Dynamic intelligent stamping production line real-time monitoring strategy based on digital twinning

Abstract

The invention relates to a dynamic intelligent stamping production line real-time monitoring strategy based on digital twinning, and adopts the technical scheme that a sensing layer acquires production total factor data related to a stamping production line in real time through multi-sensor fusion and an RFID (Radio Frequency Identification) technology, performs primary processing by utilizing edge calculation, and transmits the data to a real-time reaction layer; the real-time reaction layer quickly processes and analyzes multi-source data through edge calculation and a real-time database, detects faults in real time by using a multi-modal fusion technology, triggers an alarm mechanism, and feeds back to the interaction layer; and the interaction layer integrates intelligent production scheduling and NLP voice interaction through data interaction processing, error verification and multi-protocol adaptation to realize virtual factory visualization and equipment health real-time monitoring. Multi-modal data fusion is realized, an adaptive incremental learning and early warning mechanism is realized, an information island is broken, and millisecond-level response is realized. The strategy can improve the production efficiency, reduce the operation and maintenance cost, prolong the service life of equipment, and provide an efficient solution for intelligent transformation of the manufacturing industry.

Description

Technical Field

[0001] The present invention relates to the technical field of intelligent manufacturing and industrial Internet of Things, and specifically to a real-time monitoring strategy for a dynamic intelligent stamping production line based on digital twin technology. Background Art

[0002] Amid the global trend toward digitalization and intelligent manufacturing, manufacturing companies face the critical task of improving production efficiency and completing digital transformation. As an emerging technology, digital twinning is gradually transforming industrial production and management. By closely integrating the virtual and real worlds, it offers a new perspective and approach, driving the construction and development of smart factories. Through applications such as real-time monitoring, data analysis, process optimization, and equipment maintenance, digital twinning not only improves production efficiency and product quality, but also reduces operating costs and risks, making it an indispensable core technology for intelligent manufacturing.

[0003] Digital twin technology now enables detailed reconstruction and simulation of production pipelines, processing equipment, robots, and other facilities, providing strong support for production process optimization and production planning simulation. Consequently, issues such as equipment health management and fault warnings on the shop floor have become more pressing. Traditional monitoring strategies for stamping production lines suffer from outdated monitoring equipment and technology, insufficient sensor performance, and low sensitivity, making it difficult to accurately capture subtle changes on the production line. Simple data processing algorithms can be ineffective in extracting valuable information from massive amounts of data. Monitoring coverage is sparse, failing to fully cover every aspect of the production line. Unclear monitoring priorities can lead to missed critical production parameters and fault warning signals. Monitoring systems lack intelligence, lacking intelligent warning and diagnostic capabilities. These systems are unable to automatically identify abnormalities on the production line, requiring manual intervention and judgment. Lack of data analysis and optimization capabilities prevents in-depth mining and analysis of monitoring data to identify potential problems and areas for improvement in the production process. Furthermore, they are unable to automatically adjust production parameters based on monitoring data to achieve optimized and intelligent control of the production process. Traditional monitoring systems are disconnected from production management, resulting in severe information silos. Monitoring data cannot be transmitted to production management in a timely manner. Production management cannot obtain real-time information on the status of the production line, making it difficult to make timely and effective decisions and adjustments. The system also lacks collaborative capabilities, preventing the monitoring system from cooperating with other production links (such as raw material supply, logistics distribution, and quality inspection), hindering overall optimization of the production process.

[0004] Inadequate control over all aspects of the production line resulted in extended detection and repair times for machine failures. Furthermore, incomplete management of on-the-job workers often resulted in unsupervised operations. Furthermore, temperature and humidity in the workshop, as well as protocol compatibility issues between individual devices, impacted equipment lifespan and product quality, severely limiting production line efficiency. Therefore, a real-time monitoring strategy for intelligent stamping production lines was urgently needed to address challenges in device connectivity and intelligence, thereby improving production line efficiency. Summary of the Invention

[0005] To effectively address the problems of poor real-time performance, data silos, and insufficient intelligence in existing monitoring strategies, this paper proposes a dynamic intelligent real-time monitoring strategy for stamping production lines based on digital twins. This strategy implements multimodal data fusion, adaptive incremental learning, and an intelligent closed-loop early warning mechanism combining feature extraction, anomaly diagnosis, and decision feedback, breaking down information silos and achieving millisecond-level response. This strategy improves production efficiency, reduces maintenance costs, and extends equipment life, providing a highly effective solution for the intelligent transformation of the manufacturing industry.

