A packaging monitoring system based on a smart factory

Abstract

The application relates to the technical field of intelligent factory data processing systems, and discloses a packaging monitoring system based on an intelligent factory, which comprises the following modules: a metadata harvesting module that acquires quality metadata of an identification code; an entropy source decoupling module that acquires a substrate background reference value before printing and calibrates the quality metadata after printing; a self-calibration module that determines the health state of a scanning device by reading a fixed reference code; a batch health index generation module that calculates a batch health index based on the calibrated quality index; and a decision support module that provides decision support for the index in combination with the health state of the device. The application converts originally neglected process data into a quantifiable management index that can accurately reflect process stability by constructing an internal data verification and noise filtering mechanism, and realizes the improvement of the packaging process quality from binary result judgment to continuous spectrum quantitative management.

Description

TECHNICAL FIELD

[0001] The present application relates to a packaging monitoring system based on an intelligent factory, belonging to the technical field of intelligent factory data processing system. BACKGROUND

[0002] Currently, using automatic scanning equipment to read the identification code on the package is a basic technical means to realize product traceability, warehouse management and logistics sorting business processes. It supports efficient operation of large-scale production and supply chain management by associating individual products with data records in the information system. In the current technical field, a commonly used operation mode is to process the identification code scanning link as a discrete binary result data collection point, that is, the scanning result is only divided into two states of success or failure. The management system mainly obtains and records the data string carried in the identification code. This mode maximizes the production line running speed, but in high-speed large-scale production applications, its inherent limitations are also fixed, that is, the management system actively ignores the process information that can represent the physical printing quality of the identification code itself during the scanning process.

[0003] The direct consequence of this operation mode is that the identification code group judged as successful in the factory actually contains a large number of samples in the identifiable and non-identifiable critical state. The physical appearance of these samples has reflected the cumulative effect of various small physical disturbances in the upstream printing, conveying and other links. Although they can be read in the relatively stable scanning environment of the factory, when the products enter the variable downstream circulation links such as light, angle and wear, the failure probability of these critical state identification codes read by other scanning equipment increases significantly. Specifically, the existing technology mainly has the following deficiencies: 1. Single management information dimension, the existing management system lacks a continuous index that can quantitatively represent the stability of the whole batch of product packaging process, resulting in that the quality management activities lag behind the actual occurrence of physical problems; 2. Risk management link misplacement, that is, an internal quality stability problem that can be intervened in the source is transformed into an external quality event that may cause high commercial costs when it occurs at the end of the supply chain.

[0004] To cope with the above problems, directly improving the production line physical equipment to pursue the perfect printing of the identification code, or adding a special online visual detection system, the former will cause the continuous investment of hardware cost, the latter will increase the complexity of the production process, which does not meet the economic requirements of large-scale production. These paths do not point to the root of the problem, that is, the structural information loss in the existing data processing and management method. Therefore, how to design a data processing system, use the existing scanning equipment on the production line for product traceability, collect and analyze the process data that characterizes the physical quality of the identification code which is originally ignored, and convert it into a decision support information that can be used for administrative management or supervision purposes and can quantify the stability of the production process, becomes the technical problem to be solved by the present application. SUMMARY

[0005] The present application provides a packaging monitoring system based on intelligent factory, which mainly aims to solve the problem that the existing management system cannot effectively utilize the physical quality process data of the packaging identification code, and cannot quantitatively evaluate the stability of the production process in advance without increasing additional hardware investment.

