Predecessor sub-packing whole-process management method and system based on digital twinning

Abstract

The application relates to the technical field of industrial process control and intelligent manufacturing, in particular to a precursor dispensing whole-process management method and system based on digital twinning; the method comprises the following steps: synchronously collecting and time-correcting equipment operation, material flow and environment state through a multi-source sensing unit, and constructing a multi-dimensional dispensing state space model; a digital twin is established, precise expression and credibility evaluation of a virtual model to real dispensing behavior are realized through behavior mapping and a dynamic evolution model; unobservable disturbances are quantified through multi-scale residual decomposition, an influence path of the disturbances on the dispensing behavior is analyzed by using an asymmetric propagation model, and behavior changes are mapped to quality deviation for dynamic prediction; active intervention is realized by combining trajectory consistency evaluation and abnormal evolution inhibition, and the system has self-learning and self-adaptive capabilities through online error feedback and process knowledge base updating; the application realizes closed-loop evolution from passive response to active predictive control.

Description

Technical Field

[0001] This invention relates to the field of industrial process control and intelligent manufacturing technology, specifically to a method and system for the whole-process control of precursor packaging based on digital twins. Background Technology

[0002] Precursor materials, as key raw materials in high-end manufacturing, have a significant impact on the quality of downstream products due to the precision and stability of their packaging process. Precursor packaging involves the coordinated operation of multiple stages, including silo storage, valve control, pipeline transportation, and packaging positioning. A complex coupling relationship exists between equipment actuators, material flow characteristics, and the production environment. In actual production, dynamic changes in equipment response performance, fluctuations in material flowability, and environmental factors such as temperature and humidity can all affect packaging stability. With the rapid development of intelligent manufacturing technology, the industrial sector has placed higher demands on the visualization and refined management of production processes. Digital twin technology, as a means to achieve real-time interactive mapping between physical entities and virtual models, shows broad application prospects in industrial process monitoring and optimization control.

[0003] Chinese invention patent application CN121414112A discloses a collaborative status monitoring method for the entire tape production line based on digital twins, including the following steps: S1: Starting with the process monitoring requirements, deploying IoT sensors and vision systems to build a full-process perception network for the tape production line, collecting data through an industrial gateway, and performing preliminary processing; S2: Modeling equipment based on the processed multidimensional data, and establishing a production process twin model by combining Petri nets to achieve process parameter-driven simulation and linkage with the MES system; S3: Constructing an AI analysis unit to perform status analysis based on the processed multidimensional data, including equipment health prediction and section risk detection analysis; S4: Integrating the AI ​​analysis results with the twin to generate a multi-angle status risk view.

[0004] Applying digital twin technology to the precursor packaging process, through real-time perception and fusion analysis of the multi-dimensional states of equipment, materials, and environment, to construct a virtual model that accurately reflects the evolutionary characteristics of packaging behavior, and on this basis to achieve forward-looking prediction and control of packaging quality, has become an important research direction for improving the intelligence level of the precursor packaging process. Summary of the Invention

[0005] The purpose of this invention is to address the problems existing in the background technology by proposing a method and system for full-process control of precursor packaging based on digital twins.

[0006] The technical solution of this invention: a method for end-to-end control of precursor packaging based on digital twins, comprising the following specific implementation steps: S1. By simultaneously collecting and correcting the equipment operation status, material flow status and environmental disturbance status during the precursor packaging process from multiple sources, dynamic feature extraction and state coupling analysis are performed to construct a multi-dimensional packaging state space model and a multi-dimensional packaging state space and vector model. S2. Based on the multi-dimensional assembly state space model, an assembly behavior mapping model is constructed to transform multi-source state data into assembly behavior feature quantities, and an assembly behavior dynamic evolution model is established to predict future behavior trends. Through virtual behavior deviation calculation and model parameter adaptive correction, the digital twin is made to approximate the real assembly process, and a twin credibility assessment mechanism is introduced to quantitatively determine the reliability of the model. S3. Based on the deviation between the predicted behavior and the actual behavior of digital twins, a multi-scale residual decomposition mechanism is constructed to identify implicit perturbations and form a comprehensive perturbation intensity index. The influence path of perturbations on the packaging behavior is analyzed through a state-driven asymmetric propagation model. The behavior changes are mapped to the packaging quality offset for dynamic prediction, and a feedforward control strategy is generated in combination with a comprehensive risk assessment. S4. Construct a collaborative target trajectory for packaging behavior and quality and define the feasible region of the trajectory. Achieve dynamic control of the entire process through trajectory projection and rolling optimization. Calculate the consistency between the actual trajectory and the target trajectory in real time and identify abnormal evolution trends for proactive intervention. At the same time, use the entire process operation data to perform online error feedback correction on the digital twin model and update the process knowledge base.

[0007] Preferably, the specific process of multi-source synchronous acquisition and time correction in step S1 is as follows: Multi-dimensional status perception of packaging equipment, material conveying system and packaging environment on precursor packaging production line; The equipment control system interface collects the dispensing valve opening, valve response time, conveying pipeline pressure, hopper discharge speed, and equipment vibration amplitude as equipment operating parameters. The precursor particle flow velocity, bin density change, and particle angle of repose change are collected as material flow parameters. The temperature, humidity, and micro-vibrations generated by the equipment operation in the packaging area are collected as environmental condition parameters. By using a unified time mapping mechanism to perform time correction on the sampled data from different devices, the data from different sensing devices are mapped to a unified time axis, forming a raw data sequence of the packaging process with a consistent time scale.

