Full-process Packaging Control System and Method Based on Digital Twin Model

Abstract

The present invention discloses a full-process packaging control system and method based on a digital twin model, which relates to the technical field of packaging management. Through the digital twin model and multi-source data fusion, a full-process closed-loop control system is formed from external demand perception to internal packaging process optimization; by integrating data such as order systems, market feedback, production plans, and equipment status, the problem of data islands in traditional packaging production is solved, and the real-time fusion of external demands and internal data is realized; a multi-source data weighting and cleaning fusion mechanism is adopted, and through digital twin simulation and evaluation, it is ensured that the packaging plan is closer to the actual production environment; through a multi-objective optimization algorithm, the conflicts and coupling effects between multiple objectives are accurately captured, improving the decision-making accuracy. Through rapid deployment and a closed-loop feedback mechanism, the production line is monitored and adjusted in real time, improving production efficiency and quality, reducing production costs, increasing efficiency and product quality, and at the same time reducing environmental impacts.

Description

Technical Field

[0001] The present invention relates to the technical field of packaging management, and specifically to a full-process packaging control system and method based on a digital twin model. Background Art

[0002] With the continuous improvement of the status of the packaging industry in the modern manufacturing system, more and more enterprises have begun to introduce data acquisition systems, industrial Internet of Things platforms, and digital twin models to achieve visual and optimized management of the packaging process. Traditional packaging lines usually adjust equipment parameters through experience or limited offline experiments. Under the conditions of single production volume and few changes, the production efficiency can still be maintained. However, in the face of the current order mode of multiple varieties and small batches, as well as the frequent changes in external market demands, such traditional methods are often difficult to keep up in a timely manner. Although some existing technologies can achieve basic monitoring of the operating states of equipment such as packaging machines and labeling machines, and also use certain algorithms for data analysis, there are still common limitations in dealing with multi-source data (such as external orders, material supply batches, environmental temperature, and internal quality inspection results), such as poor data docking or only single-dimensional analysis. As a result, production decisions lack real-time response to external changes, and the packaging process still mainly relies on "post-event adjustment", making it difficult to form an efficient and flexible packaging control system.

[0003] In the Chinese invention patent with the application publication number CN115809789A, a product packaging management system based on data analysis is disclosed, specifically as follows: obtaining product information to be packaged, sending the product information to an analysis and measurement module to analyze the product information to obtain analysis data, and measuring the product in the obtained analysis data to obtain measurement data; sending the measurement data to a product packaging production module, and the product packaging generation module generates product packaging information based on the measurement data and the analysis data; the packaging evaluation module obtains the product packaging information, conducts product evaluation based on the obtained product packaging information, and selects multiple groups of packaging price information with the highest evaluation; sending the packaging price information and the product packaging information to a material calculation module, and the material calculation module calculates the materials required for the product. The present invention obtains the information of the packaged product, generates the specification information of the product, generates the packaging box body by combining the specification information with the product weight information, and effectively manages the product packaging process.

[0004] However, in the actual application process, due to the lack of a unified dynamic integration mechanism between external demand information (such as new material batches or order specification changes) and internal production process data, it is difficult for the digital twin model to timely adjust the key parameters of each node, which in turn leads to problems such as too long switching cycles, rising abnormality rates, or excessive manual intervention on the packaging line. Specifically, when an enterprise receives materials from different batches or different suppliers, if the physical characteristics or producibility parameters are not updated in a timely manner in the digital twin system, it is easy to generate small but continuously accumulating deviations in key links such as labeling and sealing, which in turn affect the production line rhythm, packaging quality, and even the overall yield. At the same time, when making changes to packaging specifications or quickly launching new products, due to the lack of a global quantitative assessment of the actual working load and process constraints of production line equipment, it often takes multiple attempts or shutdown adjustments to gradually approach the stable state, significantly reducing the flexibility and efficiency of packaging production.

[0005] To this end, the present invention provides a full-process packaging control system and method based on a digital twin model. Summary of the Invention

[0006] (I) Technical Problems to be Solved

[0007] In view of the deficiencies of the prior art, the present invention provides a full-process packaging control system and method based on a digital twin model. Through the digital twin model and multi-source data fusion, a full-process closed-loop control system from external demand perception to internal packaging process optimization is formed; by integrating data such as order systems, market feedback, production plans, and equipment status, the data island problem in traditional packaging production is solved, and the real-time fusion of external demand and internal data is realized; a multi-source data weighting and cleaning fusion mechanism is adopted, and through digital twin simulation and evaluation, it is ensured that the packaging plan is closer to the actual production environment; through a multi-objective optimization algorithm, the conflicts and coupling effects between multiple objectives are accurately captured, and the decision-making accuracy is improved. Through rapid deployment and a closed-loop feedback mechanism, the production line is monitored and adjusted in real time, the production efficiency and quality are improved, and at the same time, the environmental impact is reduced. Thus, the technical problems recorded in the background art are solved.

[0008] (II) Technical Solutions

[0009] To achieve the above objectives, the present invention is realized through the following technical solutions:

[0010] A full-process packaging control method based on a digital twin model, including

[0011] collecting external demand data and internal packaging data through a unified interface, defining the credibility index of each data source, fusing each data source through a weighting algorithm, and generating a fused data set through data cleaning and preprocessing for subsequent packaging strategy simulation and multi-objective optimization;

[0012] After importing the fused dataset into the digital twin system, a packaging strategy simulation scenario is constructed. Under the simulation scenario, simulation analysis is carried out according to the scheme parameters. After generating the corresponding performance indicators, the effectiveness of different packaging schemes is evaluated by the constructed packaging collaborative efficiency evaluation function;

[0013] After determining the optimization objectives, each packaging scheme is comprehensively evaluated through a multi-objective optimization algorithm. The constructed comprehensive offset function captures the collaborative conflicts between multiple objectives, finds a balance point among multiple objectives, automatically screens out the optimal or sub-optimal scheme according to the optimization results, and generates specific adjustment suggestions;

[0014] After deploying the selected optimal or sub-optimal packaging scheme to the actual packaging production line, small-batch trial production is carried out. The production line data is collected and the deviation from the model prediction is detected. If an abnormal deviation occurs, the feedback mechanism is immediately triggered, and the deviation data is fed back to the optimization algorithm for model correction or scheme adjustment.

[0015] Furthermore, external demand data and internal packaging data are collected. There is a data source set , where , and each represents a type of data source. The reliability index of each data source is defined . Based on the reliability index , the normalized weighted coefficient is calculated and the initial dataset of weighted fusion is obtained . The weighting formula is set as follows:

[0016] ;

[0017] In the formula: is the reliability index of the data source, , is the non-linear deviation index of the th data source, is the preprocessing function performed on the original data.

