operational risks during system maintenance. Embedding product quality prediction results into a QR code traceability system upgrades the system from a traditional passive information query mode to one capable of proactively identifying and warning of potential quality risks, enhancing the overall quality control level. Attached Figure Description
[0018] Figure 1 This is a block diagram of a metal micro-wire detection system based on edge-cloud collaboration provided in an embodiment of the present invention; Figure 2 This is a schematic diagram of the process of the metal micro-wire detection method based on edge-cloud collaboration provided in an embodiment of the present invention; Figure 3 This is a transmission layer communication architecture diagram of the metal micro-wire detection system based on end-to-cloud collaboration provided in an embodiment of the present invention; Figure 4 These are photographs of the physical hardware provided in the embodiments of the present invention. Detailed Implementation
[0019] To make the objectives, technical solutions, and advantages of this invention clearer and more explicit, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative of the invention and are not intended to limit the invention.
[0020] like Figures 1 to 4As shown, this invention provides an edge-cloud collaborative architecture. Hardware (Edge): Utilizing an STM32 microcontroller with the FreeRTOS real-time operating system. Connects to a laser diameter gauge, resistance tester, and weighing sensor. The edge not only collects data but also incorporates a lightweight data analysis module for real-time parameter trend calculation and local screen display. Edge Layer: Employs an edge gateway (e.g., based on the Huawei Cloud IoT platform) to standardize and clean the data, perform protocol conversion (RS485 / RS232 to MQTT / AMQP), and conduct preliminary rule filtering. Cloud: Deployed on a cloud server based on a Spring Boot microservice architecture. Core components include the openGauss database, message queue (AMQP), rule engine service, and AI inference service.
[0021] like Figure 2 As shown, this invention provides a method for detecting fine metal wires based on edge-cloud collaboration. The method is applied to an industrial inspection system including a detection terminal, edge nodes, and a cloud server. The method includes the following steps: S101. The detection terminal collects physical parameter data of the metal wire in real time through an integrated sensor group, and performs standardization processing and anomaly marking on the physical parameter data at the edge node to obtain the detection data to be evaluated. S102. The cloud server receives the detection data to be evaluated and triggers the rule engine evaluation process and the machine learning model evaluation process in parallel. The rule engine evaluation process generates a first evaluation result based on a preset industry standard threshold. The machine learning model evaluation process uses a pre-trained classification model to infer the detection data to be evaluated and generates a second evaluation result and the corresponding model confidence. S103, The cloud server generates the final quality judgment instruction by: automatically generating a determined quality judgment result when the first evaluation result is consistent with the second evaluation result and the model confidence is greater than or equal to a preset confidence threshold; and generating a manual review request instruction when the first evaluation result is inconsistent with the second evaluation result or the model confidence is less than the preset confidence threshold. S104. Send the final quality judgment command to the detection terminal or user interface for visual presentation or alarm control. The specific steps are as follows: Step S1: Data Input. The cloud receives multi-dimensional vector data containing wire diameter, conductivity, and tensile strength. Step S2: Parallel Evaluation. Path A (Rule Engine): The system loads the corresponding standard threshold file based on the application scenario selected in the current work order (e.g., "Aerospace Grade"). If the data is within the threshold range, it outputs "Qualified"; otherwise, it outputs "Unqualified". This is the first evaluation result R1. Path B (Machine Learning Model): The data is input into a pre-trained classification model (in this embodiment, a random forest model trained based on 100,000 historical data points is used). The model outputs the predicted category (Qualified / Unqualified) as the second evaluation result R2, and simultaneously outputs the probability value of this prediction as the confidence level C (e.g., 0.85). Step S3: Intelligent Decision. Judgment Logic: If R1 = R2 and C = 0.8 (preset threshold), the system automatically accepts this result and marks it as "Automatic Judgment". If C < 0.8, the system marks the record as "Pending Review" and pushes it to the manual review terminal. Exception handling: If the ML service response times out or becomes unavailable, the system will automatically degrade and use only R1 as the final result to ensure uninterrupted production.