[0008] Preferably, the specific process of dynamic feature extraction and state coupling analysis in step S1 is as follows: After obtaining unified time series data, dynamic feature extraction of various state variables is performed using a sliding time window. The rate of change of equipment parameters, material parameters and environmental parameters within a time window is calculated, and the raw monitoring data is transformed into a dynamic response characteristic sequence that can reflect the changing trend of the packaging behavior. By calculating the coupling degree index between different state variables, the interaction between equipment operating status, material flow status and environmental status is analyzed, key disturbance channels that have a significant impact on the stability of dispensing are identified, and the degree of correlation between the main disturbance factors and dispensing behavior is determined.

[0009] Preferably, the specific process of constructing the multidimensional distributed state space and vector model in step S1 is as follows: After identifying key state variables, the equipment operating parameters, material flow characteristic parameters, and environmental state parameters are uniformly mapped to the state space of the packaging process. Define a dispensing process state vector that includes equipment operating state sub-vectors, material flow state sub-vectors, and environmental state sub-vectors: The equipment operating status subvector should include at least valve opening degree, pressure, and vibration parameters; The material flow state sub-vector must include at least the parameters of velocity and density variation; The environmental state vector includes at least temperature, humidity, and micro-vibration parameters; Construct a multidimensional state vector model that can describe the overall operational characteristics of the packaging behavior, and complete the instantiation expression of the multidimensional packaging state space.

[0010] Preferably, the process of establishing the digital twin in step S2 is as follows: Based on the state space vector of the dispensing process, a dispensing behavior mapping model is constructed. The equipment operating state, material flow state and environmental state are structurally integrated and mapped into behavioral characteristic quantities such as dispensing flow stability, dispensing execution response efficiency, material conveying continuity, silo release uniformity and energy consumption changes during the dispensing process. The initial structure of the digital twin model of dispensing behavior is established. Based on the packaging behavior mapping model, a dynamic evolution model of packaging behavior is constructed. By introducing a time recursion mechanism, the evolution process of behavior state changes over time is described. The current behavior state is coupled with the packaging state space to calculate and predict the future trend of packaging behavior changes. By comparing the actual packaging behavior collected in real time with the predicted behavior of the digital twin model, the deviation of the virtual behavior is calculated, and the twin model parameters are adaptively adjusted based on the deviation results to make the model continuously approach the real packaging process.

[0011] Preferably, the specific process of introducing the twin credibility assessment mechanism in step S2 is as follows: After completing the twin model calibration, the virtual-real deviation is normalized by constructing a twin credibility evaluation index to quantitatively assess the reliability of the current digital twin model. Determine the model's running status based on the credibility results: When the credibility is insufficient, the model reconstruction and parameter optimization mechanism is triggered to ensure that the digital twin maintains high accuracy and stability in complex packaging environments.

[0012] Preferably, the specific process of constructing the multi-scale residual decomposition mechanism and asymmetric propagation model in step S3 is as follows: Based on the deviation between twin predicted behavior and actual behavior, short-term residuals and cumulative residuals are constructed. Through weighted fusion, a comprehensive perturbation intensity index is formed to characterize the influence of two types of perturbations during the packaging process: short-term sudden changes and slow accumulation. After obtaining the perturbation intensity, a state-driven asymmetric propagation model is constructed. By introducing a propagation coefficient that dynamically changes with the operating conditions, the transmission path of the perturbation between equipment, materials and behavior is analyzed. The mapping relationship of the perturbation's influence on the packaging behavior is established to reveal the differentiated action mechanism of different perturbations under different states.

[0013] Preferably, the specific process of mapping behavioral changes to packaging quality offset for dynamic prediction in step S3 is as follows: After obtaining the behavioral changes caused by perturbations, a multi-dimensional state model of the dispensing quality is constructed, and the behavioral layer changes are mapped into a multi-dimensional quality state vector that includes dispensing quantity deviation, dispensing uniformity and material density stability. By constructing a quality evolution model and combining it with a trend smoothing mechanism for dynamic prediction, continuous evolution analysis of the quality deviation in the packaging process can be achieved. Based on the quality prediction results, a comprehensive risk assessment index is constructed to perform weighted analysis on the quality deviation of each dimension. A constrained optimization model is introduced to generate control strategies including fine-tuning valve opening, adjusting dispensing cycle, regulating silo pressure, and starting environmental control.

[0014] Preferably, step S4 specifically includes: Construct a collaborative target trajectory for packaging behavior and quality, dynamically generate a reference trajectory by combining current batch characteristics, historical best operating data and real-time environmental status, define the trajectory feasible region composed of quality constraints, equipment constraints and behavior-quality coupling constraints, and introduce a risk-driven contraction mechanism; Based on the optimal control results, a trajectory projection mechanism is introduced to map the control quantity to the target trajectory constraint space. Combined with the rolling optimization strategy, the control sequence is dynamically updated step by step to ensure that the control execution meets the trajectory constraints. During the packaging process, the consistency index between the actual trajectory and the target trajectory is calculated in real time. Combined with the consistency change rate, potential abnormal evolution trends are identified and suppression strategies are automatically triggered to constrain and adjust key control variables. By utilizing full-process operation data, online error feedback correction is performed on the digital twin model. By continuously updating the model parameters and the relationship between perturbation propagation, and combining multiple batches of operation experience to form a process knowledge base, reverse optimization of the control strategy generation mechanism and trajectory construction rules is achieved.

[0015] The technical solution of this invention: A precursor packaging end-to-end control system based on digital twins, comprising: Memory; processor; A computer program stored in the memory and capable of running on the processor; The processor executes a computer program to implement the aforementioned method for controlling the entire process of precursor packaging based on digital twins.

