Intelligent manufacturing system and method based on digital twinning

Abstract

The application relates to the technical field of intelligent manufacturing systems, in particular to an intelligent manufacturing system and method based on digital twinning, a data acquisition module is used for acquiring manufacturing data in a manufacturing system, a data analysis module is connected with the data acquisition module to receive the manufacturing data and process the manufacturing data into feature data, a visualization module is connected with the data analysis module to receive the feature data and construct a virtual manufacturing scene, a real-time control module performs dynamic task control based on the feature data, and an early warning module is connected with the real-time control module to receive abnormal information and send adjustment instructions to the real-time control module; the modules in the intelligent manufacturing system form a closed loop of synergistic effect through data flow and information flow, the data analysis module performs multilevel decomposition and high-dimensional feature fusion, extracts key features in a manufacturing process, and finally displays the key features in the virtual manufacturing scene through the visualization module, so that accurate mapping of a manufacturing state is realized, and adjustment can be made in time when an abnormal state is found.

Description

Technical Field

[0001] This invention relates to the field of intelligent manufacturing system technology, and more specifically, to an intelligent manufacturing system and method based on digital twins. Background Technology

[0002] Currently, with the accelerated development of automation and intelligence in manufacturing, digital twin technology is gradually demonstrating its enormous potential in intelligent manufacturing systems. The core of a digital twin system lies in constructing a virtual model that reflects the physical manufacturing process in real time. By collecting, analyzing, and controlling the data and information flows between different modules, it achieves accurate simulation and real-time monitoring of the manufacturing process. However, existing intelligent manufacturing systems still face several technical bottlenecks in practical applications, particularly in terms of the depth of data processing, the real-time nature of system response, the flexibility of task scheduling, and the accurate identification and feedback of anomalies, which have not yet met the needs of complex manufacturing environments.

[0003] Current intelligent manufacturing systems primarily rely on single-level feature extraction and analysis methods, lacking the ability for high-dimensional data fusion and multi-layer feature decomposition. These systems typically employ fixed priority schemes for task scheduling and are limited to basic threshold judgment methods for data anomaly detection. Therefore, when complex dynamic changes or anomalies occur in the manufacturing process, existing systems struggle to respond and adjust promptly, easily leading to task delays, resource waste, and product quality degradation. Furthermore, the lack of feedback correction mechanisms in these systems causes a gradual accumulation of discrepancies between data analysis and actual manufacturing conditions, further impacting the system's control accuracy and stability. Existing technologies cannot effectively utilize the correlation between historical and real-time data to optimize the manufacturing process, especially in highly automated environments. Their processing modes are too simplistic and lack adaptability to complex environments, limiting the overall production efficiency and flexibility of the manufacturing system. Summary of the Invention

[0004] To address the aforementioned technical problems, this invention provides an intelligent manufacturing system and method based on digital twins. By establishing the synergistic effect of functional modules such as data acquisition, data analysis, visualization, real-time control, and early warning, it aims to solve the deficiencies of existing intelligent manufacturing systems in terms of real-time performance, data processing depth, and scheduling flexibility.

[0005] To solve the above-mentioned technical problems, the present invention adopts the following technical solution:

[0006] A digital twin-based intelligent manufacturing system includes a data acquisition module, a data analysis module, a visualization module, a real-time control module, and an early warning module. The data acquisition module collects manufacturing data from the manufacturing system. The data analysis module is connected to the data acquisition module to receive the manufacturing data and process it into feature data. The visualization module is connected to the data analysis module to receive the feature data and construct a virtual manufacturing scenario. The real-time control module performs dynamic task control based on the feature data. The early warning module is connected to the real-time control module to receive abnormal information and send adjustment commands to the real-time control module. The modules in the intelligent manufacturing system form a synergistic closed loop through data and information flows.

[0007] Preferably, the data analysis module further includes an adaptive multi-level feature decomposition unit for performing hierarchical feature decomposition on the manufacturing data, for generating multi-level feature data to obtain the decomposed feature layer data. The decomposed feature layer data F i ,j (L) Defined by the following formula:

[0008]

[0009] in, Let φ be the decomposition feature of the i,j unit in the Lth layer, and φ be the adaptive decomposition function used to layer the features of the manufacturing data. The input features are the i-th and j-th units of the previous layer, μ is the decomposition parameter representing the importance of the features, and σ is the hierarchical weight used to dynamically adjust the hierarchical ratio between features. The adaptive multi-level feature decomposition unit transmits the hierarchical feature data to the subsequent modules to form the basis of the data stream.