[0006] To achieve the above objectives, the technical solution of the present invention is:

[0007] The real-time monitoring strategy of the dynamic intelligent stamping production line based on digital twin includes the following steps:

[0008] S1. Multimodal data acquisition at the perception layer

[0009] Through multi-sensor fusion and radio frequency identification (RFID) technology, all production factor data related to the stamping production line is collected in real time and aggregated to the edge layer;

[0010] S2. Edge layer data preprocessing

[0011] All production factor data is denoised and normalized through edge computing to generate standard stream data and transmit it to the real-time reaction layer;

[0012] S3. Real-time reaction layer intelligent analysis

[0013] Analyze standard stream data through a real-time database, use multimodal fusion technology to detect faults in real time, trigger alarm mechanisms, and provide feedback to the interaction layer;

[0014] S4. Interaction layer visualization and intelligent control

[0015] The interaction layer integrates intelligent production scheduling and NLP voice interaction through data interaction processing, error checking and multi-protocol adaptation, realizing virtual factory visualization and real-time monitoring of equipment health.

[0016] Furthermore, the perception layer in step S1 includes radio frequency identification technology (RFID) and multi-sensor technology,

[0017] The radio frequency identification technology (RFID) integrates the management of personnel information and receives information about business processes;

[0018] The multi-sensor technology is the perception data of various sensors on the environmental information of the stamping production line.

[0019] Furthermore, the radio frequency identification technology (RFID) regulates product warehousing, equipment quantity and model verification, and personnel management, checks in real time whether there are any omissions during the handover process, automatically identifies goods in and out of the warehouse, and labels and tracks parts and products.

[0020] Furthermore, the edge layer adopts edge computing technology to place data processing and computing resources on edge devices, sensors or user devices close to the source of data generation to provide computing and data analysis capabilities.

[0021] Furthermore, the real-time reaction layer in step S3 includes a model analysis platform, which includes a KNN-VIT composite model algorithm. The ViT model is first used to extract features from the image to obtain a high-dimensional feature representation of the image, and the monitoring area is monitored in real time to capture video images. These feature representations are used as input to the KNN algorithm, and classification is performed by calculating the distance between the new image and the training image in the feature space.

[0022] Furthermore, the alarm mechanism in step S3 is to identify data anomalies in real time through anomaly detection and big data analysis, display information in combination with a visualization platform, and push it instantly through multiple channels.

[0023] Specifically, the interaction layer includes a visualization platform and a business management platform.

[0024] The visualization platform is based on the Unity engine and uses digital twin technology to achieve three-dimensional visualization simulation of the entire stamping production line;

[0025] The business management platform is based on full-process business transparency integration technology, including behavioral process modeling, standard step formulation, production line operation monitoring and process compliance judgment.

[0026] Furthermore, the interaction layer includes a voice assistant based on natural language processing technology (NLP), which realizes cloud platform information interaction and business progress broadcasting, and supports voice query of production line process status.

[0027] Furthermore, the real-time database uses multiple types of databases to support its real-time data processing and analysis needs.

[0028] Specifically, the real-time reaction layer is integrated into the interaction layer through the OPC UA architecture supporting multi-protocol conversion.

[0029] Beneficial effects of the present invention:

[0030] 1. This invention, based on the deep integration of digital twin technology and stamping production lines, builds a specialized real-time monitoring system architecture. Through the systematic integration of functional modules, it provides assurance for the manufacturing process in the factory manufacturing industry, reduces factory production risks, improves the safety and simplicity of workers' working hours, and extends machine life.

[0031] Second, the visualization platform simplifies equipment inspections. Using digital twin technology to integrate sensor data, it transparently displays production processes and enables refined management. Multiple account logins improve supervisory efficiency, while automated alarm mechanisms predict issues, ensuring workers can respond promptly and prevent safety risks.

[0032] 3. The real-time database, by comprehensively scheduling multiple database resources and integrating them with edge computing technology, can complete computing tasks on the infrastructure side in the shortest possible time. Furthermore, by deploying multiple sensors to collect real-time data, the port compiler can perform simultaneous computations while data is being collected, facilitating simultaneous real-time monitoring and management.