[0016] Compared with the prior art, the above-mentioned technical solution of the present invention has the following beneficial technical effects: This invention designs a method and system for the full-process control of precursor packaging based on digital twins. By constructing a multi-dimensional packaging state space model, it achieves comprehensive perception and deep integration of equipment operation, material flow, and environmental disturbances, significantly improving the situational awareness capability for complex packaging processes. Based on the behavior mapping and dynamic evolution mechanism of digital twins, the virtual model can accurately reflect real packaging behavior and evaluate model credibility in real time, effectively enhancing the system's ability to track the evolution of the packaging process. By introducing multi-scale residual decomposition and asymmetric propagation models, it transforms unobservable implicit perturbations into quantifiable mass offset predictions, breaking through the limitations of traditional control methods for explicit disturbances. This invention overcomes the limitations of traditional methods, significantly improving the early identification and warning capabilities for abnormal evolution. It constructs a full-process trajectory constraint and rolling optimization control mechanism, elevating single-point control to trajectory-level dynamic regulation. Combined with risk-driven contraction and abnormal evolution suppression strategies, it achieves stable and precise proactive intervention while ensuring process boundaries, effectively improving the stability and quality consistency of the packaging process. Furthermore, this invention possesses online error feedback correction and process knowledge base self-learning capabilities, enabling continuous optimization of model parameters and control strategies based on multi-batch operating experience. This achieves a closed-loop evolution from passive response to proactive predictive control, significantly enhancing the system's adaptability to different operating conditions and its continuous optimization potential. Attached Figure Description

[0017] Figure 1 This is a flowchart of a method for controlling the entire process of precursor packaging based on digital twins, as proposed in this invention. Detailed Implementation

[0018] Example 1, as Figure 1 As shown, the present invention proposes a method for end-to-end control of precursor packaging based on digital twins, which includes the following specific implementation steps: S1. By synchronously collecting data from multiple sources on the equipment operation status, material flow status, and environmental disturbance status during the precursor packaging process, and performing dynamic feature extraction and state coupling analysis on the raw data under a unified time reference, key disturbance factors affecting packaging stability are further identified. Finally, a multi-dimensional packaging state space model capable of describing the overall characteristics of packaging behavior is constructed, thus providing a structured and dynamic input data foundation for subsequent digital twin behavior modeling. The specific implementation process is as follows: S11. Multi-source sensing units are deployed at key nodes of the packaging equipment to collect equipment operating parameters, material flow parameters, and environmental status parameters in real time. A unified time mapping mechanism is used to correct the time of the sampled data from different devices, mapping the data from different sensing devices to a unified time axis, thereby forming a raw data sequence of the packaging process with a consistent time scale. Specifically: On the precursor packaging production line, multi-dimensional status perception is carried out on the packaging equipment, material conveying system and packaging environment, and a unified time calibration system is established to form a raw data sequence of the packaging process with a unified time scale. In actual production environments, the precursor packaging process involves equipment such as silo storage systems, packaging control valves, conveying pipelines, and packaging container positioning devices, and is also affected by material characteristics and environmental conditions. Therefore, in this embodiment, multi-source sensing units are deployed at key nodes of the packaging equipment to synchronously collect the equipment operating status, material flow status, and environmental disturbance status during the packaging process. Specifically, the following equipment operating parameters are collected through the equipment control system interface: dispensing valve opening, valve response time, conveying pipeline pressure, hopper discharge speed, and equipment vibration amplitude. Collect material state information such as precursor particle flow velocity, silo density change, and particle repose angle change; at the same time, acquire environmental information such as temperature, humidity, and micro-vibration generated by equipment operation in the dispensing area. It should be noted that, for precursor systems containing solid particles, the change in particle packing angle serves as a key characteristic parameter reflecting material flowability. To ensure that data from different sources accurately reflects the packaging status at the same moment, a unified time mapping mechanism is introduced: ; in, Indicates the corrected unified timestamp; This represents the original sampling time recorded by the i-th acquisition device; Indicates the time correction amount; S12. After obtaining unified time series data, dynamic features of various state variables are extracted using a sliding time window. The rate of change of equipment parameters, material parameters, and environmental parameters within the time window is calculated, thereby transforming the original monitoring data into a dynamic response feature sequence that reflects the changing trend of the packaging behavior. This enables the system to identify changes in the operating trend during the packaging process and provides basic indicators for subsequent coupling relationship analysis. Specifically: After completing the synchronization of multi-source data, behavioral features are further extracted from the raw data to identify key dynamic indicators that can reflect the stability and flow state changes of the packaging process. Since many anomalies during precursor packaging are caused by parameter change trends, such as slower valve response or a gradual decrease in material flow rate, a sliding time window is used to calculate the rate of change of state variables, thereby obtaining a dynamic response quantity that can describe the changing trend of packaging behavior. ; This allows us to obtain the rate of change of equipment parameters, material parameters, and environmental parameters within a certain time range; for example: the rate of change of dispensing valve opening, the rate of change of material flow rate, the rate of change of silo pressure, and the rate of change of ambient humidity. in, This represents the dynamic response of the j-th type of state variable at time t; This represents the measured value of the state variable at time t; The state value at the start of the time window; Indicates the feature extraction time window; S13. After obtaining the dynamic response characteristics, the interaction between equipment operating status, material flow status, and environmental status is analyzed by calculating the coupling degree index between different state variables. This identifies key disturbance channels that significantly affect the stability of the dispensing process and determines the degree of correlation between the main disturbance factors and the dispensing behavior. Specifically: After obtaining the dynamic response characteristics, the interaction between equipment state, material state, and environmental state is further analyzed to identify key disturbance factors affecting the stability of the dispensing process; the degree of mutual influence is assessed by calculating the coupling degree between different state variables. ; in, This represents the degree of coupling between state variable m and state variable n; This represents the dynamic response of variable m at time t; This represents the dynamic response of variable n at time t; This represents the average response value of variable m; The average response of variable n is represented by N; N represents the number of data samples. It should be noted that the higher the coupling degree value, the stronger the correlation between the two state variables; for example, when the coupling degree between changes in ambient humidity and changes in material flow rate is high, it means that changes in humidity may directly affect the flowability of the precursor. S14. After identifying key state variables, the equipment operating parameters, material flow characteristic parameters, and environmental state parameters are uniformly mapped to the state space of the packaging process. A multi-dimensional state vector model that can describe the overall operating characteristics of the packaging behavior is constructed, so that the complex operating state of the packaging process can be expressed in a structured form, specifically: After identifying key state variables, the equipment operating status, material flow status, and environmental status are uniformly mapped to the state space of the packaging process, thereby constructing a state vector model that can describe the overall characteristics of the packaging behavior and completing the instantiation expression of the multi-dimensional packaging state space. Define the state vector for the packaging process: ; To elaborate further: ; in, Represents the state vector of the packaging process; Represents a subvector of the device's operating state; Represents the sub-vector of material flow state; Represents the environment state sub-vector; This indicates the equipment's operating parameters, such as valve opening, pressure, and vibration. This indicates material flow parameters, such as flow velocity and density changes. These represent environmental parameters, such as temperature, humidity, and micro-vibration; p, q, and r represent the number of the corresponding parameters.