[0010] Preferably, the data analysis module further includes a complex feature mapping and nonlinear fusion unit, which receives the decomposed feature layer data. The high-dimensional feature matrix M is generated using the following formula. fusion :

[0011]

[0012] Among them, M fusion For the fused high-dimensional feature matrix, ψ k The feature fusion weighting factor is dynamically adjusted based on the relevance of the features. The nonlinear feature mapping function projects hierarchical data into a high-dimensional space, where the high-dimensional feature matrix M... fusion As input to the prediction unit for subsequent analysis.

[0013] Preferably, the data analysis module further includes a progressive dynamic prediction and feedback correction unit, which is based on a high-dimensional feature matrix M. fusion The current manufacturing status is dynamically predicted, and corrections are made based on feedback information. The predicted value P is obtained using the following formula. t+1 :

[0014] P t+1 =γ·η(P t M fusion )+k·Ω(F real -P t )

[0015] Among them, P t+1 The predicted value for the next moment is given by γ, which is a recursive adjustment factor that controls the weight of historical predictions. η is a recursive prediction function that performs dynamic prediction based on historical predictions and a high-dimensional feature matrix. κ is a feedback coefficient used to correct prediction bias, and Ω is a feedback correction function that adjusts the actual value F of the current manufacturing state. rea Compared with the predicted value P t The predicted value P is compared and corrected. t+1 It serves as input to the anomaly detection unit for anomaly analysis.

[0016] Preferably, the data analysis module further includes an adaptive anomaly detection and aggregation analysis unit, used for analyzing the predicted value P. t+1 And actual manufacturing data F real Obtain abnormal deviation detection value A detect The detected value is defined by the following formula:

[0017]

[0018] Among them, A detect δ is the current anomaly detection value, ρ is the adaptive adjustment factor used to regulate the sensitivity of anomaly deviation, ρ is the deviation amplification factor used to enhance the sensitivity of anomaly deviation detection, and ω is the current anomaly detection value. j Θ is the anomaly deviation weighting factor, used to adjust weights according to the anomaly type; Θ is the anomaly aggregation function, used to aggregate and analyze multidimensional anomaly information; and the anomaly deviation detection value A is... detect Used for dynamic priority decision-making in the task scheduling unit.

[0019] Preferably, the real-time control module further includes a task optimization priority scheduling and feedback control unit, used for adjusting the abnormal deviation detection value A. detect and predicted value P t+1 Obtain the optimized priority sequence O priority The priority sequence is defined by the following formula:

[0020]

[0021] Among them, O priority For the optimized priority sequence, α i τ is the priority weighting factor, dynamically adjusted according to the urgency of the task; τ is the task priority scheduling function, based on task type T. i The abnormal detection values ​​and predicted values ​​are used to sort the tasks. The priority sequence is used to adjust the robot operation order in real time to ensure that the manufacturing tasks are executed according to priority.

[0022] Preferably, the visualization module is further used for hierarchical feature data. and high-dimensional feature matrix M fusion A virtual manufacturing scenario is generated, which is used to display current manufacturing information in real time and to visualize future predictions.

[0023] Preferably, the early warning module is further used to base its warning on the anomaly detection value A. detect and the optimized priority sequence O priority It identifies abnormal states in the manufacturing process and sends adjustment instructions to the real-time control module when an abnormal state is detected, so as to adjust the operation tasks in the manufacturing process.

[0024] Preferably, the data acquisition module is further configured to acquire real-time data in the manufacturing environment through industrial image acquisition equipment and sensors, and encode the real-time data before sending it to the data analysis module.

[0025] A digital twin-based intelligent manufacturing method based on the system includes the following steps: acquiring manufacturing data through a data acquisition module; performing hierarchical feature decomposition on the manufacturing data based on the data analysis module to obtain multi-layer feature data; generating a high-dimensional feature matrix based on a complex feature mapping and nonlinear fusion algorithm; progressively predicting and feeding back to correct the high-dimensional feature matrix; obtaining anomaly detection values ​​based on adaptive anomaly deviation detection and aggregation analysis; and dynamically scheduling manufacturing tasks according to an optimized priority sequence.

[0026] Compared with the prior art, the beneficial effects of the present invention are reflected in the following aspects:

[0027] The system captures key data from the physical manufacturing process in real time through a data acquisition module. A data analysis module then performs multi-level decomposition and high-dimensional feature fusion to extract crucial manufacturing characteristics. Finally, a visualization module displays this data in a virtual manufacturing scenario, achieving precise mapping of the manufacturing state. Through the coordinated control of a real-time control module and an early warning module, the system makes timely adjustments upon detecting abnormal states, ensuring the safety and stability of the manufacturing process.