[0022] In this embodiment, the system integrates the DeepSeek large language model, but it is not a simple dialogue; instead, it achieves industrial control through "tool use" technology. Tool definition: 11 tool functions are predefined in the system, such as query_device_status(device_id), analyze_quality_trend(batch_id), and query_wire_specs(spec_id). Execution flow: The user inputs: "Is the average diameter of the most recent batch of wire from machine #3 abnormal?" The LLM parses the semantics and identifies that the query_device_status and analyze_quality_trend functions need to be called. The LLM generates function call parameters, and the agent program executes the function to access the openGauss database. The database returns structured data (e.g., average diameter 0.05mm, standard deviation 0.0001). The LLM generates a natural language report based on the returned data: "The average diameter of the most recent batch of wire from machine #3 is 0.05mm, which is within the normal range. The standard deviation is small, indicating stable process." This embodiment incorporates a secure upgrade mechanism: Version Management: Upgrade packages are generated in the cloud and include digital signatures. Distribution and Verification: After downloading, the device first verifies the signature to ensure reliable origin, and then verifies file integrity (MD5 / SHA256). Atomic Update and Rollback: The system employs a dual-partition mechanism (A / B partitions). The new version is installed on partition B, and the bootloader switches after successful verification. If startup fails or malfunctions, the watchdog mechanism automatically resets and switches back to the old version on partition A, ensuring the device remains securely inoperable.
[0023] In this embodiment, when generating a product QR code, the predictive traceability system not only writes static production time and raw material information, but also calls a cloud-based quality trend prediction model (based on time series analysis) to predict the potential quality trend of the product in subsequent processing or use (such as oxidation risk level). This information is encoded into the QR code, and downstream manufacturers can obtain "early warnings" by scanning the code, thereby optimizing their processing technology.
[0024] Furthermore, the present invention also provides a computer-readable storage medium, wherein the computer-readable storage medium stores an intelligent detection program for fine metal wires based on edge-cloud collaboration, and when the intelligent detection program for fine metal wires based on edge-cloud collaboration is executed by a processor, it implements the steps of the metal fine wire detection method based on edge-cloud collaboration as described above.
[0025] In summary, this invention provides a method, system, and storage medium for detecting fine metal wires based on edge-cloud collaboration. The method is applied to an industrial inspection system including a detection terminal, edge nodes, and a cloud server. The method includes the following steps: the detection terminal collects physical parameter data of metal wire in real time through an integrated sensor group, and the edge node performs standardization processing and anomaly marking on the physical parameter data to obtain inspection data to be evaluated; the cloud server receives the inspection data to be evaluated and triggers a rule engine evaluation process and a machine learning model evaluation process in parallel; wherein, the rule engine evaluation process generates a first evaluation result based on a preset industrial standard threshold; the machine learning model evaluation process uses a pre-trained classification model to infer the inspection data to be evaluated, generating a second evaluation result and a corresponding model confidence score; the cloud server generates a final quality judgment instruction including: automatically generating a determined quality judgment result when the first evaluation result is consistent with the second evaluation result and the model confidence score is greater than or equal to a preset confidence threshold; generating a manual review request instruction when the first evaluation result is inconsistent with the second evaluation result, or the model confidence score is less than the preset confidence threshold; and sending the final quality judgment instruction to the detection terminal or user interface for visualization or alarm control. This invention employs a mechanism that combines a rule engine with a machine learning model, supplemented by a confidence threshold judgment strategy. This retains the determinism and interpretability of rule-based judgments while leveraging the model's adaptability in complex scenarios, effectively overcoming the shortcomings of single methods in terms of flexibility or interpretability, and significantly improving the consistency and reliability of detection results. It overcomes the limitations of traditional detection architectures in response latency by deploying a lightweight analysis module on resource-constrained embedded devices (such as STM32) and performing deep inference in the cloud, balancing real-time performance and computing power, and optimizing the overall detection response speed. A natural language operation interface is introduced, supporting querying and managing the industrial database through natural language-like commands; simultaneously, an industrial-grade OTA upgrade mechanism with security verification and rollback functions is constructed, effectively reducing operational risks during system maintenance. Product quality prediction results are embedded into a QR code traceability system, upgrading the system from a traditional passive information query mode to one that proactively identifies and warns of potential quality risks, enhancing the overall quality control level.
[0026] 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 terminal 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 terminal. 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 terminal that includes that element.
[0027] Of course, those skilled in the art will understand that all or part of the processes in the above embodiments can be implemented by a computer program instructing related hardware (such as a processor, controller, etc.). The program can be stored in a computer-readable storage medium, and when executed, it can include the processes described in the above method embodiments. The computer-readable storage medium can be a memory, magnetic disk, optical disk, etc.
