A beverage easy open end production line control system and method thereof

Abstract

The present application relates to the technical field of easy-open lid production control, and discloses a beverage easy-open lid production line control system and method. The sensing and monitoring module of the system collects lid deformation sensing data streams and assembly pressure sensing data streams in real time; the quality feature detection module performs geometric feature analysis based on preset lid deformation constraint conditions and in combination with the lid deformation sensing data streams, and generates lid quality distribution feature information; the process decision module generates a primary process decision scheme based on an embedded process decision space and process fitness constraint conditions, in combination with the lid quality distribution feature information and the assembly pressure sensing data streams; the trend prediction module performs production trend prediction on the two types of data, and obtains predicted quality distribution feature information and predicted assembly pressure data; the process compensation module compensates and optimizes the primary process decision scheme based on the predicted data, and generates an optimized process decision scheme; and the execution control module controls the stamping and forming mechanism based on the optimized scheme.

Description

Technical Field

[0001] This invention relates to the field of easy-open bottle production control technology, specifically to a control system and method for an easy-open bottle beverage production line. Background Technology

[0002] In the beverage packaging industry, easy-open lids are a key component of beverage containers, and their production quality directly impacts the user experience and safety performance of the product. With the continuous expansion of the beverage market, the demand for easy-open lids is constantly increasing, placing higher demands on the efficiency and precision of production lines. Currently, easy-open lid production lines on the market generally suffer from problems such as limited monitoring dimensions, delayed quality judgment, and reactive process adjustments.

[0003] Traditional monitoring methods for easy-open lid production lines often rely on manual sampling inspections or single-type sensor data collection, making it difficult to comprehensively capture key parameters in the production process. For example, some production lines only focus on the external dimensions of the lid, ignoring pressure changes during assembly; or they only collect assembly pressure data, failing to detect subtle deformations that occur during the stamping stage. This single-dimensional monitoring model makes it difficult to detect potential quality issues in a timely manner, often only discovering defects after the product is finished, resulting in significant waste of raw materials and production time.

[0004] In the process decision-making stage, most existing production lines use fixed process parameters, lacking dynamic response to real-time production data. Even those lines that have introduced simple automated decision-making mechanisms are often limited to adjusting parameters based on current single data points, failing to combine preset process logic with comprehensive analysis of multi-dimensional data. This decision-making approach is ill-suited to the differences in the characteristics of different batches of raw materials and cannot cope with fluctuations in external factors such as temperature and humidity in the production environment, easily leading to unstable production processes and problems such as excessive deformation of the cover and insecure assembly.

[0005] The existing production line lacks the ability to predict production trends and cannot anticipate potential quality fluctuations and parameter deviations during subsequent production based on historical and real-time data patterns. When production parameters become abnormal, remedial adjustments can only be made after the problem occurs, preventing preventative measures from being taken in advance. This further exacerbates the instability of the production process, reduces the overall efficiency of the production line, and increases the risk of substandard products entering the market, negatively impacting the company's brand image and economic benefits. Summary of the Invention

[0006] The purpose of this invention is to provide a control system for an easy-open beverage production line to solve the problems mentioned in the background art.

[0007] To achieve the above objectives, the present invention provides a control system for an easy-open beverage production line, the system comprising:

[0008] The sensing and monitoring module is used to collect real-time data streams of cover deformation sensing and assembly pressure sensing from the production line.

[0009] The quality feature detection module performs geometric feature analysis based on preset cover deformation constraints and the cover deformation sensing data stream to generate cover mass distribution feature information.

[0010] The process decision module, based on the embedded process decision space and process adaptability constraints, combined with the cover mass distribution characteristics and the assembly pressure sensing data stream, generates a preliminary process decision scheme.

[0011] The trend prediction module performs production trend prediction on the cover mass distribution feature information and the assembly pressure sensing data stream to obtain predicted mass distribution feature information and predicted assembly pressure data.

[0012] The process compensation module compensates and optimizes the primary process decision scheme based on the predicted quality distribution characteristic information and the predicted assembly pressure data, and generates an optimized process decision scheme.

[0013] The execution control module controls the stamping mechanism based on the optimized process decision scheme.

[0014] Preferably, the quality feature detection module includes:

[0015] The lid basic information acquisition unit is used to acquire the material parameters and structural specifications of easy-open lids in the production line;

[0016] A geometric feature modeling unit is used to construct a deformation prediction model of the cover body based on the material parameters and structural specifications.

[0017] The sensor data parsing unit is used to input the cover deformation sensor data stream into the cover deformation prediction model and output a cover deformation distribution cloud map.

[0018] The quality feature generation unit is used to determine the quality distribution feature information of the cover body based on the preset cover body deformation constraint conditions and the cover body deformation distribution cloud map.

[0019] Preferably, the construction of the quality feature detection module includes:

[0020] A cap setting unit of the same specification is used to retrieve the production records of easy-open caps of the same model as the current production line;

[0021] The historical feature library loading unit is used to load the historical database of quality features of the same specification cover body;

[0022] The feature analysis model construction unit is used to construct a quality feature analysis model based on the historical database of quality features of the same specification cover body.