[0006] To achieve the above purpose, the packaging monitoring system based on intelligent factory provided by the present application comprises:

[0007] A metadata harvesting module configured to obtain quality metadata characterizing the physical quality of the identification code associated with each reading operation from a scanning device for reading the identification code on the packaging;

[0008] An entropy source decoupling module configured to obtain a reference value characterizing the optical characteristics of the substrate background of the to-be-printed area of the identification code before the identification code is printed on the packaging substrate, and calibrate the quality metadata obtained by the metadata harvesting module after printing based on the reference value, to generate a calibrated quality index that has compensated for the influence of the optical characteristics of the substrate background;

[0009] A self-calibration module configured to trigger the scanning device to read a fixed reference identification code with a preset physical quality level to obtain reference quality metadata, and determine the current health status of the scanning device based on the comparison between the reference quality metadata and the preset physical quality level;

[0010] A batch health index generation module configured to calculate a batch health index characterizing the stability of the packaging process of a production batch based on the statistical distribution of the calibrated quality indexes of a plurality of packages in the production batch;

[0011] a decision support module configured to attach a current health status of the scanning device as a data trustworthiness indicator to the batch health index, and to generate decision support information for administrative and management purposes when the batch health index meets preset rules.

[0012] Preferably, the quality metadata comprises one or more of the following data defined according to ISO / IEC standards: decoding degree, symbol contrast, axial non-uniformity, grid non-uniformity, and comprehensive quality grade, and the metadata harvesting module is further configured to store the quality metadata in association with the package unique identifier corresponding to the identification code.

[0013] Preferably, the batch health index generation module is configured to count the proportion of the number of packages in the production batch that are rated as different comprehensive quality grades after calibration, and to generate the batch health index by a weighted summation formula. The weighted summation formula is wherein, is a set of preset comprehensive quality grades, is a specific grade in the set, is a preset weight coefficient corresponding to the grade , and is the percentage of the number of packages in the batch that are rated as the grade to the total number.

[0014] Preferably, the decision support module is configured to automatically trigger a non-physical control level management workflow when the batch health index is lower than a preset first threshold value and the data trustworthiness indicator is high trustworthiness, and the management workflow comprises one or more of the following: updating quality management records associated with the production batch, adjusting finished product inspection strategies for the production batch, and generating package process stability early warning notifications.

[0015] Preferably, the self-calibration module is further configured to determine the data trustworthiness indicator as low trustworthiness when it is determined that the current health status of the scanning device does not meet the preset conditions, and to automatically generate an electronic maintenance work order for the scanning device in the device maintenance management system, with the content being a performance drift suggestion for cleaning or calibration, so as to convert the maintenance needs of the monitoring tool into a traceable management event.

[0016] Preferably, the entropy source decoupling module comprises an optical sensor installed upstream of the identification code printing device, the optical sensor being configured to measure the reflectivity of the area where the identification code is about to be printed as a reference value of the background optical property of the packaging substrate at the moment before printing; the entropy source decoupling module is further configured to calculate a theoretical mass attenuation expectation value from the reference value according to a pre-set calibration model, and subtract the theoretical mass attenuation expectation value from the actually measured mass metadata, and the difference is taken as the calibrated mass indicator.

[0017] Preferably, the system further comprises a context awareness module configured to identify a product line change event in the production process based on the change of the product identification code in the decoded content of the identification code acquired from the scanning device; and an adaptive baseline management module configured to automatically suspend the early warning function based on the batch health index in response to the identification of the product line change event, and enter a learning state of constructing or loading a historical baseline model for a production context corresponding to the new product identification code, so that the monitoring logic is dynamically synchronized with the production plan of the factory.

[0018] Preferably, the adaptive baseline management module is further configured to record the time consumed from the identification of the product line change event to the batch health index of the new product being higher than a pre-set health threshold value in consecutive, pre-set number of monitoring periods, and the fluctuation value of the batch health index in a sliding time window being lower than a pre-set stability threshold value, to generate a line change process stability time management indicator for quantifying and evaluating the efficiency of the line change operation.

[0019] Preferably, the system further comprises a device event log interface module configured to acquire the device event logs with accurate time stamps of a plurality of key devices on the production line from the manufacturing execution system or the device controller of the factory; and an entropy increase tracing module configured to perform statistical correlation analysis on the time sequence of the low-quality event whose calibrated mass indicator is lower than a pre-set quality threshold value, and the time sequence of the device event logs from different devices, to determine the possible physical source causing the decrease of the batch health index and generate corresponding diagnostic information when the batch health index decreases.