[0019] S2. Based on the state space vector of the packaging process constructed in step S1, a packaging behavior mapping model is established to transform multi-source state data into behavioral state variables that can characterize the packaging operation features. On this basis, a dynamic evolution model of packaging behavior is constructed to predict the future state of the packaging process. Furthermore, through virtual behavior deviation calculation and model parameter adaptive correction mechanisms, the digital twin can continuously approximate the real packaging process. Simultaneously, a twin credibility assessment mechanism is introduced to quantitatively determine the reliability of the model operation, thus forming a closed-loop packaging behavior digital twin system of state-driven, behavior mapping, dynamic evolution, and self-correction. The specific implementation process is as follows: S21. Based on the dispensing state space vector output in step S1, construct a dispensing behavior mapping model. This model structurally integrates the equipment operating state, material flow state, and environmental state, mapping them to behavioral characteristics such as dispensing flow stability, conveying continuity, and response efficiency. This achieves the transformation from state representation to behavioral representation and establishes the initial structure of a digital twin model of dispensing behavior, enabling the virtual model to describe the actual dispensing operation characteristics. Specifically: After obtaining the state space vector of the packaging process constructed in step S1, a structured mapping relationship between the real packaging behavior and the virtual twin model is first established, thereby constructing the basic model of the digital twin of the packaging behavior. It should be noted that the precursor packaging process has obvious multi-factor coupling characteristics. For example, the action of the equipment actuator will directly affect the material conveying behavior, while the material flow characteristics will be affected by environmental humidity, temperature and other factors. The sub-packaging behavior mapping model describes the operational characteristics of the sub-packaging process, enabling the twin model to reflect the actual sub-packaging behavior; the sub-packaging behavior state is obtained by mapping the state space vector of the sub-packaging process. ; To elaborate further: ; The characteristics of the repackaging process include, but are not limited to: repackaging flow stability, repackaging execution response efficiency, material conveying continuity, silo release uniformity, and changes in energy consumption during the repackaging process. in, This represents the state vector of the packaging behavior; This represents a mapping function for the packaging behavior; Let k represent the i-th packaging behavior feature; k represents the number of behavior features. It should be noted that the packaging behavior is a mapping function. This tool is used to collaboratively integrate and nonlinearly reconstruct the equipment operating state, material flow state, and environmental disturbance state in the state space of the packaging process. By assigning weights, matching associations, and constraining trends to multi-source state variables, it transforms the original state information into behavioral feature quantities that can characterize the stability and execution characteristics of the packaging process. In the mapping process, it focuses on the coupling relationship between state variables and their impact on the continuity and uniformity of packaging. By introducing a dynamic response priority mechanism, state variables that are more sensitive to the impact of packaging results are given higher mapping weights, thereby achieving accurate expression and real-time reflection of packaging behavior. S22. Based on the packaging behavior mapping model, a dynamic evolution model of packaging behavior is constructed. This model introduces a time recursion mechanism to describe the evolution of the behavior state over time, coupling the current behavior state with the packaging state space for calculation. This predicts future trends in packaging behavior, enabling a dynamic characterization of the continuous operation characteristics of the packaging process. This allows the digital twin model to possess forward-looking analytical capabilities. Specifically: After completing the mapping of the packaging behavior, a dynamic evolution model of the packaging behavior is further established to simulate the behavioral change trend of the packaging process in the future time range. Since the equipment state, material state, and environmental state all change over time during the precursor packaging process, a dynamic model that can describe the behavioral change patterns is constructed. A packaging behavior evolution model is established through behavioral recursion. ; in, Indicates the state of the packaging behavior at the next moment; Indicates the time step; Represents a function that changes behavior; It should be noted that the behavior change function This describes the combined effects of changes in the operation of the dispensing equipment, changes in material flow, and environmental disturbances on dispensing behavior. For example, when the pressure in the silo decreases, the stability of the dispensing flow rate will gradually decrease; when the ambient humidity increases, the flowability of material particles may decrease, thus affecting the continuity of conveying; when the equipment vibration increases, the uniformity of silo release may fluctuate. It should be further explained that the behavior change function This function is used to characterize the comprehensive evolution trend of dispensing behavior under the combined effects of equipment operation changes, material flow characteristics, and environmental disturbances based on the current state. Its essence is a dynamic mapping of the results of the coupling effect of multi-source state variables. This function integrates the modeling of dispensing valve response characteristics, silo discharge inertia, particle interaction, and the influence of environmental humidity and vibration on flowability. It transforms the matching degree between equipment driving capability and material flowability into the rate of behavior change. At the same time, it introduces historical behavior trajectory as a reference to reflect process inertia and hysteresis effect, so that the change of dispensing behavior not only reflects the instantaneous state, but also reflects the evolution trend over continuous time, realizing a forward-looking description and controllable expression of the change of dispensing stability. S23. By comparing the actual packaging behavior collected in real time with the predicted behavior of the digital twin model, the deviation of the virtual behavior is calculated, and the twin model parameters are adaptively adjusted based on the deviation results to make the model continuously approach the real packaging process. Specifically: During the operation of the digital twin model, the differences between the model's predicted behavior and the actual assembly behavior are continuously compared to ensure that the virtual model accurately reflects the real system state; the model's accuracy is evaluated by calculating the virtual behavior deviation. ; When the deviation exceeds the preset threshold, the twin model parameter adjustment mechanism is activated to re-match the virtual model with the actual packaging process. The model parameter adjustment expression is as follows: ; in, This indicates a deviation from the expected behavior; This indicates the characteristics of actual repackaging behavior; This indicates the behavioral characteristics predicted by the twin model; Indicates the original model parameters; This represents the updated model parameters; Indicates the parameter update coefficients; Indicates the actual state of behavior; Indicates the state of twin-predicted behavior; S24. After completing the twin model calibration, a twin reliability evaluation index is constructed to normalize the virtual-real deviation, quantitatively assess the reliability of the current digital twin model, and determine the model's operating status based on the reliability results. When the reliability is insufficient, a model reconstruction or parameter optimization mechanism is triggered to ensure that the digital twin maintains high accuracy and stability in complex assembly environments. Specifically: After model calibration, the reliability of the digital twin model is further evaluated to determine whether it can reliably reflect the current packaging process status. The reliability of the virtual model is measured by calculating a twin reliability index. ; When the credibility index is higher than the set threshold, it indicates that the twin model can accurately reflect the actual packaging behavior; when the credibility is lower than the threshold, the model structure is further updated or the behavior mapping relationship is readjusted. in, Indicators representing the credibility of twins; This indicates the maximum allowable deviation.