[0018] Furthermore, outlier detection and consistency correction are performed on the output initial dataset of weighted fusion . A multi-attribute offset function is defined to measure the degree of deviation of the i-th record from the reference pattern, as follows:

[0019] ;

[0020] In the formula: is the actual value of the th record on the th key attribute. The The benchmark distribution center value of a key attribute, is the total number of attribute dimensions; is the overall sensitivity amplification factor, The non - linear deviation index of each individual attribute; is the trade - off coefficient of the attribute collaborative offset term, is the non - linear index of the collaborative offset term;

[0021] If exceeds the given deviation threshold, it is determined that the data of this record is abnormal, and an appropriate correction method is selected according to the business scenario and subsequent usage requirements.

[0022] Furthermore, the dataset after anomaly filtering and correction is normalized in format, and finally a fused dataset is output , for the input interface and model requirements of the target digital twin system, perform position mapping, data type conversion, and timestamp alignment on the data fields, and name the fused dataset after format conversion as .

[0023] Furthermore, load the fused dataset into the digital twin environment, and define the equipment models, process flow structures, and adjustable packaging scheme parameter sets involved in the actual production line;

[0024] Through the equipment description module of the digital twin system, associate the equipment structure and dynamic characteristics so that it can call the equipment status fields in the fused dataset ; register the unique identifier of all workstations in the digital twin system ;

[0025] Introduce the packaging scheme parameters to be evaluated, denoted as , each corresponds to a packaging parameter combination, and pre - configure adjustable machine parameters in the digital twin system that match .

[0026] Furthermore, perform simulations on each packaging scheme parameter in the digital twin environment in sequence to obtain the performance indicators of each key workstation under the packaging scheme, and define the following packaging collaborative efficiency evaluation function :

[0027] ;

[0028] Among them: represents the offset relative to the reference process parameters set for workstation under scheme , is the workstation Weight coefficient in the overall packaging process is the work station nonlinear amplification factor is the attribute collaboration item trade-off coefficient is the exponential term of collaborative offset is the work station number is the total number of work stations

[0029] Packaging collaboration effectiveness evaluation function The lower the value of the smaller the collaborative deviation of the solution under the current working conditions

[0030] For all packaging solutions summarize the simulation results generated; for each packaging solution generated results and performance indicators are integrated to form a simulation result table and provide a visual display for key indicators, marking special situations or abnormal scenarios

[0031] Furthermore, extract the target dimensions that meet the enterprise's requirements from the simulation result table and define them as the input of the multi-objective optimization problem, including four target dimensions, where represents the average quality deviation index of solution ; represents the total cost per unit output of solution ; represents the total production cycle or changeover time of solution at the specified output; represents the environmental impact index of solution ;

[0032] Define the comprehensive offset function to quantify the overall performance of solution under the four target dimensions and their interactive effects

[0033] ;

[0034] In the formula: is the reference value or expected value of target ; is the weight factor of target in the overall evaluation is the single-objective nonlinear deviation index is the amplification coefficient of the collaborative offset term is the nonlinear index of the multi-objective interaction term; when the comprehensive offset function has a smaller value, it indicates that the packaging solution The overall is closer to the enterprise's ideal goal And the multi-objective interaction conflict is less than expected. Otherwise, it indicates that there is a comprehensive deviation greater than expected.

[0035] Furthermore, execute a multi-objective optimization algorithm to search for a set of candidate optimal solutions. An improved evolutionary algorithm, mixed integer programming, or other high-performance solvers can be used; after the iteration is completed, a series of approximate Pareto front solutions are output, that is, a set of solutions that achieve a relative balance among multiple objective dimensions;

[0036] Screen and visualize the output multi-objective candidate solutions, and finally form a list of optimal or sub-optimal solutions, specifically as follows: For the parameter set of all packaging solutions as candidate solutions Sort them in ascending order according to their comprehensive deviation function values, and mark the top several solutions with the best performance;

[0037] If the enterprise has a strong preference for a certain objective, it can be re-screened based on the objective importance factor or reference value to narrow the range of the solution set available for deployment; in the visualization view, highlight the solutions that have met the key constraint conditions, and record the finally selected optimal or sub-optimal solutions as , and attach the corresponding numerical indicators and recommended parameter adjustment information to form a decision list .

[0038] Furthermore, quickly deploy the selected packaging solution , including: Write the packaging materials, label designs, size specifications, and corresponding machine adjustment parameters in the optimal or sub-optimal solution into the control system of the actual packaging production line according to the established mapping relationship; register its corresponding hardware interface mark ;

[0039] Load the optimal or sub-optimal solution into the production line in the form of a configuration file or container image; use a unified parameter management module to inject the core variables of the optimal or sub-optimal solution into each work station ; Compare the hardware parameters with the digital twin model parameters for consistency, and enter the trial production state after confirming no conflicts.

[0040] Furthermore, conduct small-batch production on the deployed packaging solution and conduct real-time monitoring through the digital twin system, and collect the deviation information between the actual output and the model prediction. Among them, set the trial production batch to start the actual operation of the production line, and install high-frequency data acquisition devices at each work station to transmit the process data back to the digital twin system in real time;

[0041] Introduce a dynamic offset increment function For the th real-time record, evaluate at time : If exceeds the predetermined threshold, it is determined as an abnormal deviation, triggering an alarm or an automatic fallback strategy, and storing all real-time deviations , that is, the deviation detection results are stored in the intermediate result table .

[0042] Furthermore, introduce a dynamic offset increment function , and the evaluation method for the th real-time record at time is as follows:

[0043] ;

[0044] In the formula: represents a certain quality index of the th product at time , which is the corresponding value predicted by the digital twin model, representing the process variable related to the product, is the model prediction value; and respectively control the relative weights of single-index deviation and multi-index collaborative deviation; is the number of indicators; and are non-linear amplification factors.

[0045] Furthermore, for the recorded abnormal deviations found, compare the corresponding process data with the input features of the digital twin model to identify possible sources of errors; according to the deviation distribution, adaptively update relevant model parameters or external process assumptions to reduce future prediction errors;

[0046] If the deviation is severe enough to affect the feasibility of the plan, then return the newly collected data to the multi-objective optimization algorithm, supplement and update the decision space, and use the corrected digital twin model to recalculate the comprehensive offset for the plan or other possible plans to determine whether to switch or fine-tune the current deployment plan; if the deviation after the corrected trial production has returned to the acceptable range, set it as the default mass production plan.