[0028] The technical architecture of this invention employs a unique multi-module collaborative mode, enabling efficient transmission of data and information flows between modules and generating synergistic effects. For example, the data acquisition module provides sufficient and accurate real-time data to the data analysis module, which in turn extracts key information from the data through feature decomposition and high-dimensional fusion algorithms, providing a reliable input basis for real-time control and task scheduling. Furthermore, the feature data generated by the data analysis module can be transformed into an intuitive virtual manufacturing scenario by the visualization module, allowing operators to monitor the status changes of the manufacturing process in real time. This multi-module collaborative structure not only solves the shortcomings of existing technologies where a single module works independently and lacks flexible adjustment, but also improves the overall efficiency and intelligence level of the manufacturing system through the complementary and synergistic effects between modules.

[0029] This invention effectively resolves the contradiction between task scheduling and anomaly handling in manufacturing systems. In traditional systems, the interdependence between priority scheduling and anomaly detection makes it difficult to simultaneously meet the demands for efficient scheduling and timely response. This invention, by introducing high-dimensional feature fusion and progressive dynamic prediction mechanisms into the data analysis module, enables the system to adjust task priorities in real time based on accurate predictions, ensuring efficient scheduling while rapidly responding to anomalies. Furthermore, the system's feedback correction mechanism adaptively corrects data deviations during dynamic prediction, making system control and prediction more accurate and fully resolving the technical contradictions arising from dynamic changes and environmental complexity in the manufacturing process.

[0030] Through the aforementioned innovative system architecture and collaborative mechanism, this invention demonstrates significant beneficial effects in the practical application of intelligent manufacturing. First, the introduction of multi-layer feature decomposition and high-dimensional feature fusion enhances the depth of data analysis, enabling the system to accurately identify various key features in the manufacturing process and ensuring rapid response to dynamic tasks and complex data. Second, the collaborative work of the real-time control module and the early warning module achieves rapid task priority scheduling and anomaly response, improving the system's responsiveness and production flexibility. Especially in scenarios with intensive tasks and high production pressure, this invention can significantly improve production efficiency. Furthermore, through the digital twin scenario of the visualization module, system operators can more intuitively monitor and grasp changes in various indicators during the production process, effectively improving management efficiency and decision-making quality.

[0031] In summary, while overcoming the shortcomings of the prior art, this invention achieves breakthroughs in the real-time performance, flexibility, and intelligence of the manufacturing system through the collaborative operation of various modules, providing the manufacturing industry with a highly intelligent and precise solution. Attached Figure Description

[0032] Figure 1 This is a flowchart illustrating the overall system flow of the present invention.

[0033] Figure 2 This is a logic block diagram of the data acquisition module of the present invention.

[0034] Figure 3 This is a logic block diagram of the data analysis module of the present invention.

[0035] Figure 4 This is a logic block diagram of the visualization module of the present invention.

[0036] Figure 5 This is a logic block diagram of the real-time control module of the present invention.

[0037] Figure 6 This is a logic block diagram of the early warning module of the present invention. Detailed Implementation

[0038] The solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. All other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of the present invention.

[0039] like Figure 1-6 As shown, this invention provides an intelligent manufacturing system and method based on digital twins, which achieves precise control, real-time monitoring and dynamic optimization of intelligent manufacturing through the coordinated operation of a series of modules and algorithms.

[0040] This invention proposes an intelligent manufacturing system based on digital twins. The system includes a data acquisition module 101, a data analysis module 102, a visualization module 103, a real-time control module 104, and an early warning module 105. First, the data acquisition module 101 is responsible for collecting various raw data during the manufacturing process, such as real-time data on equipment status, temperature, humidity, and pressure. This data can be obtained through sensors, image recognition systems, or on-site monitoring equipment, and is preferably processed by hierarchical encoding according to the acquisition frequency. By configuring the data acquisition module 101, the system can effectively improve the quality of the raw data, providing stable input for the subsequent analysis modules. For example, in temperature data acquisition, the acquisition frequency can be set to 1 second to ensure that the acquired data has sufficient temporal resolution, guaranteeing the ability to capture real-time changes in subsequent steps.