[0020] S3. Based on the digital twin behavior prediction and credibility output formed in step S2, a closed-loop mechanism is constructed, driven by behavioral residuals, to identify implicit perturbations, analyze state correlation propagation, predict multidimensional quality mapping, and adjust risk constraints. This mechanism transforms unobservable perturbation factors into quantifiable quality deviation prediction results and further generates stable and executable control strategies. This achieves a shift from passive monitoring to proactive intervention, improving the stability and quality controllability of the packaging process. The specific implementation process is as follows: S31. Based on the deviation between the predicted behavior and the actual behavior of digital twins, a multi-scale residual decomposition mechanism is constructed to divide the overall deviation into short-term abrupt change components and long-term cumulative components. A comprehensive perturbation intensity index is then formed through weighted fusion to characterize the influence of different types of perturbations during the assembly process, thereby achieving refined identification and structured expression of implicit perturbations. Specifically: In obtaining twin predictive behavior With actual behavior Subsequently, the deviation between the two is decomposed into multiple scales to identify implicit perturbations at different time scales during the packaging process. Considering that perturbations have two characteristics, namely "short-term abrupt changes" and "slow accumulation", a dual-scale residual decomposition mechanism is introduced: ; Further construct the short-time residuals and cumulative residuals: ; ; Based on this, a comprehensive perturbation intensity index is defined: ; in, Represents the residual of the original behavior; Represents short-term abrupt change residuals (reflecting transient disturbances); The value represents the long-term cumulative residual (reflecting asymptotic disturbances); L represents the length of the time window. Indicates the overall perturbation intensity; and This represents the weighting coefficient (reflecting sensitivity to sudden / gradual disturbances); Let represent the short-time residual of the i-th behavioral feature; This represents the cumulative residual of the i-th behavioral feature; S32. After obtaining the perturbation intensity, a state-driven asymmetric propagation model is constructed to analyze the transmission path of the perturbation between equipment, materials, and behaviors. By introducing a propagation coefficient that dynamically changes with operating conditions, a mapping relationship between the perturbation and the packaging behavior is established, thereby revealing the differentiated action mechanism of different perturbations under different states, specifically: After obtaining the perturbation intensity, we further identify the propagation path of the perturbation between equipment, materials, and behavior, and construct an asymmetric response model of the perturbation effect. The propagation of disturbances during the packaging process exhibits obvious directionality and asymmetry. For example, vibration has a significant impact on flow rate, but flow rate, in turn, has a small impact on vibration; humidity has a strong impact on material flowability, but a weak reverse impact. Constructing an asymmetric propagation model: ; Further details: ; It should be noted that the propagation coefficient As conditions change, for example, in high humidity environments, the influence coefficient of humidity on flowability increases; under high load conditions, the impact of vibration on dispensing stability is enhanced. in, This indicates behavioral changes caused by perturbations; Represents the state-related propagation matrix; This represents the propagation coefficient from the j-th perturbation to the i-th action; Indicates the residual dimension; This represents the change in the i-th behavioral variable; This represents the j-th residual component; S33. Based on the behavioral changes caused by perturbations, a multi-dimensional state model of packaging quality is constructed. Behavioral changes are mapped to changes in quality indicators such as packaging quantity deviation, uniformity, and stability. A trend smoothing mechanism is then used for dynamic prediction, thereby enabling continuous evolution analysis of packaging quality deviations. Specifically: After obtaining the behavioral changes, a multi-dimensional prediction model for packaging quality is further established, mapping the behavioral changes to quality offsets and constructing a multi-dimensional quality state vector: This includes, but is not limited to, deviations in dispensing quantity. Uniformity of dispensing Material density stability ; Further develop a quality evolution model: ; To elaborate further: ; To enhance forecast stability, a trend smoothing term is introduced: ; in, Bv represents the current packaging quality state vector; Bv represents the behavior-to-quality mapping matrix, obtained by fitting historical data. This represents the weight of the impact of the j-th behavior on the i-th quality indicator; This represents the smoothed quality prediction value; This represents the smoothing coefficient, ranging from 0 to 1. S34. Based on the quality prediction results, a comprehensive risk assessment index is constructed. Weighted analysis of quality deviations across various dimensions is performed, and a constrained optimization model is introduced to generate control strategies. Under the premise of ensuring stable regulation, feedforward adjustments are made to key parameters such as valve opening and dispensing cycle time. Specifically: After obtaining the quality prediction results, a risk assessment is further conducted, and an actionable control strategy is generated. Define comprehensive risk indicators: ; Based on this, a constrained optimization control model is introduced: ; It should be noted that the optimization objectives are balanced: reducing quality risks and avoiding drastic fluctuations in control. The final control strategy includes: fine-tuning valve opening, adjusting dispensing cycle, regulating silo pressure, and initiating environmental control (such as dehumidification). in, This represents a comprehensive quality risk indicator; This represents the weight of the i-th quality indicator; represents the target value of the i-th quality indicator, which is derived from the process setting standard; U represents the control parameter vector. Indicates the optimal control parameters; Indicates the current control parameters; This represents the control smoothing weight coefficient.