[0023] The model sensitivity optimization unit is used to optimize the sensitivity of the quality feature analysis model based on the historical quality inspection records of the current production line, and generate the quality feature detection module.

[0024] Preferably, the model sensitivity optimization unit includes:

[0025] The sensitivity coefficient testing subunit is used to test the quality feature analysis model based on the historical quality inspection records of the current production line to obtain the feature detection sensitivity coefficient.

[0026] The incremental learning triggering subunit is used to perform incremental learning optimization of the quality feature analysis model based on the historical quality inspection records of the current production line when the feature detection sensitivity coefficient is lower than a preset sensitivity threshold.

[0027] Preferably, the process decision module generates a preliminary process decision scheme including:

[0028] The decision space calibration unit is used to perform correlation calibration on the process decision space based on the mass distribution characteristic information of the cover and the assembly pressure sensing data stream, so as to obtain a calibrated process decision space.

[0029] A decision trigger interval identification unit is used to identify process decision feature trigger intervals from the calibration process decision space;

[0030] The decision scheme generation unit is used to generate candidate process decisions based on the process decision features triggering intervals.

[0031] The decision fitness verification unit is used to verify whether the candidate process decision meets the process fitness constraints, and if it does, add it to the primary process decision scheme.

[0032] Preferably, the decision space calibration unit performs correlation calibration including:

[0033] Traverse the historical decision records in the process decision space to extract the sample cover mass distribution characteristics and sample assembly pressure data;

[0034] The quality feature similarity calculation subunit is used to compare the similarity depth coefficient between the current cover mass distribution characteristics and the sample cover mass distribution characteristics;

[0035] The pressure data similarity calculation subunit is used to compare the similarity depth coefficient between the current assembly pressure sensing data stream and the sample assembly pressure data.

[0036] The weighted correlation determination subunit is used to fuse the similarity depth coefficient of the quality feature and the similarity depth coefficient of the pressure data according to the preset weight coefficient to generate the correlation degree of process decision;

[0037] The decision record filtering subunit is used to add the corresponding sample decision to the calibration process decision space when the correlation degree of the process decision reaches the correlation threshold.

[0038] Preferably, the process compensation module performs compensation optimization including:

[0039] A prediction decision space generation unit is used to calibrate the process decision space based on the predicted quality distribution characteristic information and the predicted assembly pressure data to obtain a prediction calibration decision space.

[0040] A prediction decision triggering unit is used to identify the prediction process decision feature triggering space from the prediction calibration decision space;

[0041] The iterative decision generation unit is used to perform iterative decision-making within the predicted process decision feature triggering space based on the process fitness constraints, and generate a compensation process decision scheme.

[0042] The decision fusion unit is used to fuse and optimize the compensation process decision scheme with the primary process decision scheme.

[0043] Preferably, the system further includes:

[0044] The production line balance equation construction module is used to divide the process nodes of the production line, extract the material deformation characteristics of each process node, and construct the production line balance equation by combining the production disturbance factors acting on the process nodes.

[0045] The stamping equipment performance testing module is used to test the working status of the stamping forming mechanism at different pressure stages and obtain the dynamic performance parameters of the equipment.

[0046] The equipment model fusion module is used to integrate the dynamic performance parameters of the equipment into the production line balance equation.

[0047] Preferably, the system further includes:

[0048] The process flow calculation module is used to calculate the total material deformation of the production line within a preset cycle, and generate a theoretical process load by combining the dynamic performance parameters of the equipment.

[0049] The control command generation module is used to determine the opening and closing sequence of the stamping valve based on the theoretical process load and the unit deformation adjustment amount of the stamping valve.

[0050] The segmented control verification module is used to divide the preset cycle into discrete time periods, verify the actual deformation adjustment of the stamping valve in each time period, and dynamically adjust the opening and closing sequence of subsequent time periods.

[0051] Preferably, the present invention also includes a control method for an easy-open beverage production line, the method comprising all the modules and process flow of the aforementioned easy-open beverage production line control system.

[0052] Compared with the prior art, the beneficial effects of the present invention are:

[0053] This beverage easy-open cap production line control system provides a comprehensive optimization solution for the easy-open cap production process through the collaborative work of multiple modules. The sensor monitoring module can collect real-time data streams of cap deformation and assembly pressure, breaking through the limitations of traditional single-dimensional monitoring in production lines. This allows key parameters in the production process to be presented comprehensively and in real time, enabling even subtle changes in the production process to be captured in a timely manner, providing comprehensive data support for subsequent quality analysis and process decisions.

[0054] The quality feature detection module, based on preset lid deformation constraints, performs geometric feature analysis on the lid deformation sensing data stream and generates lid quality distribution feature information. This allows quality judgment to no longer rely on manual experience or single-dimensional inspection, but to achieve an objective and comprehensive assessment of lid quality through standardized constraints and precise data analysis. It can accurately identify subtle defects in the geometric shape of the lid and prevent unqualified products from flowing into subsequent stages due to inaccurate quality judgment.