[0047] The full-process packaging control system based on the digital twin model includes,

[0048] The data acquisition and fusion module collects external requirement data and internal packaging data through a unified interface, defines the reliability indicators of each data source, fuses each data source through a weighted algorithm, and generates a fused data set after data cleaning and preprocessing , for subsequent packaging strategy simulation and multi-objective optimization;

[0049] The packaging strategy simulation module imports the fused data set into the digital twin system to build a packaging strategy simulation scenario. Under the simulation scenario, simulation analysis is carried out according to the scheme parameters. After generating the corresponding performance indicators, the effectiveness of different packaging schemes is evaluated by the constructed packaging collaborative efficiency evaluation function;

[0050] The multi-objective optimization module, after determining the optimization objectives, comprehensively evaluates each packaging scheme through a multi-objective optimization algorithm. The constructed comprehensive offset function captures the collaborative conflicts between multiple objectives, finds a balance point between multiple objectives, automatically screens out the optimal or sub-optimal scheme according to the optimization results, and generates specific adjustment suggestions;

[0051] The scheme deployment and verification module deploys the selected optimal or sub-optimal packaging scheme to the actual packaging production line for small-batch trial production, collects production line data and detects the deviation from the model prediction. If an abnormal deviation occurs, the feedback mechanism is immediately triggered, and the deviation data is fed back to the optimization algorithm for model correction or scheme adjustment.

[0052] (III) Beneficial effects

[0053] The present invention provides a full-process packaging control system and method based on a digital twin model, having the following beneficial effects:

[0054] Through the digital twin model and multi-source data fusion, a full-process closed-loop control system from external demand perception to internal packaging process optimization is formed. By effectively integrating different data sources (such as order systems, market feedback, production plans, and equipment status, etc.), the data island problem in traditional packaging production is solved, and the real-time fusion of external demand and internal packaging data is realized, providing accurate data support for packaging strategy simulation, evaluation, and optimization decisions.

[0055] Adopting a multi-source data weighting and cleaning fusion mechanism, different data sources are weighted to ensure that high-quality and time-sensitive data participates in decision-making preferentially. By standardizing the data and unifying the time stamps, the consistency between different data sources is ensured; by introducing the fused data set into the digital twin system for packaging scheme simulation and evaluation, and by introducing a packaging collaborative efficiency evaluation function, the collaborative effects of multiple workstations and multiple dimensions are incorporated into the evaluation, ensuring that the simulation results are closer to the actual production environment, and improving the feasibility and accuracy of the scheme. By considering the collaborative effects between workstations, the limitations of single-workstation optimization are avoided.

[0056] Comprehensively evaluate the packaging plan through a multi-objective optimization algorithm to solve the limitation of only considering a single objective (such as cost or quality) in traditional methods. By introducing a non-linear deviation measure and a collaborative offset term, the optimization algorithm can accurately capture the conflicts and coupling effects between multiple objectives, improving the flexibility and accuracy of the optimization decision-making, and finding a reasonable balance among quality, cost, production efficiency, and environmental impact.

[0057] The optimized plan is applied to the actual production line through a rapid deployment and closed-loop feedback mechanism. Through small-batch trial production and real-time monitoring, it can detect and correct the deviation from the model prediction in real time, ensure the flexible adjustment of the packaging plan, ensure the rapid response and continuous optimization of the production line, and improve production efficiency and quality.

[0058] The present invention solves various problems in packaging production through data fusion, digital twin simulation, multi-objective optimization, and a closed-loop feedback mechanism, provides an innovative packaging control solution, reduces production costs, improves production efficiency and product quality, and at the same time reduces environmental impact. BRIEF DESCRIPTION OF THE DRAWINGS

[0059] Figure 1 It is a schematic flow chart of the full-process packaging control method of the present invention;

[0060] Figure 2 It is a schematic structural diagram of the full-process packaging control system of the present invention. DETAILED DESCRIPTION OF THE INVENTION

[0061] The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.

[0062] Please refer to Figure 1 , the present invention provides a full-process packaging control method based on a digital twin model, including:

[0063] Step 1: Collect external demand data and internal packaging data through a unified interface, define the credibility index of each data source, fuse each data source through a weighted algorithm, and generate a fused data set after data cleaning and preprocessing , for subsequent packaging strategy simulation and multi-objective optimization;

[0064] The content of the first step is as follows:

[0065] Step 101: Collect external demand data and internal packaging data, and weight each data source to reflect the reliability and timeliness of the data. There is a data source set , where , and each represents a type of data source (such as an order system, market feedback, production plan, equipment status, etc.); define the reliability index of each data source , which is used to characterize the data integrity and update frequency of the data source. Based on the reliability index calculate the normalized weighting coefficient and obtain the initial data set for weighted fusion , can be positive or negative, and the weighting formula is set as follows:

[0066] ;

[0067] In the formula: is the reliability index of the data source, , the larger the value, the more reliable the data or the more timely the update. It is an adjustable scale factor used to control the amplification degree of high-reliability data sources. is the normalized weighting coefficient of the data source; is the th non-linear deviation index of the data source, which is used to control the amplification degree of abnormal deviations of attributes, ;

[0068] is the preprocessing function performed on the original data (such as unifying timestamps, field mapping), which ensures the consistency of different data sources in terms of time series and field names. is the initial data set after weighted fusion;

[0069] When in use, by performing hierarchical weighting on the data sources, the negative impact of low-quality or lagged data on overall decision-making is reduced; by first performing basic standardization on the fields of each data source, data structure conflicts in subsequent steps are reduced, The form avoids using traditional statistical variance or standard deviation and other indicators, and has more flexible adjustability and non-linear amplification effects;

[0070] Step 102. Perform outlier detection and consistency correction on the initial data set for weighted fusion output in Step 101 To identify outlier records in multiple dimensions, define a multi-attribute offset function to measure the deviation degree of the i-th record from the reference pattern, as follows:

[0071] ;

[0072] In the formula: is the th record's actual value on the th key attribute, is the The reference mode or benchmark distribution center value of each key attribute (which can be obtained from business experience or prior models), is the total number of attribute dimensions;

[0073] is the overall sensitivity amplification factor, Used to control the overall amplification of the deviation accumulation. Separate nonlinear deviation index for each attribute, , The more sensitive the attribute, the It can be set larger to more obviously amplify the abnormality of the attribute; is the trade-off coefficient of the attribute collaborative offset term, Used to balance the contribution of single attribute bias and multi-attribute collaborative bias; is the nonlinear index of the cooperative offset term, The larger the value, the more it can amplify the impact of the simultaneous deviation of multiple attributes; multi-attribute collaborative items , used to capture linkage anomalies where multiple attributes deviate from the reference pattern simultaneously, and can identify complex outliers that are difficult to detect with a single attribute. When this value is large, it means that at least two or more attributes have significant deviations and should be given extra attention;

[0074] like If the deviation exceeds a given threshold, the data is deemed abnormal and needs to be corrected or removed. For abnormal data records, appropriate correction methods are selected based on the business scenario and subsequent usage requirements. For example, mapping corrections are performed according to the reference model, and severely abnormal records that cannot be corrected are directly removed.