[0041] The data analysis module 102 is closely integrated with the data acquisition module 101, and is used to receive and process the raw data transmitted by the data acquisition module. This module performs feature extraction and in-depth analysis on the data based on a series of algorithms to generate characteristic data of the manufacturing process. The visualization module 103 is connected to the data analysis module 102, and its core function is to construct a virtual manufacturing scenario based on the analysis results, providing a real-time display of the manufacturing status. For example, in a production line equipment monitoring scenario, the visualization module 103 displays the current status of each machine in real time, such as operating frequency and temperature fluctuations, providing operators with intuitive status information.

[0042] After receiving the feature data generated by the visualization module 103, the real-time control module 104 dynamically adjusts the manufacturing process based on this data. Specifically, the real-time control module 104 adjusts the robot's operating parameters according to data changes to optimize the production process. The early warning module 105 monitors the feedback information from the real-time control module, ensuring that an alarm is issued when a potential anomaly is detected or a set threshold is exceeded. This early warning threshold can be flexibly adjusted according to the actual application scenario; for example, the system will automatically activate an early warning when the equipment temperature exceeds 80 degrees Celsius for timely intervention.

[0043] In one embodiment of the present invention, the data analysis module 102 further includes an adaptive multi-level feature decomposition unit, which is used to decompose manufacturing data hierarchically to generate multi-level feature data. For example, by setting appropriate hierarchical parameters and weights, the adaptive multi-level feature decomposition unit can decompose the data layer by layer according to the feature importance and complexity of the data, thereby constructing a clear multi-dimensional data structure. Specifically, the decomposed feature layer data... It can be obtained through the following formula:

[0044]

[0045] In this formula, F i ,j (L) Let φ be the decomposition feature of the i,j unit in the Lth layer, and φ be the adaptive decomposition function used to decompose the input data layer by layer. The input features are the i-th and j-th units of the previous layer. Preferably, the decomposition parameter μ is set to 0.5, indicating that the importance of the features is appropriately preserved in the data decomposition; the hierarchical weight σ is dynamically adjusted to 1.2 according to the needs of the scenario to adapt to different data structures and further improve the responsiveness and accuracy of the system in complex environments.

[0046] By setting up this decomposition unit, the system can effectively extract core feature data from the manufacturing process, providing a data foundation for complex feature analysis and multi-level feature combination in downstream modules.

[0047] In the data analysis module 102, a complex feature mapping and nonlinear fusion unit is preferably provided. This unit is connected to the output of the aforementioned adaptive multi-level feature decomposition unit and is used to further fuse the decomposed multi-level feature data into a high-dimensional feature matrix. This unit generates a high-dimensional feature matrix M through various feature mapping functions and weight combinations. fusion It can capture complex nonlinear relationships in the manufacturing process. Specifically, the fusion formula is as follows:

[0048]

[0049] Here, M fusion For the final high-dimensional feature matrix, ψ k This is a feature fusion weighting factor used to adjust the correlation strength between different features. This is a non-linear mapping function that projects the features from each decomposed layer into a higher-dimensional space to enhance the cross-correlation of features. Preferably, the feature weights ψ... k A value of 0.8 can be used to ensure that important features receive higher priority during the fusion process. This high-dimensional feature matrix serves as input to the recursive prediction model, ensuring that the manufacturing system can capture critical data in the rapidly changing manufacturing environment.

[0050] To achieve dynamic prediction of the manufacturing process, this invention further incorporates a progressive dynamic prediction and feedback correction unit in the data analysis module 102. This unit utilizes a high-dimensional feature matrix M. fusion This enables dynamic prediction of manufacturing status and optimizes the prediction results using a feedback mechanism. The core of the prediction algorithm lies in recursively adjusting the prediction based on feedback from historical data and current feature data. The specific calculation formula is as follows:

[0051] P t+1 =γ·η(P t M fusion )+κ·Ω(F real -P t )

[0052] Among them, P t+1 Here, γ is the predicted value for the next moment, γ is the recursive adjustment factor used to balance the weights of historical and current predictions, preferably set to 0.7 to ensure an appropriate balance between long-term trends and short-term fluctuations; η is the recursive prediction function responsible for combining the high-dimensional feature matrix with historical data to predict the next state; κ is the feedback coefficient used to control the feedback strength, preferably 0.5; and Ω is the feedback correction function, which improves the prediction accuracy of the system by correcting the difference between the current actual value and the predicted value.

[0053] This prediction module helps to control the manufacturing process precisely, especially in dynamically changing scenarios, where the system can respond and adjust in a timely manner to ensure the stability of the manufacturing process.