[0055] The process decision module combines the embedded process decision space and process adaptability constraints with the cover mass distribution characteristics and assembly pressure sensor data stream to generate a preliminary process decision scheme. This makes process decision no longer a mechanical execution based on fixed parameters, but fully integrates actual production data and preset process logic, making the generated preliminary scheme more in line with the current production situation, adaptable to changes in the characteristics of different batches of raw materials and the production environment, and reducing production fluctuations caused by mismatched process parameters.

[0056] The trend prediction module uses the cover mass distribution characteristics and assembly pressure sensor data stream to predict production trends, obtaining predicted mass distribution characteristics and predicted assembly pressure data. This changes the passive situation of traditional production lines that can only remedy problems after they occur. By predicting production trends in advance, it can detect potential parameter deviations and quality fluctuations in future production processes, giving the initiative to adjust processes and avoiding large-scale quality problems and production stoppages caused by the accumulation of issues.

[0057] The process compensation module optimizes the initial process decision-making scheme based on predictive data, generating an optimized scheme that makes the process decisions more comprehensive and accurate. The initial scheme may not adequately consider future changes, while the compensation optimization stage fully utilizes predictive information to adjust it. This ensures that the final process scheme not only adapts to the current production situation but also addresses potential future changes, further enhancing the rationality and foresight of the process decisions.

[0058] The execution control module controls the stamping mechanism based on optimized process decision-making schemes, ensuring that the stamping process operates strictly according to optimal process parameters. This reduces problems such as excessive cap deformation and non-standard forming caused by deviations in mechanism operating parameters. Through the close cooperation of various modules, the entire system forms a complete closed loop in easy-open cap production, from monitoring, quality analysis, process decision-making to execution control. This improves the stability and accuracy of the production process, reduces the generation of defective products, reduces raw material waste and production time loss, and also reduces the pressure of manual intervention, allowing the production line to operate more efficiently and stably. Attached Figure Description

[0059] Figure 1 This is a schematic diagram of the working principle of the beverage easy-open cap production line control system described in this invention;

[0060] Figure 2 This is a schematic diagram illustrating the working principle of the quality feature detection module.

[0061] Figure 3 A schematic diagram illustrating the working principle of the quality feature detection module;

[0062] Figure 4 This is a schematic diagram illustrating the working principle of an additional module in the control system for an easy-open beverage production line. Detailed Implementation

[0063] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0064] Please see Figure 1 This invention provides a control system for an easy-open beverage production line, the system comprising:

[0065] The system includes a sensor monitoring module, a quality characteristic detection module, a process decision-making module, a trend prediction module, a process compensation module, and an execution control module.

[0066] The sensing and monitoring module collects real-time data streams from the lid deformation and assembly pressure sensors on the production line. The quality feature detection module, based on preset lid deformation constraints, performs geometric feature analysis using the lid deformation sensor data stream to generate lid mass distribution characteristic information. The process decision module, based on the embedded process decision space and process adaptability constraints, combines the lid mass distribution characteristic information and assembly pressure sensor data stream to generate a preliminary process decision scheme. The trend prediction module predicts production trends using the lid mass distribution characteristic information and assembly pressure sensor data stream, obtaining predicted mass distribution characteristic information and predicted assembly pressure data. The process compensation module compensates and optimizes the preliminary process decision scheme based on the predicted mass distribution characteristic information and predicted assembly pressure data, generating an optimized process decision scheme. The execution control module controls the stamping mechanism based on the optimized process decision scheme.

[0067] The sensing and monitoring module employs high-precision strain and pressure sensors, installed at key workstations on the production line, to collect real-time deformation and assembly pressure data of the cover during the stamping process. The data stream is transmitted to the central processing unit in time-series format. The quality characteristic detection module receives the deformation data stream, analyzes geometric features using a built-in algorithm, and, combined with deformation constraints, outputs information on the cover's quality distribution characteristics, including deformation uniformity, stress concentration areas, and potential defect locations. The process decision module stores historical process parameters and their corresponding quality results in its process decision space. By matching current quality characteristics and pressure data, it filters decision schemes that meet fitness constraints. The trend prediction module uses time-series analysis models, such as autoregressive integral moving average models, to perform multi-step predictions on quality characteristics and pressure data, outputting future production states. The process compensation module uses the prediction results to adjust the initial scheme, generating decisions more suited to future production conditions through iterative optimization. The execution control module converts the optimized scheme into control commands, driving the stamping mechanism to adjust pressure, speed, or stroke parameters.

[0068] Example 1: See Figure 2 In the specific operation of the easy-open beverage cap production line, the quality characteristic detection module, as the core link in the system's perception and understanding of the cap's condition, forms a complete processing chain from raw data to advanced quality information through the collaborative work of its internal units. The module's operation begins with the cap basic information acquisition unit, which actively retrieves detailed specifications of the current batch of easy-open caps from the production line's central database. These parameters include not only basic aluminum alloy material grades, thickness tolerance ranges, material elastic modulus, and yield strength, but also specific structural dimensions of the cap, such as the initial diameter of the cap blank, the design height and angle of the rolled edge, the precise width and depth of the pull ring groove, and the geometry of the pre-set weak points in the engraved areas. This information is encapsulated into a structured dataset, providing a solid physical basis for modeling and analysis.