[0028] It should be understood that the application of the present invention is not limited to the examples above. Those skilled in the art can make improvements or modifications based on the above description, and all such improvements and modifications should fall within the protection scope of the appended claims.
Claims
1. A method for detecting fine metal wires based on edge-cloud collaboration, characterized in that, The method is applied to an industrial inspection system that includes a detection terminal, an edge node, and a cloud server, and the method includes the following steps: The detection terminal collects physical parameter data of the metal wire in real time through an integrated sensor group, and performs standardization processing and anomaly marking on the physical parameter data at the edge node to obtain the detection data to be evaluated. The cloud server receives the detection data to be evaluated and triggers the rule engine evaluation process and the machine learning model evaluation process in parallel. The rule engine evaluation process generates a first evaluation result based on a preset industry standard threshold. The machine learning model evaluation process uses a pre-trained classification model to infer the detection data to be evaluated and generates a second evaluation result and the corresponding model confidence. The cloud server generates the final quality judgment instruction by: automatically generating a determined quality judgment result when the first evaluation result is consistent with the second evaluation result and the model confidence is greater than or equal to a preset confidence threshold; and generating a manual review request instruction when the first evaluation result is inconsistent with the second evaluation result or the model confidence is less than the preset confidence threshold. The final quality judgment command is sent to the detection terminal or user interface for visualization or alarm control.
2. The method for detecting metallic micro-wires according to claim 1, characterized in that, The method further includes: real-time monitoring of the service status of the machine learning model evaluation process; when the machine learning model service is detected to be unavailable, automatically switching to a single rule engine mode, and generating the final quality judgment instruction based solely on the first evaluation result.
3. The method for detecting metallic micro-wires according to claim 1, characterized in that, The industrial testing system also integrates a large language model interaction module, and the method further includes: Receive a natural language query request input by the user; parse the query request using a large language model, and match the target tool function from a pre-built tool function library; the tool function library includes device status query functions, wire data query functions, and quality diagnosis functions for accessing the database; The large language model autonomously calls the target tool function to access the database and obtain real-time detection information. The real-time detection information is integrated into an analysis report in natural language format and fed back to the user.
4. The method for detecting metallic micro-wires according to claim 1, characterized in that, The detection terminal is based on an embedded microcontroller and real-time operating system architecture, performs lightweight data analysis, calculates parameter trends in real time before uploading to the cloud, and displays them on the local terminal.
5. The method for detecting metallic micro-wires according to claim 1, characterized in that, The method also includes: Obtain the historical time-series characteristics of the detection data to be evaluated; Use a quality trend prediction model to predict the quality trend of this batch of wires within a preset future time period; The final quality assessment instruction, the model confidence level, and the predicted quality trend data are encoded to generate a product QR code containing a dynamic risk identifier.
6. The method for detecting metallic micro-wires according to claim 1, characterized in that, It also includes: a cloud server constructing an upgrade package containing a version verification code and distributing it to the detection terminal; the detection terminal verifying the source reliability and file integrity of the upgrade package; after the verification is successful, the firmware update is executed, and if an abnormality occurs during the update process, it is automatically rolled back to the previous stable version.
7. A metal micro-wire detection system, characterized in that, include: The data acquisition and edge processing module is configured at the testing site to acquire wire diameter, conductivity and weight parameters through sensors, and to perform preliminary data cleaning and anomaly marking. The cloud communication module is used to realize the data streaming transmission between the collection terminal and the cloud through a message queue; the dual-core evaluation module is configured on the cloud server and includes a rule engine unit and a machine learning inference unit that run in parallel, used to output the first evaluation result and the second evaluation result with confidence level respectively. The intelligent decision control module is used to compare the consistency between the first evaluation result and the second evaluation result and to determine whether the confidence level meets the preset threshold, and select the automatic judgment path or the manual review path accordingly. The data storage and management module is used to store detection data and judgment results, and supports full lifecycle traceability.
8. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores an intelligent detection program for fine metal wires based on edge-cloud collaboration. When the intelligent detection program for fine metal wires based on edge-cloud collaboration is executed by a processor, it implements the steps of the method for detecting fine metal wires based on edge-cloud collaboration as described in any one of claims 1-6.