[0054] The present invention further includes an adaptive anomaly deviation detection and aggregation analysis unit in the data analysis module 102, used to further determine whether the deviation in the manufacturing process exceeds a preset threshold based on recursive prediction, thereby deciding whether to issue a warning signal. The anomaly detection formula is:

[0055]

[0056] Among them, A detect δ is the current anomaly detection value, an adaptive adjustment factor used to adjust the sensitivity of the deviation, with a preferred value of 1.5 to ensure sensitivity to abnormal changes; ρ is the deviation amplification factor, preferably set to 2, so that abnormal deviations are sufficiently amplified when they significantly exceed the threshold; ω j Θ is the weighting factor in the aggregation analysis, used to control the importance of multidimensional anomaly information. It is usually taken between 0.4 and 0.6 to balance the influence of various factors; Θ is the aggregation function, which performs comprehensive analysis on multiple anomaly signals.

[0057] Preferably, in this embodiment, the threshold is set to when A detect An alert is triggered when the value exceeds 2.5. This setting is based on the actual monitoring needs in the manufacturing process, which can effectively reduce false alarms and ensure timely feedback when there are real anomalies.

[0058] Through the above modular design and algorithm structure, the intelligent manufacturing system and method based on digital twins of the present invention realizes the synergistic effect between the modules, enabling the system to achieve high-precision data analysis and dynamic control in complex manufacturing environments, as well as issue reliable early warnings in abnormal situations, thereby ensuring the intelligence and stability of the entire manufacturing process.

[0059] In a preferred embodiment of the present invention, the real-time control module 104 includes a task optimization priority scheduling and feedback control unit. This unit receives the anomaly detection value A output from the data analysis module 102. detect and predicted value P t+1 Based on this data, the priority of manufacturing tasks is dynamically determined to ensure the timely completion of critical tasks. Task prioritization is calculated using the following formula.

[0060]

[0061] In this formula, O priority Let α represent the optimized task priority sequence. iis a weighting factor used to adjust the sensitivity of task priority to real-time anomaly detection and prediction results. It is usually preferred to take a value between 0.5 and 0.7 to achieve a balance between normal and abnormal states; p is a task scheduling function responsible for prioritizing tasks based on the current anomaly detection value and prediction results.

[0062] For example, when A detect When the priority exceeds the set threshold of 3.0, the system will increase the priority of critical tasks to respond quickly to anomalies in the manufacturing process. This dynamic priority adjustment significantly improves the adaptability and flexibility of the manufacturing system, enabling it to dynamically adjust the task sequence according to changes in the manufacturing environment, achieving efficient and stable production control.

[0063] Preferably, the visualization module 103 further includes a unit for constructing a virtual manufacturing scene, which is based on hierarchical feature data. and high-dimensional feature matrix M fusion Virtual manufacturing scenarios are generated to display key states and characteristics of the manufacturing process in real time within a digital twin environment. The visualization module transforms complex manufacturing data into intuitive graphics or images, helping operators quickly identify key nodes, task priorities, and abnormal conditions in the production process, thereby gaining a better understanding of the system's current operating status.

[0064] For example, the virtual manufacturing scenario can intuitively present the status and corresponding parameters of different equipment on the production line, including the equipment's operating temperature and frequency. Especially during dynamic task scheduling, when the real-time control module 104 adjusts task priorities, the visualization module 103 can display the priority changes of each task in the virtual scenario in real time, thus making the operator's understanding of the system adjustment process clear at a glance. Through this design, the virtual manufacturing scenario effectively improves the operator's understanding and real-time control of the manufacturing process.

[0065] In one specific embodiment of the present invention, the early warning module 105 is further configured to be based on the abnormal detection value A. detect And the optimized task priority sequence O priority Determine abnormal states in the manufacturing process. When the early warning module 105 detects an abnormal state (e.g., A...), it will detect the abnormal state. detect When the abnormal threshold (4.0) is exceeded, the system will issue an early warning signal and send an adjustment command to the real-time control module to ensure that the system can react quickly in abnormal conditions.

[0066] The threshold settings for this early warning module can be adjusted according to different needs in the manufacturing process. For example, in highly precise manufacturing processes, A detectThe threshold can be set to 3.5 to detect potential problems at an early stage. In a more relaxed process flow, the threshold can be increased to 4.5 to reduce unnecessary warning signals and improve the system's sensitivity. This flexible warning mechanism allows the invention to adapt to various manufacturing scenarios, providing strong assurance for production stability and safety.