[0069] After acquiring the basic information, the geometric feature modeling unit is activated. This unit uses the acquired material and structural data to construct a body deformation prediction model based on physical laws. This model is not a simple empirical formula, but a digital twin that integrates the principles of materials mechanics and finite element analysis. It calculates the flow and deformation behavior of materials under complex stress states by simulating the interaction between the upper and lower dies and the body blank during the stamping process. The model can predict the strain distribution, stress concentration, and thickness reduction trend of various parts of the body, such as the center of the panel, the rolled edge area, and the edge of the pull ring groove, under different pressures and stamping speeds. The output of this model is not a single numerical value, but a virtual field that reflects the deformation state of the entire body surface.

[0070] The sensor data analysis unit continuously receives data streams from a high-frequency strain sensor array deployed on the production line. These sensors capture the microscopic deformation of the cover blank during the stamping process in real time with an extremely high sampling rate, generating a three-dimensional data stream containing timestamps, spatial coordinates, and deformation values. The core task of this unit is to efficiently inject this massive amount of spatiotemporally labeled raw sensor data into the aforementioned cover deformation prediction model. The model uses the real-time sensor data as boundary conditions and validation benchmarks, performs rapid calculations and data assimilation, dynamically corrects its prediction results, and finally outputs a high-fidelity cover deformation distribution cloud map. This cloud map visually presents the deformation distribution of the entire cover at the moment of stamping completion in an intuitive color gradient or contour line format, where darker colored areas represent more severe deformation.

[0071] The quality feature generation unit is responsible for converting "morphology" into "quality." It internally stores a detailed set of cover deformation constraints, based on product standards, long-term production experience, and destructive testing results. These constraints typically include, but are not limited to: the maximum permissible arch height in the center area of ​​the panel, the minimum compression ratio at the rolled edge, the integrity threshold of the groove around the pull ring, and the uniformity index of the overall cover deformation distribution (e.g., prohibiting drastic local deformation gradients). This unit performs a pixel-by-pixel, region-by-region fine-grained comparison and calculation between the deformation distribution cloud map generated by the sensor data analysis unit and these constraints. This process identifies any abnormal areas in the cloud map that exceed the permissible range, such as a tiny local bulge, an incompletely formed groove, or an unevenly thick rolled edge. This unit generates a structured report of cover quality distribution characteristics. This report is no longer just an image, but a document containing quantitative data, which clearly marks the scope of qualified areas, the types of defects (such as wrinkles, cracks, and unfilled areas), the specific location coordinates of defects, and their severity ratings, providing a precise and actionable basis for quality judgment in process decisions.

[0072] Example 2: See Figure 3 In the construction and continuous optimization of the beverage easy-open cap production control system, the construction method of the quality feature detection module embodies a modeling concept based on historical data-driven and adaptive learning. The implementation of this method begins with a cap-setting unit of the same specification. This unit, as the starting point of the entire construction process, has the core function of accurately identifying a range of historical data highly comparable to the current production task. When the system is ready to build a quality detection model for a new production line or a new batch of caps, this unit initiates a query to the production line management system. The search criteria are strictly limited to all historical production records with the same model, material specifications, and structural design as the current target product. These records not only include the final quality inspection report but also detailed records of the process parameter sets used in each batch of production, such as stamping speed, pressure curve, mold number, and corresponding raw material batch information. In this way, the system ensures that the historical data on which the modeling relies is fundamentally homogeneous, minimizing modeling biases introduced by differences in product design.

[0073] After successfully identifying historical production records of the same specifications, the historical feature library loading unit is activated. This unit is responsible for extracting deeper historical quality feature data corresponding to these historical records from the central quality database. This data goes far beyond simple "qualified" or "unqualified" labels; it is a multi-dimensional time series and feature set. It may include raw waveform data of the cover deformation sensor collected every second in history, processed deformation distribution cloud map sequences, surface defect image feature vectors captured by the online visual inspection system, and quantitative indicators such as strength and sealing performance obtained from the final destructive physical test. All of this data is aligned and correlated according to production batch and timestamp, forming a rich and clearly structured historical database of quality features of covers of the same specifications. This database forms a solid foundation for training a high-precision quality feature analysis model.

[0074] The feature analysis model building unit utilizes this carefully prepared historical database to begin its core task, typically employing supervised machine learning algorithms to construct the model. Taking a convolutional neural network as an example, deformation distribution maps or defect images from the historical database serve as input features, while corresponding quality labels determined manually or by sophisticated instruments (such as "qualified," "slightly wrinkled," and "critically fractured") act as training targets. Through extensive iterative learning, the model attempts to discover and extract microscopic patterns and abstract features hidden within complex deformation maps or sensor data that are highly correlated with the final quality outcome. This process enables the final quality feature analysis model to infer the cap's quality state from new, unseen sensor data; essentially, it learns complex mapping relationships from historical experience.