[0075] When used, the quality and accuracy of the dataset are significantly improved through outlier detection and consistency correction. The introduction of multi-attribute offset functions can accurately detect outliers in multi-dimensional data, especially multi-attribute collaborative deviations, avoiding anomalies that cannot be detected by a single attribute. Classification and correction based on the size of the deviation avoids subsequent simulation errors or decision-making errors caused by data anomalies, providing high-quality input for the output of the final fused dataset.

[0076] Step 103: Normalize the format of the dataset after abnormal filtering and correction in step 102, and finally output the fused dataset , based on the input interface and model requirements of the target digital twin system, the data fields are mapped, data types are converted, and timestamps are aligned. This can be specifically divided into the following three aspects:

[0077] Field normalization: Align the names and orders of each record with the input elements required by the digital twin to avoid subsequent model ambiguity about the field meaning; Unified data type: For the possible discrete / continuous mixed fields, convert them into numerical or enumeration types that are convenient for the digital twin to parse; Timestamp recalibration: When the external demand data is inconsistent with the internal packaged data in the acquisition frequency or time base, achieve a unified time reference system through interpolation or truncation to support subsequent simulation and optimization;

[0078] Finally, name the fused dataset after format conversion as and clarify its metadata such as field definitions, time range, and value range that can be taken, so that in the next step, it can be directly read in the digital twin system and the packaging strategy simulation can be carried out;

[0079] When in use, the fused dataset after format conversion matches the interface and data model of the digital twin platform, reducing the additional processing cost of the docking link; Clearly define the meaning and available range of each field in the form of metadata to ensure that the same meaning interpretation can be maintained when called in subsequent steps. Through the finally output fused dataset realize the unified expression from external demands to internal packaged data, laying a foundation for the strategy simulation and multi-objective optimization in the next step;

[0080] After completing the weighted and preliminary fusion of multi-source data, the initial dataset is obtained; Apply the above formula to each record of , identify and quantify the deviation degree of each record, and then decide to retain, correct or remove. After the processing is completed, it can enter the format conversion and final fusion output of step 103. Mark the evaluation value and whether it has been corrected or removed in the output dataset to ensure traceability and consistency for the next step (i.e., the packaging strategy simulation of the digital twin system).

[0081] Step 2. After importing the fused dataset into the digital twin system, construct a packaging strategy simulation scenario. Under the simulation scenario, conduct simulation analysis according to the scheme parameters. After generating the corresponding performance indicators, evaluate the effectiveness of different packaging schemes by the constructed packaging collaborative efficiency evaluation function;

[0082] The said step 2 includes the following contents:

[0083] Step 201. In step 201, first, the fused dataset output in step 1 Load it into the digital twin environment; based on this environment, define the equipment models, process flow structures, and adjustable packaging scheme parameter sets involved in the actual production line; through the equipment description module of the digital twin system, associate the structures and dynamic characteristics of packaging machines, labeling machines, conveyor devices, etc., so that it can call the fusion data set The equipment status fields related to the equipment operation status, production capacity limit, etc. in the ; to ensure connection with the simulation process in the subsequent steps, it is necessary to register the unique identifiers of all workstations or key components in the digital twin system here ;

[0084] Introduce a set of packaging scheme parameters to be evaluated, denoted as , each corresponds to a combination of packaging parameters such as possible packaging materials, label styles, size specifications, etc. Pre-configure the adjustable machine parameters (such as encapsulation temperature, labeling accuracy, conveyor speed, etc.) in the digital twin system that match to support one-by-one loading, switching, and running during subsequent simulations;

[0085] When in use, it can ensure that the digital twin model is consistent with the actual production line configuration, reduce the modeling deviation of subsequent simulations, and through unified workstation [[ID=I8]]and marking information, which can be directly called in the subsequent steps without secondary matching or repeated conversion, forming a complete set of equipment-process-parameter set foundation for parallel simulation of multiple packaging strategies;

[0086] Step 202. For each packaging scheme parameter defined in Step 201 Perform discrete event or hybrid-driven simulations in the digital twin environment in sequence, aiming to obtain performance indicators such as production efficiency, quality indicators, and switching time at each key workstation under the packaging scheme. To comprehensively measure the pros and cons of the scheme, the following packaging collaboration effectiveness evaluation function can be defined :

[0087] ;

[0088] Among them: represents the offset relative to the reference process parameters set for workstation under scheme , which can be dynamically calculated by the digital twin according to the actual equipment characteristics or production capacity limit during the simulation process; is the weight coefficient of workstation in the overall packaging process, , used to reflect its relative importance to the output quality or beat; is the non-linear amplification factor of workstation , with a value greater than 0, determining the sensitivity to the offset of a single workstation; is the trade-off coefficient of attribute collaboration items, usually taking positive values or 0, and is used to control the linkage amplification effect when deviations occur simultaneously at multiple workstations; is the exponential term of collaborative offset, and the larger its value, the more the influence of multi-station coupling deviation will be amplified: is the workstation number; is the total number of workstations;

[0089] Among them, the packaging collaborative efficiency evaluation function with a lower numerical value indicates that the solution has a smaller collaborative deviation under the current working conditions, and the overall execution efficiency or quality potential is higher;

[0090] It should be noted that: when performing the simulation, the following procedures need to be followed:

[0091] Scheme loading: Send the parameters corresponding to the current scheme to each workstation module in the digital twin system;

[0092] Scheduling execution: Simulate the production process, including process queuing, processing cycle, transfer waiting, etc., and record the actual output quantity, duration, and quality inspection information;

[0093] Index extraction: Calculate the such as offset indexes for each workstation, and substitute them into the packaging collaborative efficiency evaluation function for numerical solution, and other auxiliary indexes (such as defect rate, unit energy consumption, etc.) can be recorded at the same time;

[0094] When in use, the multi-station coupling influence is incorporated into the same evaluation function through the packaging collaborative efficiency evaluation function , which is beneficial to quickly identify efficiency bottlenecks or quality shortboards on complex production lines; the introduction of non-linear exponential amplification and the collaborative item can amplify the potential risks brought by the synchronous offset of key workstations and avoid only focusing on single-point optimization: after execution, each packaging scheme will correspond to a set of clear simulation results and offset indexes, providing a basis for the next-step visualization and summary analysis;

[0095] Step 203. Summarize the simulation results (including numerical values and other auxiliary indexes) generated by all packaging schemes ; Integrate the results, performance indexes such as defect rate, switching duration, and energy consumption generated by each packaging scheme to form a result table ;

[0096] For key indexes (such as , production, and quality pass rate), provide visual displays such as line charts, histograms, or heat maps, enabling subsequent decision-making processes to quickly identify advantageous and disadvantageous solutions; mark special situations or abnormal scenarios (such as station downtime, excessive deviation), facilitating their use as references or for exclusion during multi-objective trade-offs in the next stage.