[0067] In another preferred embodiment of the present invention, the data acquisition module 101 not only acquires real-time data during the manufacturing process, but also integrates industrial image acquisition equipment, sensors, and other devices. These devices collect data from different dimensions through a distributed data acquisition network, and after being encoded and processed by the data acquisition module 101, the data is sent to the data analysis module 102.

[0068] For example, in a complex automated production line, the data acquisition module can be set to capture images once per second to ensure accurate images of each machine's movement. Sensors can collect environmental data, such as temperature, humidity, and equipment pressure, to ensure high accuracy and real-time performance. This data is encoded into a specific format to meet the needs of the data analysis module, laying the foundation for subsequent data processing and feature extraction. Through this design, the data acquisition module ensures a stable and accurate data source for the manufacturing process, providing effective data support for the realization of a digital twin system.

[0069] This invention also provides a smart manufacturing method based on digital twins, comprising the following main steps:

[0070] First, the data acquisition module 101 collects raw manufacturing data and encodes it. This data includes, but is not limited to, ambient temperature, humidity, equipment pressure, and industrial image data, to ensure that the system obtains comprehensive manufacturing information.

[0071] Secondly, the manufacturing data is subjected to hierarchical feature decomposition by the data analysis module 102 to obtain multi-level feature data. During the feature decomposition process, multi-level decomposition is performed according to different data types, so that each feature of the data is extracted step by step, ensuring the accuracy of subsequent processing.

[0072] Based on complex feature mapping and nonlinear fusion algorithms, the data analysis module 102 integrates the decomposed multi-layer feature data into a high-dimensional feature matrix M. fusion This allows us to capture the complex nonlinear relationships in the manufacturing process.

[0073] Next, in the dynamic prediction and feedback correction module, the system uses the regression function and feedback correction algorithm to dynamically predict the high-dimensional feature matrix and obtain the predicted value P. t+1 And adjust the control parameters of the manufacturing process in real time.

[0074] In the adaptive anomaly detection and aggregation analysis module, anomaly detection value A is obtained by adaptively detecting the difference between predicted values ​​and actual data. detect It is used to determine deviations in the manufacturing process and to perform aggregation analysis.

[0075] Finally, based on the optimized task priority sequence, the system dynamically schedules manufacturing tasks to ensure that the manufacturing process proceeds in priority order, maximizing production efficiency and controlling quality.

[0076] Through the above implementation steps, this invention provides a complete intelligent manufacturing process based on digital twins. From data acquisition and analysis to final anomaly detection and scheduling optimization, each step is progressive and interconnected, forming a highly intelligent and flexible manufacturing system. In application, the system can monitor and adjust the manufacturing process in real time to achieve refined management of complex manufacturing processes, improve manufacturing efficiency, and ensure product quality. The superiority of this invention is verified through embodiments and comparative examples. The test data is based on the innovative points of the multi-dimensional intelligent manufacturing system of this invention. The test indicators are set around the core innovative points such as the system's real-time control efficiency, anomaly detection accuracy, task priority scheduling response speed, and data fusion accuracy, aiming to comprehensively demonstrate the superior performance of this invention. The following are detailed experimental procedures, testing methods, test results, and analysis of specific embodiments and comparative examples.

[0077] Example: The intelligent manufacturing system of the present invention, based on the synergistic effect of the data acquisition module, data analysis module, visualization module, real-time control module and early warning module, and combined with innovative algorithms such as hierarchical feature decomposition, high-dimensional feature mapping and fusion, dynamic prediction and feedback correction algorithm, abnormal deviation detection and aggregation analysis, and task priority optimization scheduling, is applied to an intelligent production line to verify its superiority in dynamic scheduling and abnormal response.

[0078] Comparative Example: Traditional intelligent manufacturing systems employ single-layer feature analysis models, lacking high-dimensional feature fusion and dynamic feedback prediction mechanisms. Task scheduling is primarily based on fixed priority schemes, lacking flexible real-time control and high-precision anomaly detection methods. The comparative example systems are applied on the same intelligent production line to facilitate performance index comparison.

[0079] The testing indicators, standards, and methods are as follows:

[0080] 1. Real-time control efficiency: Evaluate the real-time control response speed of the system under different production task loads. The detection method is based on the frequency of task switching per unit time, and the standard is that the task switching reaches the priority requirement set by the system within 5 seconds.

[0081] 2. Anomaly detection accuracy: The anomaly response sensitivity and accuracy of the detection system are simulated to detect manufacturing anomalies (such as a sudden rise in equipment temperature). The standard setting is that the correct detection rate within 1 second after the anomaly occurs reaches more than 95%.