[0075] A model trained solely on historical general data may not perform perfectly when faced with a specific production line, equipment condition, and environmental circumstances. The model sensitivity optimization unit is designed to address this issue. This unit incorporates historical quality inspection records from the current production line as a "calibration set." These records reflect the unique equipment characteristics, sensor biases, and micro-environmental fluctuations of the production line. The unit quantifies the model's performance in the current environment by calculating a metric called the "feature detection sensitivity coefficient." This coefficient comprehensively measures the model's recognition rate of real defects and its false alarm rate for normal operating conditions. If the coefficient is below a preset threshold, it indicates that the model is not sensitive enough to the "personality" of the current production line, and the incremental learning trigger subunit is activated. This subunit does not discard the original model but uses the historical data from the current production line as new training samples to incrementally fine-tune and optimize the original model's parameters. This approach allows the model to retain past general knowledge while quickly adapting to the specific characteristics of the current production line, thereby improving its discrimination accuracy and reliability in practical applications.

[0076] Example 3: In the control system of an easy-open beverage production line, the process decision module generates a preliminary process decision scheme through a complex reasoning process that integrates historical experience and real-time data-driven approaches. The core function of this module lies in its embedded process decision space, which is essentially a vast database storing process parameters, quality characteristics, pressure data, and their final outcomes corresponding to countless successful and failed production cases throughout history. When the module starts working, the decision space calibration unit is activated first. This unit receives real-time cap quality distribution characteristic information from the quality characteristic detection module and assembly pressure sensing data streams from the sensor monitoring module. Its task is to filter out the samples most similar to the current production situation from the vast historical decision records, thereby constructing a calibrated and more targeted decision subspace. To achieve this goal, two parallel similarity calculation subprocesses run within this unit.

[0077] The quality feature similarity calculation subunit is responsible for processing the cover mass distribution feature information. The current feature information is typically a high-dimensional feature vector, which may contain multiple quantitative indicators such as deformation uniformity index, coordinates of the maximum deformation region, and stress concentration factor. This subunit traverses the cover mass distribution features of each sample in the historical record and uses a weighted distance metric to calculate the similarity depth coefficient between the current feature and the features of historical samples. The coefficient is calculated as follows:

[0078]

[0079] in: The first eigenvector representing the current mass distribution feature vector of the cover body Values ​​of each dimension The first eigenvector representing the characteristic vector of the mass distribution of the historical sample cover Values ​​of each dimension It is to give the first The weight coefficients for each feature dimension reflect the importance of that feature dimension to the final quality. These weight coefficients are obtained through statistical analysis of historical data. This represents the total dimension of the feature vectors. Meanwhile, the pressure data similarity calculation subunit processes the assembly pressure sensing data stream. The current pressure data stream is a time series reflecting the dynamic changes in pressure during the stamping process. This subunit employs a dynamic time warping algorithm to handle potential minute phase drifts and velocity variations in the time series, calculating the morphological similarity depth coefficient between the current pressure sequence and historical sample pressure sequences. This coefficient focuses on the similarity between two sequences in terms of shape and trend of change, rather than alignment at absolute points in time.

[0080] The weighted correlation determination subunit receives the outputs from the two subunits mentioned above. and This subunit is based on weighting coefficients determined in advance through extensive experiments and analysis. and (satisfy This integrates two similarity depth coefficients into a comprehensive process decision correlation. ,Right now The decision record filtering sub-unit then uses this fused correlation as its basis. Filter through all historical records The value exceeds the preset association threshold. The selected sample decision records will collectively form the calibration process decision space. This calibrated space narrows the decision search scope, making it highly focused on historical scenarios similar to the current situation.

[0081] The decision trigger interval identification unit then performs an in-depth analysis of the calibration process decision space. This unit does not simply list historical decisions, but rather attempts to abstract common patterns from these similar samples. Through methods such as cluster analysis or multidimensional spatial density estimation, it identifies high-density regions in the calibration decision space where process parameters, quality characteristics, and pressure data are jointly distributed. These regions are called process decision feature trigger intervals. These intervals indicate which process parameters were most frequently used historically under what combinations of quality characteristics and pressure data, and which achieved good results. The decision scheme generation unit then operates within these identified trigger intervals. Based on the statistical characteristics of historical decisions within the intervals (such as mean, mode, or through interpolation methods), it generates a set of candidate process decisions. These decisions may include target setpoints for stamping pressure, adjustments to stamping speed, or fine-tuning suggestions for holding time.

[0082] The decision fitness verification unit performs feasibility checks on each generated candidate process decision. The process fitness constraints embedded in this unit are a set of hard boundary rules, such as the stamping force must not exceed the equipment's safety limit, the adjusted production cycle time must match the downstream process, and any process modification must not cause energy consumption to exceed the rated range. Each candidate decision must pass these constraints; only candidate solutions that fully meet all constraints are ultimately adopted and added to the primary process decision set, awaiting further processing and optimization by subsequent modules.