[0021] S4. Based on the quality prediction results and optimal control strategy output in step S3, a trajectory constraint and rolling optimization control mechanism for the entire packaging process is constructed. By elevating single-moment control decisions to full-process trajectory-level management and control, trajectory consistency assessment and anomaly evolution suppression methods are introduced. Combined with the online self-evolution of the digital twin model and reverse updates of process knowledge, the packaging system achieves a closed-loop evolution from "passive response control" to "active predictive control + continuous self-optimization". The specific implementation process is as follows: S41. Construct a collaborative target trajectory for packaging behavior and quality, and dynamically generate a reference trajectory by combining the characteristics of the current batch, historical best operating data, and real-time environmental status. Simultaneously, define the feasible region of the trajectory, composed of quality constraints, equipment constraints, and behavior-quality coupling constraints, to ensure that subsequent control always operates within the range that satisfies the process and equipment boundary conditions. Specifically: Define the target trajectory: ; The trajectory is generated based on the fusion of three types of information: the current prediction deviation (from step 3); the historical best packaging curve (for similar materials / processes); and the current level of environmental disturbance (determined by the intensity of the disturbance I(t). For example, when a strong disturbance is detected, the trajectory will automatically slow down to avoid aggressive adjustment. Define the initial feasible region: ; During operation, risk-driven contraction is introduced: ; in, This represents the set of target trajectories throughout the entire packaging process; The target quality trajectory is generated from historical data of excellent batches and current prediction deviation corrections. The target behavior trajectory is derived by inversely calculating the optimal operating mode and quality trajectory of the equipment. This represents the initial feasible region, derived from the physical limits of the equipment and process constraints (such as maximum flow rate and pressure limit). Represents the dynamic feasible region; This represents a dynamic risk threshold, set according to the production stage (gradually decreasing over time), meaning: larger fluctuations are allowed in the early stages (to ensure adjustment space), and the threshold is gradually tightened in the later stages (to ensure stability and consistency). S42. Based on the optimal control result obtained in step S3, a trajectory projection mechanism is introduced to map the control quantity to the target trajectory constraint space. Combined with a rolling optimization strategy, the control sequence is dynamically updated step-by-step, ensuring that the control execution closely approximates the optimal solution while satisfying the trajectory constraints. Simultaneously, real-time feedback continuously corrects the future control direction, thereby achieving stable and continuous dynamic regulation throughout the entire process. Specifically: Constructing predictive control sequences: ; Set optimization goals: ; Incorporate constraints during the optimization process: Ensure that: control actions always remain within the feasible domain and do not result in unexecutable or dangerous controls; Only execute the current control variable: In the next moment: update the state and recalculate the entire sequence; in, This represents a control sequence for a future period of time. This represents the control quantity at the current moment, including but not limited to: valve opening, dispensing rate, and silo pressure, which directly affects the equipment execution layer; H represents the prediction time domain length (control step size), which is set according to the system response time; Indicates the future number The quality state of step S3 is predicted by the model. Indicates the future number Walking is the current state; Indicates the future number Step target trajectory; This represents the risk weighting coefficient; Indicates the actual amount of control that is executed; S43. During the assembly process, the consistency index between the actual trajectory and the target trajectory is calculated in real time, and the potential anomaly evolution trend is identified by combining the consistency change rate. When a continuous deviation or aggravation of deviation is detected, a suppression strategy is automatically triggered to constrain and adjust key control variables and strengthen stability control, thereby achieving early intervention before the anomaly becomes explicit. Specifically: Define trajectory deviation: ; Determine the evolutionary trend: ; like If so, the deviation is widening, meaning the risk is increasing; if If so, the system is converging; And based on the trajectory deviation and evolutionary trends Combination decision control strategy: For example, if the deviation is small and convergence is achieved, it is determined to be in a stable state, and normal control is performed; if the deviation is small and divergence is achieved, it is determined to be in a slight anomaly, and the control weight is increased; if the deviation is large and divergence is achieved, it is determined to be in a moderate anomaly, and key variables are forcibly constrained; if the deviation is extremely large, it is determined to be in a severe anomaly, and protection strategies (flow limiting / pause) are activated. in, Indicates the degree of deviation from the current trajectory; Indicates the rate of change of trajectory deviation; S44. Utilize full-process operational data to perform online error feedback correction on the digital twin model. By continuously updating model parameters and the relationship between perturbation propagation, and combining multiple batches of operational experience to form a process knowledge base, reverse optimization of the control strategy generation mechanism and trajectory construction rules is achieved, enabling the system to possess self-learning and adaptive capabilities. Specifically: Obtaining error Update model parameters: ; Regarding the key model parameter: perturbation propagation coefficient Quality and mapping weights The update adopts a condition-triggered approach: when a certain type of error persists, the weights of relevant paths are automatically increased and invalid paths are suppressed. Establish a process knowledge transfer mechanism and construct batch experience vectors: ; Used for: initializing models for new batches, and quickly matching similar working conditions; Further measures to prevent overfitting include setting an update threshold (no update for small errors) and introducing historical weight decay. in, Represents the error function of the twin model; Indicates the actual state of behavior; This indicates the behavior predicted by the twin model; Represents the set of parameters for the twin model; This represents the updated model parameters; Indicates the learning rate; Represents the gradient of the error function; Represents the batch experience vector; Indicates the average perturbation intensity; Indicates the average risk level; Indicates control mode characteristics; This indicates the final quality result.