[0082] 3. Task priority scheduling response speed: measures the speed at which the system responds to priority changes in a multi-tasking environment. The detection method is the response delay after the task scheduling command is issued, and the standard is that the priority switching response time is less than 1 second.

[0083] 4. Data fusion accuracy: During feature fusion, the accuracy of the fused data is evaluated through data verification methods (such as fitting tests based on the original data), with the standard set at a fusion error of less than 2%.

[0084] The table below shows the test results of the examples and comparative examples on the above indicators. The results show that the examples are significantly better than the comparative examples.

[0085] Real-time control efficiency (seconds) 0.7 3.2 Anomaly detection accuracy (%) 98.5 76.3 Task priority scheduling response time (seconds) 0.5 2.1 Data fusion accuracy (error %) 1.2 4.5

[0086] Regarding real-time control efficiency, the average response time for task switching in the embodiment is 0.7 seconds, significantly lower than the 3.2 seconds in the comparative example. This indicates that, supported by efficient data decomposition and dynamic prediction mechanisms, the present invention can quickly adapt to real-time changes in tasks during the manufacturing process, optimizing production efficiency. This data reflects the synergistic effect of the combination of multi-layer feature analysis, dynamic prediction correction, and priority scheduling in the present invention, ensuring precise scheduling of manufacturing tasks.

[0087] Regarding anomaly detection accuracy, the anomaly response accuracy of the embodiment is 98.5%, while that of the comparative system is 76.3%. This significant difference illustrates the advantages of the high-dimensional feature fusion and adaptive detection algorithm in the embodiment in complex manufacturing environments. By quickly identifying and accurately responding to abnormal states in the manufacturing process, it effectively reduces the possibility of false alarms and false negatives, thereby improving the safety and reliability of the system.

[0088] In the task priority scheduling response speed test, the average response latency of the embodiment system was 0.5 seconds, far lower than the 2.1 seconds of the comparative example. This result demonstrates the significant improvement brought by the real-time task optimization scheduling algorithm in this invention, especially under multi-tasking conditions, it can quickly complete priority switching and ensure the priority execution of critical tasks in the manufacturing process.

[0089] Regarding data fusion accuracy, the example system has an error of 1.2%, while the comparative system has an error of 4.5%. This result demonstrates that the high-dimensional feature mapping and nonlinear fusion methods employed in the example system significantly improve the accuracy of data fusion, thereby providing high-quality input data for subsequent system analysis and control.

[0090] In summary, the comparison shows that the embodiments of the present invention outperform the traditional comparative examples in all key indicators. These test results verify the superiority of the present invention in intelligent manufacturing environments. Through innovative data analysis and dynamic scheduling schemes, significant improvements in real-time performance, accuracy, and security are achieved, further demonstrating the application value of the present invention in intelligent manufacturing systems.

[0091] The above description is merely a preferred embodiment of the present invention; however, the scope of protection of the present invention is not limited thereto; any substitutions or modifications made by those skilled in the art within the scope disclosed in the present invention, based on the scheme and improved concept of the present invention, should be covered within the scope of protection of the present invention.

Claims

1. A digital twin-based intelligent manufacturing system, characterized in that, The system includes a data acquisition module, a data analysis module, a visualization module, a real-time control module, and an early warning module. The data acquisition module collects manufacturing data from the manufacturing system. The data analysis module is connected to the data acquisition module to receive the manufacturing data and process it into feature data. The visualization module is connected to the data analysis module to receive the feature data and construct a virtual manufacturing scenario. The real-time control module performs dynamic task control based on the feature data. The early warning module is connected to the real-time control module to receive abnormal information and send adjustment commands to the real-time control module. The various modules in the intelligent manufacturing system form a synergistic closed loop through data and information flows. The data analysis module further includes an adaptive multi-level feature decomposition unit for performing hierarchical feature decomposition on manufacturing data, used to generate multi-level feature data to obtain decomposed feature layer data. The decomposed feature layer data Defined by the following formula: in, Let Φ be the decomposition feature of the i,j unit in the Lth layer, and Φ be the adaptive decomposition function used to layer the features of the manufacturing data. The input features are the i-th and j-th units of the previous layer, μ is the decomposition parameter representing the importance of the features, and σ is the hierarchical weight used to dynamically adjust the hierarchical ratio between features. The adaptive multi-layer feature decomposition unit transmits multi-layer feature data to subsequent modules to form the basis of the data stream.