[0083] Example 4: See Figure 4In the operation of the easy-open beverage cap production line, the process compensation module plays a core role in forward-looking optimization decision-making. This module's operation is not based on a single real-time data source, but rather deeply integrates future production status information provided by the trend prediction module. Its compensation optimization process is a multi-stage, iterative calculation and decision-making flow. The prediction decision space generation unit, as the starting point of the process, operates similarly to the decision space calibration unit in Example 3, but the key difference lies in its input data source. This unit no longer receives purely real-time cap quality distribution characteristic information and assembly pressure sensor data streams, but rather predicted quality distribution characteristic information and predicted assembly pressure data output by the trend prediction module. These predicted data represent the system's estimation of cap quality and pressure changes over a future period (such as the next production batch or the next few minutes) based on historical time-series patterns and the current operating status. Using this predicted data, the unit also employs correlation calculation and threshold filtering methods to retrieve all historically most similar sample records from the global process decision space. Since the input is a predicted value, which inherently contains a certain degree of uncertainty, this unit may introduce a leniency factor when calculating the correlation degree, appropriately relaxing the strictness of the correlation threshold, thereby constructing a broader and more inclusive prediction calibration decision space. This space not only includes highly matched historical records but also incorporates some possible edge cases, providing richer possibilities for decision exploration.

[0084] After constructing the predictive calibration decision space, the predictive decision triggering unit begins operation. This unit's task is to identify valuable decision exploration areas from this broader space, namely, the predictive process decision feature triggering space. Unlike the "trigger interval" identified in Example 3, this identifies a "trigger space," a multi-dimensional, conceptually defined region with relatively ambiguous boundaries. It identifies areas where, although the correlation is not the highest, the corresponding process decisions have historically effectively addressed similar predictive scenarios by analyzing the distribution density and clustering of data points in the predictive calibration decision space. For example, if the system predicts that future material batches may experience slight fluctuations leading to a slight increase in deformation resistance, then the space comprised of all historically successful decision records that have handled similar material fluctuations will be identified as the current triggering space. This space definition focuses more on the adaptability and effectiveness of the decisions, rather than absolute similarity to the current state.

[0085] The iterative decision generation unit searches and optimizes within the identified predictive process decision feature trigger space. This unit employs iterative computation algorithms, such as genetic algorithms or particle swarm optimization, to automatically generate and evaluate a large number of candidate process parameter combinations within this space. Each candidate combination, i.e., a hypothetical process adjustment scheme, is simulated and verified using embedded process fitness constraints. These constraints include equipment physical limits, energy consumption limitations, production cycle synchronization requirements, and, most importantly, quality compliance standards. The algorithm iterates multiple times, eliminating candidate schemes that violate constraints or have poor simulation results, and generating new candidate schemes based on the characteristics of the winning schemes, gradually approaching the optimal solution. After multiple iterations, this unit outputs one or more high-performance compensatory process decision schemes, specifically tailored to address the predicted future production state.

[0086] The decision fusion unit is the final stage of the process, responsible for organically integrating future-oriented compensatory process decision schemes with primary process decision schemes based on real-time status. This fusion is not a simple weighted average or replacement, but an intelligent superposition and conflict resolution. An internal decision fusion matrix is ​​established to compare and coordinate every adjustable process parameter (such as punching force, speed, and position) in both types of schemes.

[0087] Table 1: Decision Fusion Matrix Data Table.

[0088]

[0089] This matrix defines the fusion strategy for different parameters. For example, for the main stamping pressure, a weighted average is used, giving higher weight to the real-time solution to maintain stability; for the stamping speed, a lower speed is predicted for future demand to avoid defects, so the compensation solution value is directly adopted; for the holding time, the longer of the two values ​​is selected to ensure sufficient forming; and for the die offset, the current real-time setting is maintained. Finally, this unit outputs a consistent and optimized process decision scheme that takes into account both the present and the future.

[0090] The production line balance equation construction module, as a parallel supporting subsystem, operates independently of the main control flow but provides it with deep model support. This module decomposes the continuous production process into several discrete process nodes, such as material feeding, pre-stamping, main stamping, hemming, and online inspection. For each node, the module extracts deformation characteristic parameters of the material being processed from the material library, such as work hardening index, strain rate sensitivity coefficient, and elastic recovery. Furthermore, the module models production disturbances acting on each node, including but not limited to natural fluctuations in material properties, progressive die wear, minor changes in lubrication conditions, and drift in ambient temperature and humidity. All this information is integrated into a mathematical framework to construct a balance equation describing the dynamic characteristics of the entire production line. This equation essentially reflects the quantitative relationship between material deformation, process parameters, and external disturbances. The stamping equipment performance testing module, through a specialized testing program, drives the stamping mechanism to operate under different pressure settings from low to high and collects its response data to obtain dynamic performance parameters of the equipment, such as pressure build-up delay time, punch positioning repeatability accuracy, and energy consumption curves under different loads. The equipment model fusion module ultimately integrates the measured equipment performance parameters into the production line balance equation, evolving the equation from an ideal material model into a digital twin that more realistically reflects the "material-machine" joint system, providing a deeper physical basis and predictive capability for the decision-making of the entire control system.