[0097] During use, a complete mapping is formed from the scheme parameters to the core evaluation value and then to the detailed station indicators , which can be comprehensively weighed for subsequent decision-making. By means of visualization, the understanding threshold is reduced, the identification efficiency of key anomalies or station bottlenecks is improved, and index inputs verified through simulation are provided for the multi-objective optimization process, avoiding blind optimization directly on the original data.

[0098] Step 3: After determining the optimization objectives, comprehensively evaluate each packaging scheme through a multi-objective optimization algorithm. The constructed comprehensive deviation function captures the collaborative conflicts among multiple objectives, finds a balance point among multiple objectives, automatically selects the optimal or sub-optimal solution according to the optimization results, and generates specific adjustment suggestions;

[0099] The content of the above Step 3 is as follows:

[0100] Step 301: First, extract the target dimensions that meet the enterprise's requirements from the simulation result table output in Step 2 , and clarify them as the input of the multi-objective optimization problem. To more flexibly describe the complex requirements in the packaging process, the main optimization objectives can be defined as four target dimensions: quality, cost, production efficiency, and environmental load, denoted respectively as: , , , , where:

[0101] represents the average quality deviation index of the scheme (such as a non-linear amplification form of the defect rate); represents the total cost per unit output of the scheme ; represents the total production cycle or changeover time of the scheme at the specified production volume; represents the environmental impact index of the scheme (such as a weighted measure of carbon emissions, energy consumption, etc.);

[0102] To better capture the collaborative trade-off effect in the multi-objective combined evaluation, define the comprehensive deviation function , to quantify the overall performance and interactive effects of the scheme under the four target dimensions:

[0103] ;

[0104] Where: Target Reference or expected value (which can be given by historical data or corporate benchmark strategy);

[0105] Target The weight factor in the overall evaluation, , is the single target nonlinear deviation index, A larger index indicates greater sensitivity to deviation from the target;

[0106] is the amplification factor of the cooperative offset term, Used to balance the influence of coupling effects among multiple objectives: is the nonlinear index of the multi-objective interaction term, Used to amplify or suppress the cumulative impact when multiple goals deviate from the ideal state at the same time;

[0107] pass and coproduct Achieve nonlinear amplification of large deviations, and thus more accurately evaluate multi-target conflicts. When the integrated deviation function The smaller the value, the better the packaging solution The overall goal is closer to the company's ideal and the multi-objective interaction conflict is small, otherwise it means there is a large comprehensive offset;

[0108] When used, it accurately describes the comprehensive performance of the packaging strategy in a multi-objective form, avoiding the one-sided optimality under a single indicator. The collaborative offset term captures the cumulative impact of multiple objectives deviating at the same time. This is more in line with the characteristics of multiple objectives restraining each other in real production. It clarifies the input source and symbol meaning, so that the optimization algorithm in the subsequent steps can accurately call the corresponding indicators.

[0109] Step 302: Based on the comprehensive offset function defined in step 301 , execute a multi-objective optimization algorithm to search for a set of candidate optimal solutions that take into account quality, cost, cycle time and environment. Improved evolutionary algorithm, mixed integer programming or other high-performance solvers can be used. The specific process is as follows:

[0110] According to the simulation results table , the feasible solution and its corresponding , , , The value is loaded into the initial solution set of the optimization algorithm; if the search space needs to be further expanded, it can be Fine-tune around the parameters (such as fine-tuning the sealing temperature, labeling speed, etc.) to generate a small number of new solutions and incorporate them into the initial solution set;

[0111] Set the number of iterations Or the convergence criterion. When the algorithm has not converged, continuously based on Evaluate and update the solution set. In evolutionary algorithms, operations such as crossover and mutation can be used to generate new solutions And according to Compare their advantages and disadvantages and select the better-performing solutions. Perform non-dominated sorting or multi-objective crowding distance calculation on the candidate solutions for each iteration to keep the candidate solutions diverse in the objective space. After multiple iterations, output a series of approximate Pareto front solutions, that is, a set of solutions that achieve a relative balance among multiple objective dimensions such as quality, cost, cycle, and environment. For the parameter set of each packaging solution Its ( ), ), ), ) numerical value and the final Evaluation value, which is convenient for visualization and ranking recommendation in the next step;

[0112] When in use, through systematic search and iteration, taking into account the conflicts and balances under multi-objective attributes, provide enterprises with multiple potential feasible solutions. Using methods such as non-dominated sorting helps to retain diverse high-quality solutions in the complex objective space, facilitating subsequent solution selection, and is closely connected to the simulation result table in Step 2 Data docking to achieve an organic connection from numerical simulation to global optimization;

[0113] Step 303: Screen and visualize the multi-objective candidate solutions output in Step 302, and finally form a list of optimal or sub-optimal solutions, specifically as follows:

[0114] For the parameter sets of all packaging solutions that are candidate solutions Sort them in ascending order according to their comprehensive offset function Value, and mark the top several solutions with the best performance; if the enterprise has a strong preference for a certain objective (such as particularly emphasizing low cost or high quality), it can be re-screened based on the objective importance factor Or reference value To narrow the range of the solution set available for deployment;

[0115] Select common multi-objective visualization methods (such as parallel coordinate plots, radar charts, or heat maps) to display in the solution set ( ), ), ), For the convenience of subsequent operations, the solutions that have met the key constraint conditions (such as the minimum quality deviation, the maximum production line utilization rate, etc.) can be marked in a highlighted manner in the visualization view, and the finally selected optimal or sub-optimal solution is recorded as , and attach the corresponding numerical indicators and recommended parameter adjustment information to form a decision list named . .

[0116] When in use, through the comprehensive sorting and visualization based on the comprehensive deviation function , excellent solutions that meet the production strategy can be quickly locked, and the clear decision list enables step four to directly execute the deployment, reducing repeated confirmation or data docking. The multi-dimensional visualization provides an intuitive understanding of the complex decision space, helps to detect potential risks in a timely manner and make effective trade-offs;

[0117] By defining the multi-objective comprehensive deviation function to quantify the deviation degree of the packaging scheme in multiple dimensions such as quality, cost, cycle, and environment, and introducing a collaborative deviation term in the form of a product to capture the linkage conflicts between the objectives, significantly improving the characterization accuracy of the coupling of multiple objectives; providing richer search space information for evolutionary algorithms or other solution methods; with the help of non-dominated sorting or other high-order optimization strategies, the distribution diversity of each solution in the objective space can be maintained on the premise of ensuring the core needs of the enterprise, thus generating a batch of alternative Pareto front solutions, and finally assisting managers to make decision-making suggestions that conform to the actual production scenario in a visual and sorted manner. By keeping the naming and field mapping consistent with the simulation result table , the data can be closely connected with the previous process within this step.