[0033] 4. Utilizing the KNN-VIT composite monitoring algorithm, a deep learning-based real-time production monitoring system was developed. The VIT module uses the Transformer architecture to process production line images in parallel, enabling rapid extraction of abnormal features and improving monitoring response speed. The KNN component, based on a dynamic feature library based on historical normal samples, uses similarity comparison to provide production deviation warnings, addressing the lag inherent in traditional single-sample libraries. The algorithm combines multimodal monitoring data and synchronously analyzes equipment status and process parameters. Once an image instance is detected to deviate from the process standard beyond a threshold, a three-level warning mechanism is triggered. The system dynamically updates the monitoring model through an adaptive learning mechanism, addressing blind spots in the sensor network and establishing an intelligent closed-loop monitoring system, providing a decision-making basis for production quality control. The KNN-VIT model integrates an incremental learning framework and updates the feature library every 24 hours, ensuring the timeliness of the monitoring system and the advancement of networked training.

[0034] 5. The business management platform uses full-process visualization integration technology to conduct in-depth analysis of business lists, build business models and formulate a series of process divisions to ensure that each link has corresponding steps, which are fed back to the monitoring system, ultimately achieving the effect of step-by-step monitoring, which is conducive to the unified formulation of production process standards and quality control of finished products. BRIEF DESCRIPTION OF THE DRAWINGS

[0035] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. The drawings are used to provide a further understanding of the present invention and constitute a part of the specification. Together with the embodiments of the present invention, they are used to explain the present invention and do not constitute a limitation of the present invention. For those skilled in the art, other drawings can be obtained based on the drawings without paying any creative work.

[0036] Figure 1 Schematic diagram of the system architecture of the present invention;

[0037] Figure 2 Schematic diagram of the operating mode of the present invention. DETAILED DESCRIPTION

[0038] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0039] like Figure 1-2 The real-time monitoring strategy of the dynamic intelligent stamping production line based on digital twin shown in the figure includes the following steps:

[0040] S1. Multimodal data acquisition at the perception layer: This uses multi-sensor fusion and RFID technology to collect all production factor data related to the stamping line in real time and aggregate it to the edge layer.

[0041] S2. Edge layer data preprocessing: All production factor data is de-noised and normalized through edge computing, generating standard stream data for transmission to the real-time reaction layer.

[0042] S3. Real-time reaction layer intelligent analysis: This layer analyzes standard stream data through a real-time database, uses multimodal fusion technology to detect faults in real time, triggers alarm mechanisms, and provides feedback to the interaction layer.

[0043] S4. Interaction layer visualization and intelligent control: The interaction layer integrates intelligent production scheduling and NLP voice interaction through data interaction processing, error checking, and multi-protocol adaptation to achieve virtual factory visualization and real-time equipment health monitoring.

[0044] The perception layer of multimodal data collection in step S1 includes radio frequency identification technology (RFID) and multi-sensor technology. With the help of the coordinated integration of multiple sensors in the workshop, various sensors such as temperature, humidity, vibration and vision are fully deployed, and radio frequency identification technology (RFID) is applied to collect environmental parameters, equipment operation status and personnel activity data in real time. At the same time, it closely tracks the material flow, product production progress, personnel movement trajectory and other production data information of all factors and aggregates them to the edge layer through the ZigBee gateway.

[0045] The radio frequency identification technology (RFID) includes tags, readers and antennas. RFID can be used to regulate product warehousing, equipment quantity and model verification, and personnel management. It can also check in real time whether there are any omissions during the handover process, automatically identify goods entering and leaving the warehouse, identify and track parts and products, and realize the integrated management of personnel information and the reception of information on business processes.

[0046] The multi-sensor technology is the perception of environmental information by various sensors. Various sensors cooperate with compilers in various devices and air circulation areas. Equipment that is sensitive to environmental factors needs to monitor environmental changes in real time, while also providing guarantees for product quality. The addition of sensors such as temperature sensors, humidity sensors, and visual sensors enables rapid data collection.

[0047] The edge layer in step S2 edge layer data preprocessing uses edge computing technology, placing data processing and computing resources on edge devices, sensors, or user devices close to the data source to provide computing and data analysis capabilities. Production full-factor data is aggregated to edge nodes via the ZigBee gateway, and the Flink engine performs denoising and normalization processing. Edge computing technology combined with an edge compiler can process production full-factor data information at the data source, generate standard stream data, and push it to the real-time database via the MQTT protocol. This enables rapid data processing, reduces latency, and achieves real-time results.