[0091] Example 5: In the refined control of an easy-open beverage production line, the focus is on the synergy between macro-management of production cycle and micro-adjustment of actuators. The core lies in converting theoretically calculated production load into precise control commands for key actuators, such as the stamping valves, and ensuring control effectiveness through a continuous verification and feedback mechanism. The process flow calculation module, as the starting point of the entire process, assesses the overall output demand and inherent capacity of the production system from a relatively macro-timescale perspective. This module defines a preset cycle, typically corresponding to the time span of a production batch, a maintenance cycle, or a stable material supply period. Within this cycle, the module accumulates the total material deformation of all process nodes on the production line (from feeding to stamping to forming) through integral calculations. This total deformation is an abstract concept that transforms different parts and forms of deformation (such as tension, compression, and bending) into a comparable scalar using the principles of energy or equivalent work, characterizing the "total workload" required by the production line within this cycle. This module incorporates dynamic performance parameters of the stamping equipment obtained from the stamping equipment performance testing module, particularly the average working efficiency, maximum sustainable output, and performance degradation curve of the stamping mechanism under continuous operation. By combining these equipment capability parameters with the calculated total material deformation, the module generates an index called theoretical process load. This index quantifies the average workload that the production line needs to maintain to achieve the expected output within a preset period, given the current equipment state. It considers both the demands of the production task and the physical limitations of the equipment itself, providing a benchmark reference for the generation of control commands.

[0092] The control command generation module is responsible for the crucial task of decomposing macroscopic load indicators into specific execution actions. The core inputs to this module are the theoretical process load and the unit deformation adjustment of the stamping valve. The unit deformation adjustment is a characteristic parameter measured through extensive experiments; it precisely describes the change in material deformation caused by each unit change in the stamping valve's opening (e.g., 1% opening or 1 mm displacement). This is a conversion coefficient that directly links control actions to actual production results. Based on the total workload requirement indicated by the theoretical process load and combined with the unit deformation adjustment, the module calculates the total adjustment stroke that the stamping valve needs to complete within the entire preset cycle. Furthermore, the module needs to plan how these adjustments are allocated across the timeline, i.e., determine the opening and closing sequence of the stamping valve. This sequence is not a simple on / off command, but a refined timing plan that specifies in detail when the valve begins to operate, the duration of the operation, the rate (slope) of the opening change, and the target opening value. When generating this sequence, the module must take into account a variety of constraints, including the mechanical response delay of the valve, the smoothness of the action to avoid impact on the material, and the synchronization of the action with other production line equipment (such as feeding mechanism and mold) to ensure the continuity and stability of the entire production process.

[0093] The segmented control verification module introduces a feedback loop, enabling the control system to adapt to real-time changes. This module divides the entire preset cycle into a series of continuous, shorter time periods. These discrete time periods allow for frequent checks and adjustments. At the end of each time period, the module collects the actual deformation adjustment generated by the valve's operation. This is typically obtained using a high-precision displacement sensor or a direct material deformation monitoring sensor. The module compares the actual deformation adjustment with the theoretically expected adjustment set by the control command at the beginning of the time period. The deviation may stem from various factors, such as valve aging or drift, minor changes in material properties, or unforeseen external disturbances. After calculating this deviation, the module does not passively record it but actively uses this information to dynamically adjust the opening and closing sequence plan for subsequent time periods. If the actual adjustment is consistently lower than expected, it may appropriately advance the valve's action time or increase the rate of opening change in subsequent time periods to compensate for the accumulated lag. Conversely, if the actual adjustment exceeds expectations, it may take the opposite measures to avoid overshoot. This rolling optimization method based on real-time verification enables the entire system to continuously observe the effects and fine-tune the operation like an experienced operator, thereby ensuring that at the end of the preset cycle, the total actual deformation can approach the initially calculated theoretical process load as closely as possible, thus guaranteeing the stability of production and the consistency of products.

[0094] It should be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus.