[0118] Step four: After deploying the selected optimal or sub-optimal packaging scheme to the actual packaging production line, conduct a small-batch trial production, collect production line data and detect the deviation from the model prediction. If an abnormal deviation occurs, immediately trigger the feedback mechanism, feedback the deviation data to the optimization algorithm, and perform model correction or scheme adjustment;

[0119] The said step four includes the following contents:

[0120] Step 401: According to the decision list output in step three, quickly deploy the selected packaging scheme , and the specific process includes: writing the packaging material, label design, size specification and corresponding machine adjustment parameters in the optimal or sub-optimal solution into the control system of the actual packaging production line according to the established mapping relationship; registering its corresponding hardware interface mark , and keep it in the deployment log for subsequent tracing;

[0121] If the digital twin system adopts a microservice architecture, the optimal or suboptimal solution can be Encapsulate it into a configuration file or container image and load it into the production line; use a unified parameter management module to manage the optimal or suboptimal solution. The core variables (such as temperature, pressure, labeling speed, etc.) are injected into each station at one time Ensure that the data fields correspond to those of the digital twin system; perform a consistency comparison between the hardware parameters and the digital twin model parameters. Once no conflicts are confirmed, the solution can be declared ready for trial production.

[0122] When used, implementation from decision list Efficient mapping to the actual parameters of the production line reduces the risk of repeated manual configuration, retains mapping logs and interface tags, and helps to quickly locate deployment problems in subsequent small-batch trial production. 3. Microservices or containerization mode makes the deployment process more flexible and more compatible;

[0123] Step 402: Deploy the packaging solution, i.e., the optimal or suboptimal solution. Conduct small-batch production and conduct real-time monitoring through the digital twin system to collect information on deviations between actual output and model predictions, providing a basis for subsequent closed-loop corrections;

[0124] Set trial production batch (usually much smaller than the official scale) to minimize the large-scale losses caused by immature solutions; start the actual operation of the production line and Install high-frequency data acquisition devices to transmit process data (quality inspection, output rhythm, etc.) back to the digital twin system in real time;

[0125] In order to identify the difference between actual production line performance and model prediction, a dynamic offset increment function is introduced For the first Real-time records at time Assessments will be conducted at:

[0126] ;

[0127] Where: Indicates the Products in time A certain quality indicator (such as sealing strength) of the product is predicted by the digital twin model, and the corresponding value represents the quality of the product. Process variables (such as temperature, viscosity, etc.), is the model prediction value; and Control the relative weights of single indicator deviation and multi-indicator collaborative deviation respectively; the values are all greater than 0, is the number of indicators;

[0128] and is the non - linear amplification coefficient, both of which take values greater than 0 and are used to significantly amplify large deviations;

[0129] If exceeds the predetermined threshold, it is determined as an abnormal deviation, triggering an alarm or an automatic fallback strategy, and all real - time deviations , that is, the deviation detection results are stored in the intermediate result table , with additional information such as timestamps and batch numbers, and the deviation distribution is dynamically displayed in the form of a trend chart or a heat map on the digital twin monitoring panel, facilitating timely intervention by manual or automatic systems;

[0130] During use, quickly verify the solution through small - batch trial production for its applicability in the real production environment, reducing large - scale risks, The function provides a quantitative index for monitoring the non - linear difference between the actual and the predicted, which is suitable for quickly discovering major quality or process deviations. Real - time visualization enables decision - makers to grasp the operation of the production line in the first place and provides an objective basis for subsequent callbacks or corrections;

[0131] Step 403: For the recorded abnormal deviation, compare the corresponding process data with the input features of the digital twin model to identify possible sources of errors (such as equipment aging, inconsistent material batches, abnormal environmental temperature, etc.); according to the deviation distribution, adaptively update relevant model parameters (such as small, etc. non - linear exponents) or external process assumptions (such as the upper limit of station production capacity) to reduce future prediction errors;

[0132] If the deviation is so severe that it affects the feasibility of the solution, then return the newly collected data to the multi - objective optimization algorithm in step three, supplement and update the decision space, and use the corrected digital twin model to recalculate the comprehensive offset for the solution or other possible solutions to determine whether to switch or fine - tune the current deployment plan; if the deviation after the corrected trial production

[0133] During use, the deviation information obtained through real-time monitoring is immediately applied to the digital twin model, enabling the model to maintain a high degree of consistency with the actual production line during continuous iteration. Once there is a problem with the deployment, it immediately returns to the optimization link to search again in an agile operation mode. Through multiple rounds of small-batch iteration, the packaging strategy that best matches the actual production environment can be found in the shortest time, greatly shortening the development cycle.

[0134] Through solution deployment - small-batch trial production - real-time monitoring - closed-loop feedback, a dynamic correction loop is formed that runs through the digital twin model and the actual packaging production line. Different from traditional static decision-making, a non-linear deviation measurement function and real-time data acquisition means are used to achieve precise detection and flexible response to deviations, and are highly coupled with the multi-objective optimization algorithm in the previous link. It can trigger an optimization search again based on newly emerging abnormal or deviation data, truly realizing adaptive iteration and continuous improvement.

[0135] Through the digital twin model and multi-source data fusion, a full-process closed-loop control system is formed from external demand perception to internal packaging process optimization. By effectively integrating different data sources (such as order systems, market feedback, production plans, and equipment status, etc.), the data island problem in traditional packaging production is solved, and the real-time fusion of external demands and internal packaging data is achieved, providing accurate data support for packaging strategy simulation, evaluation, and optimization decisions.

[0136] A multi-source data weighting and cleaning fusion mechanism is adopted to weight different data sources to ensure that high-quality and highly time-sensitive data participates in decision-making first. By standardizing the data and unifying the timestamps, the consistency between different data sources is ensured. By introducing the fusion data set into the digital twin system for packaging solution simulation and evaluation, and by introducing a packaging collaborative efficiency evaluation function to incorporate the collaborative effects of multiple workstations and multiple dimensions into the evaluation, the simulation results are ensured to be closer to the actual production environment, improving the feasibility and accuracy of the solution. By considering the collaborative effects between workstations, the limitations of single-workstation optimization are avoided.

[0137] By comprehensively evaluating packaging solutions through a multi-objective optimization algorithm, the limitation of only considering a single objective (such as cost or quality) in traditional methods is solved. By introducing non-linear deviation measures and collaborative offset terms, the optimization algorithm can accurately capture the conflicts and coupling effects between multiple objectives, enhancing the flexibility and accuracy of optimization decisions, and being able to find a reasonable balance among quality, cost, production efficiency, and environmental impact.