[0022] Example 2: The present invention proposes a precursor packaging end-to-end control system based on digital twins, which is used to execute the precursor packaging end-to-end control method based on digital twins proposed in Example 1, comprising: Memory; processor; A computer program stored in the memory and capable of running on the processor; The processor executes a computer program to implement the precursor packaging process control method based on digital twins in the first embodiment described above.

[0023] The embodiments of the present invention have been described in detail above with reference to the accompanying drawings. However, the present invention is not limited thereto. Various changes can be made within the scope of knowledge possessed by those skilled in the art without departing from the spirit of the present invention.

Claims

1. A method for end-to-end control of precursor packaging based on digital twins, characterized in that, The specific implementation steps include the following: S1. By simultaneously collecting and correcting the equipment operation status, material flow status and environmental disturbance status during the precursor packaging process from multiple sources, dynamic feature extraction and state coupling analysis are performed to construct a multi-dimensional packaging state space and vector model. S2. Based on the multi-dimensional assembly state space model, an assembly behavior mapping model is constructed to transform multi-source state data into assembly behavior feature quantities, and an assembly behavior dynamic evolution model is established to predict future behavior trends. Through virtual behavior deviation calculation and model parameter adaptive correction, the digital twin is made to approximate the real assembly process, and a twin credibility assessment mechanism is introduced to quantitatively determine the reliability of the model. S3. Based on the deviation between the predicted behavior and the actual behavior of digital twins, a multi-scale residual decomposition mechanism is constructed to identify implicit perturbations and form a comprehensive perturbation intensity index. The influence path of perturbations on the packaging behavior is analyzed through a state-driven asymmetric propagation model. The behavior changes are mapped to the packaging quality offset for dynamic prediction, and a feedforward control strategy is generated in combination with a comprehensive risk assessment. S4. Construct a collaborative target trajectory for packaging behavior and quality and define the feasible region of the trajectory. Achieve dynamic control of the entire process through trajectory projection and rolling optimization. Calculate the consistency between the actual trajectory and the target trajectory in real time and identify abnormal evolution trends for proactive intervention. At the same time, use the entire process operation data to perform online error feedback correction on the digital twin model and update the process knowledge base.

2. The method for end-to-end control of precursor packaging based on digital twins according to claim 1, characterized in that, The specific process of multi-source synchronous acquisition and time correction in step S1 is as follows: Multi-dimensional status perception of packaging equipment, material conveying system and packaging environment on precursor packaging production line; The equipment control system interface collects the dispensing valve opening, valve response time, conveying pipeline pressure, hopper discharge speed, and equipment vibration amplitude as equipment operating parameters. The flow velocity of precursor particles, changes in silo density, and changes in particle angle of repose are collected as material flow parameters. The temperature, humidity, and micro-vibrations generated by the equipment operation in the packaging area are collected as environmental condition parameters. By using a unified time mapping mechanism to perform time correction on the sampled data from different devices, the data from different sensing devices are mapped to a unified time axis, forming a raw data sequence of the packaging process with a consistent time scale.

3. The method for end-to-end control of precursor packaging based on digital twins according to claim 2, characterized in that, The specific process of dynamic feature extraction and state coupling analysis in step S1 is as follows: After obtaining unified time series data, dynamic feature extraction of various state variables is performed using a sliding time window. The rate of change of equipment parameters, material parameters and environmental parameters within a time window is calculated, and the raw monitoring data is transformed into a dynamic response characteristic sequence that can reflect the changing trend of the packaging behavior. By calculating the coupling degree index between different state variables, the interaction between equipment operating status, material flow status and environmental status is analyzed, key disturbance channels that have a significant impact on the stability of dispensing are identified, and the degree of correlation between the main disturbance factors and dispensing behavior is determined.