[0048] In terms of hardware, edge nodes are equipped with NVIDIA Jetson AGX Orin modules, leveraging the CUDA parallel computing platform for GPU acceleration. FPGAs (such as the Xilinx Zynq series) are also integrated for high-real-time signal processing. For data processing, a stream processing engine based on Apache Flink is used, optimizing event time windows and state management to achieve millisecond-level sensor data aggregation. Docker container deployment is combined with Kubernetes (such as the lightweight K3s) for dynamic scaling, and lightweight images such as Alpine are used to reduce resource overhead.

[0049] The real-time reaction layer in the step S3 real-time reaction layer intelligent analysis includes a model analysis platform. The model analysis platform adopts the KNN-VIT composite model algorithm. The ViT model is first used to extract features from the image to obtain a high-dimensional feature representation of the image. The monitoring area is monitored in real time using visual sensors (such as cameras) and other devices to capture video images. These feature representations are used as input to the KNN algorithm, and classification is performed by calculating the distance between the new image and the training image in the feature space.

[0050] The KNN-VIT composite model was optimized through dataset training and model parameter adjustments. Dataset training involved collecting 100,000 images of stamped parts (including both normal and defective samples) and annotating them down to the pixel level for training. Model parameter adjustments included ViT using a 12-layer Transformer with a patch size of 16. KNN used k=5, with cosine similarity as the distance metric. Finally, an adaptive online learning framework was designed, integrating the KNN-VIT model with an incremental learning framework. The feature library was updated every 24 hours to support the automatic identification of new defect types.

[0051] Specifically, the real-time database supports its real-time data processing and analysis needs by adopting multiple types of databases, including time series database (TSDB), NoSQL database and HBase. In order to enable the monitoring strategy to quickly identify fault information, the KNN-VIT composite model is adopted through machine learning, combined with ViT (Vision Transformer) to extract image features, and the KNN algorithm is used to classify and detect production anomalies, and compare them with the real-time input environmental information. The innovative multimodal fusion technology combines sensor data (temperature, humidity, vibration) with visual data and inputs the multimodal Transformer to output a comprehensive anomaly score, quickly detect whether a fault has occurred, and feed it back to the interaction layer; in order to solve the lack of accurate monitoring data caused by problems such as protocol incompatibility,

[0052] The alarm mechanism uses anomaly detection and big data analysis to identify data anomalies in real time, combines it with a visualization platform to display information, and utilizes multiple channels for instant push notifications. Specifically, after comparing data instructions using anomaly detection algorithms and big data analysis technology, users can conveniently view various types of information such as monitoring images, alarm information, and device status through an intuitive visualization interface, allowing them to quickly make judgments and take appropriate actions. In addition, this alarm mechanism also has the function of sending information to multiple parties through multiple channels such as email and phone. Information can be sent directly to the contact information recorded in the database of responsible persons and managers, enabling simultaneous transmission to multiple parties to ensure the timeliness of the alarm.

[0053] The real-time response layer integrates the OPC UA (OPC Unified Architecture) protocol and supports automatic conversion of Modbus, Profinet, and EtherCAT protocols, allowing various types of data to be effectively integrated into the interaction layer.

[0054] The interactive layer in step S4, the interactive layer visualization and intelligent control, includes a visualization platform, natural language processing (NLP), and a business management platform. The visualization platform, built on the Unity engine, utilizes digital twin technology to achieve 3D visualization simulation of the entire stamping production line. Through digital modeling of physical entities, remote real-time monitoring of production line operations is achieved. Lightweight modeling technology is used to refine the stamping line process, enabling rapid data analysis and processing, ensuring real-time data processing.

[0055] A voice assistant based on natural language processing (NLP) technology enables information exchange and business progress reporting within the visualization platform, as well as voice queries on production line status. The application of NLP technology is tightly integrated with the visualization platform, enabling the voice assistant to report issues requiring rectification in text format and provide real-time information on production environment temperature indicators, ensuring healthy and safe production. Furthermore, managers can issue verbal commands to direct the visualization platform to monitor any aspect of the situation without manual intervention, allowing them to meticulously monitor details such as the wear of specific parts on the stamping line.

[0056] The business management platform is based on full-process business transparency integration technology, including behavioral process modeling, standard step formulation, production line operation monitoring, and process compliance judgment. The business management platform compares the actual operation process with the preset path to determine whether it complies with the specification and make corresponding decisions. Specifically, the business management platform accepts the data from the above process modeling, calculates the workload of the stamping production line, reasonably arranges the production sequence of products, and exports the results in the form of charts for management personnel to review. After approval, an order is issued to start production, and the priority of the instructions can also be manually modified.