[0138] The optimized solution is applied to the actual production line through a rapid deployment and closed-loop feedback mechanism. Through small-batch trial production and real-time monitoring, the deviation from the model prediction can be detected and corrected in real time, ensuring the flexible adjustment of the packaging solution, the rapid response and continuous optimization of the production line, and the improvement of production efficiency and quality.

[0139] Through data fusion, digital twin simulation, multi-objective optimization, and a closed-loop feedback mechanism, the present invention solves various problems in packaging production, provides an innovative packaging control solution, reduces production costs, improves production efficiency and product quality, and at the same time reduces environmental impact.

[0140] Please refer to Figure 2 , the present invention provides a full-process packaging control system based on a digital twin model, including:

[0141] A data acquisition and fusion module that collects external demand data and internal packaging data through a unified interface, defines the reliability index of each data source, fuses each data source through a weighted algorithm, and generates a fused data set after data cleaning and preprocessing , for subsequent packaging strategy simulation and multi-objective optimization;

[0142] A packaging strategy simulation module that imports the fused data set into the digital twin system to construct a packaging strategy simulation scenario, conducts simulation analysis according to the scheme parameters in the simulation scenario, generates corresponding performance indicators, and then evaluates the effectiveness of different packaging schemes by the constructed packaging collaborative effectiveness evaluation function;

[0143] A multi-objective optimization module that, after determining the optimization objectives, comprehensively evaluates each packaging scheme through a multi-objective optimization algorithm, captures the collaborative conflicts between multiple objectives by the constructed comprehensive offset function, finds a balance point between multiple objectives, automatically screens out the optimal or sub-optimal scheme according to the optimization results, and generates specific adjustment suggestions;

[0144] A scheme deployment and verification module that deploys the selected optimal or sub-optimal packaging scheme to the actual packaging production line for small-batch trial production, collects production line data and detects the deviation from the model prediction. If an abnormal deviation occurs, the feedback mechanism is immediately triggered, and the deviation data is fed back to the optimization algorithm for model correction or scheme adjustment.

[0145] Those of ordinary skill in the art can realize that the units and algorithm steps of each example described in combination with the embodiments disclosed herein can be implemented by electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are executed in a hardware or software manner depends on the specific application and design constraints of the technical solution. Professional technicians can use different methods to implement the described functions for each specific application, but such implementation should not be considered to exceed the scope of this application.

[0146] Those skilled in the art can clearly understand that for the convenience and brevity of description, the specific working processes of the systems, devices, and units described above can refer to the corresponding processes in the foregoing method embodiments and will not be elaborated herein.

[0147] In several embodiments provided in the present application, it should be understood that the disclosed systems, devices, and methods can be implemented in other ways. For example, the device embodiments described above are merely illustrative. For example, the division of the units is only for some logical function divisions. In actual implementation, there may be other division methods. For example, multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the couplings, direct couplings, or communication connections shown or discussed with each other can be through some interfaces. The indirect couplings or communication connections of the devices or units can be in electrical, mechanical, or other forms.

[0148] The units described as separate components may or may not be physically separated. The components shown as units may or may not be physical units, that is, they can be located in one place or distributed to multiple network units. Some or all of the units can be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0149] As mentioned above, the above is only the specific implementation manner of the present application, but the protection scope of the present application is not limited thereto. Any person skilled in the art within the technical scope disclosed in the present application can easily think of changes or substitutions, which should all be covered within the protection scope of the present application. Therefore, the protection scope of the present application should be subject to the protection scope of the claims.

Claims

1. A full-process packaging control method based on a digital twin model, characterized in that: include, Collect external requirement data and internal packaging data through a unified interface, define the reliability index of each data source, fuse each data source through a weighted algorithm, and generate a fused data set after data cleaning and preprocessing , for subsequent packaging strategy simulation and multi-objective optimization; among them: for the initial data set output by weighted fusion Perform outlier detection and consistency correction, and define a multi-attribute deviation function To measure the deviation degree of the i-th record from the reference pattern, as follows: ； Wherein: is the actual value of the th record on the th key attribute, is the reference distribution center value of the th key attribute, is the total number of attribute dimensions; is the overall sensitivity amplification factor, the non - linear deviation index of each attribute individually; is the trade - off coefficient of the attribute collaborative offset term, is the non - linear index of the collaborative offset term; If If it exceeds the given deviation threshold, it is determined that the data is abnormal, and an appropriate correction method is selected according to the business scenario and subsequent usage requirements; After importing the fused dataset into the digital twin system, a packaging strategy simulation scenario is constructed. Under the simulation scenario, simulation analysis is carried out according to the scheme parameters. After generating the corresponding performance indicators, the effectiveness of different packaging schemes is evaluated by the constructed packaging collaborative effectiveness evaluation function. Among them, for each packaging scheme parameter simulation is sequentially executed in the digital twin environment to obtain the performance indicators of each key workstation under the packaging scheme, and the following packaging collaborative effectiveness evaluation function is defined : ； Wherein: represents the offset with respect to the reference process parameters set for the process plan under the work station ; the weight coefficient of the work station in the overall packaging process is the non-linear amplification factor of the work station ; the attribute collaboration item trade-off coefficient is the exponential term of the collaborative offset ; the work station number is the total number of work stations ; ​ After determining the optimization goal, a multi-objective optimization algorithm is used to comprehensively evaluate various packaging solutions. The constructed comprehensive offset function captures the collaborative conflicts among multiple objectives, finds the balance point among multiple objectives, and automatically selects the optimal or suboptimal solution based on the optimization results, and generates specific adjustment suggestions. After deploying the selected optimal or suboptimal packaging solution to the actual packaging production line, a small-batch trial run is conducted. Production line data is collected and deviations from the digital twin model predictions are detected. If an abnormal deviation occurs, a feedback mechanism is immediately triggered to feed the corresponding deviation data back to the optimization algorithm, adaptively updating the digital twin model parameters or external process assumptions, or adjusting the deployment plan. From the simulation result table Extract the target dimensions that meet the enterprise's requirements and specify them as the inputs of the multi-objective optimization problem, including four target dimensions, where: represents the average quality deviation index of the solution ; represents the total cost per unit output of the solution ; represents the total production cycle or changeover time of the solution at the specified output; represents the environmental impact index of the solution ; Define the comprehensive offset function , and use the quantization scheme to quantify the overall performance and its interaction effects under four target dimensions: ； Wherein: is the reference value or expected value of the target ; is the weight factor of the target in the overall evaluation, is the single-objective non-linear deviation index; is the amplification coefficient of the collaborative offset term, is the non-linear index of the multi-objective interaction term; when the comprehensive offset function has a smaller value, it indicates that the packaging scheme is closer to the ideal target of the enterprise as a whole and the multi-objective interaction conflict is less than expected, otherwise it indicates that there is a comprehensive offset greater than expected.