[0095] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A control system for an easy-open beverage production line, characterized in that, include: The sensing and monitoring module is used to collect real-time data streams of cover deformation sensing and assembly pressure sensing from the production line. The quality feature detection module performs geometric feature analysis based on preset cover deformation constraints and the cover deformation sensing data stream to generate cover mass distribution feature information. The process decision module, based on the embedded process decision space and process adaptability constraints, combined with the cover mass distribution characteristics and the assembly pressure sensing data stream, generates a preliminary process decision scheme. The trend prediction module performs production trend prediction on the cover mass distribution feature information and the assembly pressure sensing data stream to obtain predicted mass distribution feature information and predicted assembly pressure data. The process compensation module compensates and optimizes the primary process decision scheme based on the predicted quality distribution characteristic information and the predicted assembly pressure data, and generates an optimized process decision scheme. The execution control module controls the stamping mechanism based on the optimized process decision scheme; The quality feature detection module includes: The lid basic information acquisition unit is used to acquire the material parameters and structural specifications of easy-open lids in the production line; A geometric feature modeling unit is used to construct a deformation prediction model of the cover body based on the material parameters and structural specifications. The sensor data parsing unit is used to input the cover deformation sensor data stream into the cover deformation prediction model and output a cover deformation distribution cloud map. The quality feature generation unit is used to determine the quality distribution feature information of the cover body based on the preset cover body deformation constraint conditions and the cover body deformation distribution cloud map; The construction of the quality feature detection module includes: A cap setting unit of the same specification is used to retrieve the production records of easy-open caps of the same model as the current production line; The historical feature library loading unit is used to load the historical database of quality features of the same specification cover body; The feature analysis model construction unit is used to construct a quality feature analysis model based on the historical database of quality features of the same specification cover body. The model sensitivity optimization unit is used to optimize the sensitivity of the quality feature analysis model based on the historical quality inspection records of the current production line, and generate the quality feature detection module.

2. The beverage easy-open cap production line control system as described in claim 1, characterized in that, The model sensitivity optimization unit includes: The sensitivity coefficient testing subunit is used to test the quality feature analysis model based on the historical quality inspection records of the current production line to obtain the feature detection sensitivity coefficient. The incremental learning triggering subunit is used to perform incremental learning optimization of the quality feature analysis model based on the historical quality inspection records of the current production line when the feature detection sensitivity coefficient is lower than a preset sensitivity threshold.

3. The beverage easy-open cap production line control system as described in claim 1, characterized in that, The process decision module generates preliminary process decision schemes, including: The decision space calibration unit is used to perform correlation calibration on the process decision space based on the mass distribution characteristic information of the cover and the assembly pressure sensing data stream, so as to obtain a calibrated process decision space. A decision trigger interval identification unit is used to identify process decision feature trigger intervals from the calibration process decision space; The decision scheme generation unit is used to generate candidate process decisions based on the process decision features triggering intervals. The decision fitness verification unit is used to verify whether the candidate process decision meets the process fitness constraints, and if it does, add it to the primary process decision scheme.

4. The beverage easy-open cap production line control system as described in claim 3, characterized in that, The decision space calibration unit performs correlation calibration including: Traverse the historical decision records in the process decision space to extract the sample cover mass distribution characteristics and sample assembly pressure data; The quality feature similarity calculation subunit is used to compare the similarity depth coefficient between the current cover mass distribution characteristics and the sample cover mass distribution characteristics; The pressure data similarity calculation subunit is used to compare the similarity depth coefficient between the current assembly pressure sensing data stream and the sample assembly pressure data. The weighted correlation determination subunit is used to fuse the similarity depth coefficient of the quality feature and the similarity depth coefficient of the pressure data according to the preset weight coefficient to generate the correlation degree of process decision; The decision record filtering subunit is used to add the corresponding sample decision to the calibration process decision space when the correlation degree of the process decision reaches the correlation threshold.

5. The beverage easy-open cap production line control system as described in claim 1, characterized in that, The process compensation module performs compensation optimization, including: A prediction decision space generation unit is used to calibrate the process decision space based on the predicted quality distribution characteristic information and the predicted assembly pressure data to obtain a prediction calibration decision space. A prediction decision triggering unit is used to identify the prediction process decision feature triggering space from the prediction calibration decision space; The iterative decision generation unit is used to perform iterative decision-making within the predicted process decision feature triggering space based on the process fitness constraints, and generate a compensation process decision scheme. The decision fusion unit is used to fuse and optimize the compensation process decision scheme with the primary process decision scheme.

6. The beverage easy-open cap production line control system as described in claim 1, characterized in that, Also includes: The production line balance equation construction module is used to divide the process nodes of the production line, extract the material deformation characteristics of each process node, and construct the production line balance equation by combining the production disturbance factors acting on the process nodes. The stamping equipment performance testing module is used to test the working status of the stamping forming mechanism at different pressure stages and obtain the dynamic performance parameters of the equipment. The equipment model fusion module is used to integrate the dynamic performance parameters of the equipment into the production line balance equation.

7. The beverage easy-open cap production line control system as described in claim 6, characterized in that, Also includes: The process flow calculation module is used to calculate the total material deformation of the production line within a preset cycle, and generate a theoretical process load by combining the dynamic performance parameters of the equipment. The control command generation module is used to determine the opening and closing sequence of the stamping valve based on the theoretical process load and the unit deformation adjustment amount of the stamping valve. The segmented control verification module is used to divide the preset cycle into discrete time periods, verify the actual deformation adjustment of the stamping valve in each time period, and dynamically adjust the opening and closing sequence of subsequent time periods.

8. A control method for an easy-open beverage production line, characterized in that, It includes all modules and method flows of a beverage easy-open cap production line control system as described in any one of claims 1 to 7.

Images