[0057] After various sensors sense environmental information, they receive and process the data, enabling interaction between hardware and software, making data transmission more efficient. Furthermore, the interaction layer detects data errors through mechanisms such as checksums and cyclic redundancy checks (CRCs). It also provides adaptation capabilities to accommodate differences in data transmission rates and lengths between different communication protocols. This allows data to be freely transmitted and collected across different communication protocols, ensuring efficient and stable data transmission between different devices.

[0058] The above description is merely a preferred embodiment of the present invention and does not constitute any form of limitation to the present invention. Although the present invention has been disclosed as above in terms of a preferred embodiment, it is not intended to limit the present invention. Any person skilled in the art can, without departing from the scope of the technical solution of the present invention, make some changes or modifications to equivalent embodiments using the technical contents disclosed above. However, any brief modifications, equivalent changes and modifications made to the above embodiments based on the technical essence of the present invention without departing from the content of the technical solution of the present invention are still within the scope of the technical solution of the present invention.

Claims

1. A real-time monitoring strategy for a dynamic intelligent stamping production line based on digital twins, characterized by: The following steps are involved: S1. Multimodal data acquisition at the perception layer Through multi-sensor fusion and radio frequency identification (RFID) technology, all production factor data related to the stamping production line is collected in real time and aggregated to the edge layer; S2. Edge layer data preprocessing All production factor data is denoised and normalized through edge computing to generate standard stream data and transmit it to the real-time reaction layer; S3. Real-time reaction layer intelligent analysis Analyze standard stream data through a real-time database, use multimodal fusion technology to detect faults in real time, trigger alarm mechanisms, and provide feedback to the interaction layer; S4. Interaction layer visualization and intelligent control The interaction layer integrates intelligent production scheduling and NLP voice interaction through data interaction processing, error checking and multi-protocol adaptation, realizing virtual factory visualization and real-time monitoring of equipment health.

2. The real-time monitoring strategy according to claim 1, characterized in that: The perception layer in step S1 includes radio frequency identification technology (RFID) and multi-sensor technology. The radio frequency identification technology (RFID) integrates the management of personnel information and receives information about business processes; The multi-sensor technology is the perception data of various sensors on the environmental information of the stamping production line.

3. The real-time monitoring strategy according to claim 2, characterized in that: The radio frequency identification technology (RFID) regulates product warehousing, equipment quantity and model verification, and personnel management, conducts real-time checks for omissions during shift changes, automatically identifies goods entering and leaving the warehouse, and labels and tracks parts and products.

4. The real-time monitoring strategy according to claim 1, characterized in that: The edge layer uses edge computing technology to place data processing and computing resources on edge devices, sensors or user devices close to the source of data generation to provide computing and data analysis capabilities.

5. The real-time monitoring strategy according to claim 1, characterized in that: The real-time reaction layer in step S3 includes a model analysis platform, which includes a KNN-VIT composite model algorithm. The ViT model is first used to extract features from the image to obtain a high-dimensional feature representation of the image. The monitoring area is monitored in real time and video images are captured. These feature representations are used as input to the KNN algorithm, and classification is performed by calculating the distance between the new image and the training image in the feature space.

6. The real-time monitoring strategy according to claim 1, characterized in that: The alarm mechanism in step S3 is to identify data anomalies in real time through anomaly detection and big data analysis, display information in combination with a visualization platform, and push information instantly through multiple channels.

7. The real-time monitoring strategy according to claim 1, characterized in that: The interaction layer includes a visualization platform and a business management platform. The visualization platform is based on the Unity engine and uses digital twin technology to achieve three-dimensional visualization simulation of the entire stamping production line; The business management platform is based on full-process business transparency integration technology, including behavioral process modeling, standard step formulation, production line operation monitoring and process compliance judgment.

8. The real-time monitoring strategy according to claim 1, characterized in that: The interaction layer includes a voice assistant based on natural language processing technology (NLP), which realizes cloud platform information interaction and business progress broadcasting, and supports voice query of production line process status.

9. The real-time monitoring strategy according to claim 1, characterized in that: The real-time database uses various types of databases to support its real-time data processing and analysis needs.

10. The real-time monitoring strategy according to claim 1, characterized in that: The real-time reaction layer is integrated into the interaction layer by supporting multi-protocol conversion through the OPC UA architecture.

Images