2. The full-process packaging management and control method based on the digital twin model according to claim 1 is characterized by: Collect external demand data and internal packaging data, and set up a data source set , where , and each of which represents a type of data source, and define the reliability index of each data source , based on the reliability index calculate the normalized weighted coefficient and obtain the initial data set of weighted fusion , and the weighting formula is set as follows: ； In the formula: is the reliability index of the data source, , is the preprocessing function performed on the original data, is the total number of data sources participating in the weighted fusion.

3. The full-process packaging management and control method based on the digital twin model according to claim 2 is characterized in that: Normalize the format of the dataset after completing anomaly filtering and correction, and finally output the fused dataset , perform position mapping, data type conversion, and timestamp alignment on the data fields according to the input interface and model requirements of the target digital twin system, and name the fused dataset after completing the format conversion as .

4. The full-process packaging management and control method based on the digital twin model according to claim 3 is characterized by: Load the fusion dataset into the digital twin environment, and define the equipment models, process flow structures, and adjustable packaging scheme parameter sets involved in the actual production line; Through the device description module of the digital twin system, the device structure and dynamic characteristics are associated so that it can call the device status fields in the fusion data set ; Register the unique identifier of all workstations in the digital twin system ; Introduce the packaging scheme parameters to be evaluated, denoted as , each corresponds to a packaging parameter combination, and pre-configure adjustable machine parameters matching in the digital twin system.

5. The full-process packaging management and control method based on the digital twin model according to claim 4 is characterized in that: Packaging collaborative efficiency evaluation function The lower the value of the smaller the collaborative deviation of the solution under the current working conditions; Summarize the simulation results generated for all packaging solutions ; integrate the results and performance indicators generated for each packaging solution to form a simulation result table , provide a visual display for key indicators, and mark special cases or abnormal scenarios .

6. The full-process packaging management and control method based on the digital twin model according to claim 5 is characterized by: Execute a multi-objective optimization algorithm to search for a set of candidate optimal solutions, using a modified evolutionary algorithm and mixed integer programming. After the iterations are complete, a series of approximate Pareto frontier solutions are output, that is, a set of solutions that achieve a relative balance between multiple objective dimensions. Screen and visualize the output multi-objective candidate solutions, and finally form a list of optimal or sub-optimal solutions, as follows: For the parameter sets of all packaging solutions as candidate solutions sort them in ascending order according to their comprehensive offset function values, and mark the top several solutions with the best performance; If an enterprise has a strong preference for a certain goal, it can be based on the weight factor or the reference value for re-screening to narrow down the solution set available for deployment; in the visualization view, highlight the solutions that have met the key constraints, and record the finally selected optimal or sub-optimal solution as , and attach the corresponding numerical indicators and recommended parameter adjustment information to form a decision list .

7. The full-process packaging management and control method based on the digital twin model according to claim 6 is characterized in that: Perform rapid deployment on the selected packaging solution , including: writing the packaging materials, label designs, size specifications, and corresponding machine optimization parameters in the optimal or sub-optimal solution into the control system of the actual packaging production line according to the established mapping relationship; registering its corresponding hardware interface mark ; Load the optimal or sub-optimal solution into the production line in the form of a configuration file or container image; use a unified parameter management module to inject the core variables of the optimal or sub-optimal solution into each workstation ; perform a consistency comparison between the hardware parameters and the digital twin model parameters, and enter the pre-production state after confirming no conflicts.

8. The full-process packaging management and control method based on the digital twin model according to claim 7 is characterized in that: For the deployed packaging solution Conduct small-batch production and perform real-time monitoring through the digital twin system to collect deviation information between the actual output and the model prediction. Among them, set the trial production batch Start the actual operation of the production line and install high-frequency data acquisition devices at each workstation to transmit the process data back to the digital twin system in real time; Introduce the dynamic offset increment function For the th real-time record at time : If exceeds the predetermined threshold, it is determined as an abnormal deviation, triggering an alarm or an automatic fallback strategy, and all real-time deviations , that is, the deviation detection results are stored in the intermediate result table .

9. The full-process packaging management and control method based on the digital twin model according to claim 8 is characterized by: Introduce a dynamic offset increment function For the th real-time record at time is evaluated as follows: ； In the formula: represents the th quality index of the product at time , is the corresponding value predicted by the digital twin model, representing the process variable related to the product; refers to the th product at time related th real-time process parameter, which is the predicted value of the process variable by the digital twin model; and respectively control the relative weights of the single-index deviation and the multi-index collaborative deviation; is the number of indicators; and are non-linear amplification factors.

10. The full-process packaging management and control method based on the digital twin model according to claim 9 is characterized in that: For any abnormal deviation records found, the corresponding process data is compared with the input features of the digital twin model to identify the possible source of the error. Based on the deviation distribution, the relevant model parameters or external process assumptions are adaptively updated to reduce future prediction errors. If the deviation is so severe that it affects the feasibility of the solution, the newly collected data will be returned to the multi-objective optimization algorithm to supplement and update the decision space. The revised digital twin model will be used to recalculate the comprehensive offset of the solution or other possible solutions to determine whether to switch or fine-tune the current deployment plan; if the deviation during the trial production after correction has returned to an acceptable range, it will be set as the default mass production plan.

11. A full-process packaging control system based on a digital twin model, which applies the full-process packaging control method described in any one of claims 1 to 10, characterized in that: include, The data acquisition and fusion module collects external requirement data and internal packaging data through a unified interface, defines the credibility indicators of each data source, fuses each data source through a weighted algorithm, and generates a fused data set after data cleaning and preprocessing for subsequent packaging strategy simulation and multi-objective optimization; The packaging strategy simulation module integrates the dataset After importing it into the digital twin system, a packaging strategy simulation scenario is constructed. Under the simulation scenario, simulation analysis is carried out according to the scheme parameters. After corresponding performance indicators are generated, the effectiveness of different packaging schemes is evaluated by the constructed packaging collaborative effectiveness evaluation function; The multi-objective optimization module, after determining the optimization goal, comprehensively evaluates each packaging solution through a multi-objective optimization algorithm. The constructed comprehensive offset function captures the collaborative conflicts among multiple goals, finds the balance point among multiple goals, automatically selects the optimal or suboptimal solution based on the optimization results, and generates specific adjustment suggestions; The solution deployment verification module deploys the selected optimal or suboptimal packaging solution to the actual packaging production line and then conducts a small-batch trial production. It collects production line data and detects deviations from the model prediction. If an abnormal deviation occurs, the feedback mechanism is immediately triggered to feed the deviation data back to the optimization algorithm for model correction or solution adjustment.

Images