4. The method for end-to-end control of precursor packaging based on digital twins according to claim 3, characterized in that, The specific process of constructing the multidimensional packaging state space and vector model in step S1 is as follows: After identifying key state variables, the equipment operating parameters, material flow characteristic parameters, and environmental state parameters are uniformly mapped to the state space of the packaging process. Define a dispensing process state vector that includes equipment operating state sub-vectors, material flow state sub-vectors, and environmental state sub-vectors: The equipment operating status subvector should include at least valve opening degree, pressure, and vibration parameters; The material flow state sub-vector must include at least the parameters of velocity and density variation; The environmental state subvector includes temperature, humidity, and micro-vibration parameters; Construct a multidimensional state vector model that can describe the overall operational characteristics of the packaging behavior, and complete the instantiation expression of the multidimensional packaging state space.

5. The method for end-to-end control of precursor packaging based on digital twins according to claim 4, characterized in that, The process of creating the digital twin in step S2 is as follows: Based on the state space vector of the dispensing process, a dispensing behavior mapping model is constructed. The equipment operating state, material flow state and environmental state are structurally integrated and mapped into behavioral characteristic quantities such as dispensing flow stability, dispensing execution response efficiency, material conveying continuity, silo release uniformity and energy consumption changes during the dispensing process. The initial structure of the digital twin model of dispensing behavior is established. Based on the packaging behavior mapping model, a dynamic evolution model of packaging behavior is constructed. By introducing a time recursion mechanism, the evolution process of behavior state changes over time is described. The current behavior state is coupled with the packaging state space to calculate and predict the future trend of packaging behavior changes. By comparing the actual packaging behavior collected in real time with the predicted behavior of the digital twin model, the deviation of the virtual behavior is calculated, and the twin model parameters are adaptively adjusted based on the deviation results to make the model continuously approach the real packaging process.

6. The method for end-to-end control of precursor packaging based on digital twins according to claim 5, characterized in that, The specific process of introducing the twin credibility assessment mechanism in step S2 is as follows: After completing the twin model calibration, the virtual-real deviation is normalized by constructing a twin credibility evaluation index to quantitatively assess the reliability of the current digital twin model. Determine the model's running status based on the credibility results: When the credibility is insufficient, the model reconstruction and parameter optimization mechanism is triggered to ensure that the digital twin maintains high accuracy and stability in complex packaging environments.

7. The method for end-to-end control of precursor packaging based on digital twins according to claim 6, characterized in that, The specific process of constructing the multi-scale residual decomposition mechanism and asymmetric propagation model in step S3 is as follows: Based on the deviation between twin predicted behavior and actual behavior, short-term residuals and cumulative residuals are constructed. Through weighted fusion, a comprehensive perturbation intensity index is formed to characterize the influence of two types of perturbations during the packaging process: short-term sudden changes and slow accumulation. After obtaining the perturbation intensity, a state-driven asymmetric propagation model is constructed. By introducing a propagation coefficient that dynamically changes with the operating conditions, the transmission path of the perturbation between equipment, materials and behavior is analyzed. The mapping relationship of the perturbation's influence on the packaging behavior is established to reveal the differentiated action mechanism of different perturbations under different states.

8. The method for end-to-end control of precursor packaging based on digital twins according to claim 7, characterized in that, The specific process of mapping behavioral changes to packaging quality offsets for dynamic prediction in step S3 is as follows: After obtaining the behavioral changes caused by perturbations, a multi-dimensional state model of the dispensing quality is constructed, and the behavioral layer changes are mapped into a multi-dimensional quality state vector that includes dispensing quantity deviation, dispensing uniformity and material density stability. By constructing a quality evolution model and combining it with a trend smoothing mechanism for dynamic prediction, continuous evolution analysis of the quality deviation in the packaging process can be achieved. Based on the quality prediction results, a comprehensive risk assessment index is constructed to perform weighted analysis on the quality deviation of each dimension. A constrained optimization model is introduced to generate control strategies including fine-tuning valve opening, adjusting dispensing cycle, regulating silo pressure, and starting environmental control.

9. The method for end-to-end control of precursor packaging based on digital twins according to claim 8, characterized in that, Step S4 specifically includes: Construct a collaborative target trajectory for packaging behavior and quality, dynamically generate a reference trajectory by combining current batch characteristics, historical best operating data and real-time environmental status, define the trajectory feasible region composed of quality constraints, equipment constraints and behavior-quality coupling constraints, and introduce a risk-driven contraction mechanism; Based on the optimal control results, a trajectory projection mechanism is introduced to map the control quantity to the target trajectory constraint space. Combined with the rolling optimization strategy, the control sequence is dynamically updated step by step to ensure that the control execution meets the trajectory constraints. During the packaging process, the consistency index between the actual trajectory and the target trajectory is calculated in real time. Combined with the consistency change rate, potential abnormal evolution trends are identified and suppression strategies are automatically triggered to constrain and adjust key control variables. By utilizing full-process operation data, online error feedback correction is performed on the digital twin model. By continuously updating the model parameters and the relationship between perturbation propagation, and combining multiple batches of operation experience to form a process knowledge base, reverse optimization of the control strategy generation mechanism and trajectory construction rules is achieved.

10. A precursor packaging end-to-end control system based on digital twins, characterized in that, include: Memory; processor; A computer program stored in the memory and capable of running on the processor; The processor executes a computer program to implement the precursor packaging process control method based on digital twins as described in any one of claims 1 to 9.

Images