2. The intelligent manufacturing system based on digital twins according to claim 1, characterized in that, The data analysis module further includes a complex feature mapping and nonlinear fusion unit, which receives the decomposed feature layer data. The high-dimensional feature matrix M is generated using the following formula. fusion : Among them, M fusion For the fused high-dimensional feature matrix, ψ k The feature fusion weighting factor is dynamically adjusted based on the relevance of the features. The nonlinear feature mapping function projects hierarchical data into a high-dimensional space, where the high-dimensional feature matrix M... fusion As input to the prediction unit for subsequent analysis.

3. The intelligent manufacturing system based on digital twins according to claim 2, characterized in that, The data analysis module further includes a progressive dynamic prediction and feedback correction unit, which is based on a high-dimensional feature matrix M. fusion The current manufacturing status is dynamically predicted, and corrections are made based on feedback information. The predicted value P is obtained using the following formula. t+1 : P t+1 =γ·η(P t ,M fusion )+κ·Ω(F real -P t ) Among them, P t+1 The predicted value for the next moment is given by γ, which is a recursive adjustment factor that controls the weight of historical predictions. η is a recursive prediction function that performs dynamic prediction based on historical predictions and a high-dimensional feature matrix. κ is a feedback coefficient used to correct prediction bias, and Ω is a feedback correction function that adjusts the actual value F of the current manufacturing state. real Compared with the predicted value P t The predicted value P is compared and corrected. t+1 It serves as input to the anomaly detection unit for anomaly analysis.

4. The intelligent manufacturing system based on digital twins according to claim 3, characterized in that, The data analysis module further includes an adaptive anomaly detection and aggregation analysis unit, used to perform analysis based on the predicted value P. t+1 And actual manufacturing data F real Obtain abnormal deviation detection value A detect The detected value is defined by the following formula: Among them, A detect δ is the current anomaly detection value, ρ is the adaptive adjustment factor used to regulate the sensitivity of anomaly deviation, ρ is the deviation amplification factor used to enhance the sensitivity of anomaly deviation detection, and ω is the current anomaly detection value. j Θ is the anomaly deviation weighting factor, used to adjust weights according to the anomaly type; Θ is the anomaly aggregation function, used to aggregate and analyze multidimensional anomaly information; and the anomaly deviation detection value A is... detect Used for dynamic priority decision-making in the task scheduling unit.

5. The intelligent manufacturing system based on digital twins according to claim 4, characterized in that, The real-time control module further includes a task optimization priority scheduling and feedback control unit, used to optimize the abnormal deviation detection value A. detect and predicted value P t+1 Obtain the optimized priority sequence O priority The priority sequence is defined by the following formula: Among them, O priority For the optimized priority sequence, α i τ is the priority weighting factor, dynamically adjusted according to the urgency of the task; τ is the task priority scheduling function, based on task type T. i The abnormal detection values ​​and predicted values ​​are used to sort the tasks. The priority sequence is used to adjust the robot operation order in real time to ensure that the manufacturing tasks are executed according to priority.

6. The intelligent manufacturing system based on digital twins according to claim 5, characterized in that, The visualization module is further used for hierarchical feature data. and high-dimensional feature matrix M fusion A virtual manufacturing scenario is generated, which is used to display current manufacturing information in real time and to visualize future predictions.

7. The intelligent manufacturing system based on digital twins according to claim 6, characterized in that, The early warning module is further used to base on the abnormal detection value A detect and the optimized priority sequence O priority It identifies abnormal states in the manufacturing process and sends adjustment instructions to the real-time control module when an abnormal state is detected, so as to adjust the operation tasks in the manufacturing process.

8. The intelligent manufacturing system based on digital twins according to claim 7, characterized in that, The data acquisition module is further used to acquire real-time data in the manufacturing environment through industrial image acquisition equipment and sensors, and to encode the real-time data and send it to the data analysis module.

9. A digital twin-based intelligent manufacturing method based on the system described in any one of claims 1-8, characterized in that, Includes the following steps: Manufacturing data is collected through the data acquisition module. The manufacturing data is then subjected to hierarchical feature decomposition by the data analysis module to obtain multi-layer feature data. A high-dimensional feature matrix is ​​generated based on a complex feature mapping and nonlinear fusion algorithm. Dynamic prediction is performed and feedback is used to correct the high-dimensional feature matrix. Anomaly detection values ​​are obtained based on adaptive anomaly deviation detection and aggregation analysis. Manufacturing tasks are then dynamically scheduled according to an optimized priority sequence.

Images