[0020] Preferably, the entropy increase tracing module is configured to include the name of a certain device event and the quantified strength value of the positive correlation in the diagnostic information when detecting that there is a persistent positive correlation between a certain device event and the low-quality event which is higher than a pre-set correlation threshold value, to direct the attention of the manager from the batch-level quality problem to the specific and device-associated executable maintenance task.

[0021] Compared with the prior art, the present application has the following beneficial effects:

[0022] 1. A new management data dimension is established by associating the acquired identification code physical quality metadata with the production batch information to which the package belongs, and calculating a batch health index based on the statistical distribution of multiple package quality metadata within the batch. This index is not simply a presentation of the quality of a single reading, but rather it converts the continuous and dispersed physical printing quality information on the production line into a holistic and trend-oriented indicator representing the stability of the packaging process before the product leaves the factory. This allows managers to move away from the limitations of only identifying scanning failures as a discrete result, and instead focus on continuous quantitative observation and management of process entropy increase that may lead to downstream scanning failures. This converts potential commercial risks occurring at the far end of the supply chain into a controllable process management object within the factory that is strongly associated with the production batch.

[0023] 2. The present application constructs an internal logical closed loop for information verification and traceability. The self-calibration module triggers the scanning device to read the fixed reference identification code, acquires reference quality metadata and compares it with the preset level to determine the health status of the scanning device itself, and adds a data reliability status identifier to the batch health index. Based on this, when a batch health index with high reliability decreases, the entropy increase traceability module aligns and analyzes the time information of the quality metadata with the event data time information obtained from the production line equipment. This mechanism avoids the long-standing management problem of being unable to distinguish between production process fluctuations and deterioration of the monitoring tool itself when a process risk signal is received. At the same time, it further analyzes the possible physical source of the problem by correlating the time information, directing the specific equipment event that caused the problem. This allows subsequent management or maintenance instructions to be based on data that is reliable and has a traceable cause.

[0024] 3、The application has self-adaptive ability to dynamic changes of production process, the context perception module uses the identification code obtained from the scanning device to decode the content and identify the product line change production event corresponding to the product identification code change, the identification of the event as an internal state switching instruction triggers the adaptive baseline management module to automatically adjust the baseline model used to calculate the batch health index, or switch the running state of the index generation module, such as suspending the alarm function and entering the learning state of building a baseline model for the new production context, this way of hierarchical and linked processing of the content information and quality information of the identification code makes the running logic of the monitoring system keep synchronized with the actual production plan and business process of the factory, avoiding a large amount of invalid management information caused by mismatching of the baseline when the production is switched within the plan, maintaining the long-term effectiveness of the system output information and the trust of the management personnel, and improving the monitoring accuracy of the packaging process stability to a new level, the module obtains the reference value of the optical properties of the substrate background of the to-be-printed area before the identification code is printed, and calibrates the quality metadata after printing based on the reference value, generates a calibrated quality index that has partially or completely compensated for the influence of the optical properties of the substrate background, and the batch health index generation module calculates based on the calibrated quality index, this operation mechanism of self-referencing before printing and spatial difference after printing effectively separates the quality fluctuations introduced by the inherent defects of the packaging substrate surface from the quality fluctuations introduced by the printing equipment and process, so that the final batch health index can better reflect the real state of the equipment and process, avoiding the monitoring baseline drift and management misjudgment caused by replacing packaging materials of different batches and different suppliers. BRIEF DESCRIPTION OF DRAWINGS

[0025] Figure 1 The timing diagram of the self-calibration and data reliability guarantee process of the system of the application;

[0026] Figure 2 The dynamic monitoring and early warning threshold diagram of the batch health index of the application;

[0027] Figure 3 The core management function and user role interaction use case diagram of the system of the application. DETAILED DESCRIPTION

[0028] To make the purpose technical solutions and advantages of the application clearer, the application will be further described in detail below, obviously, the described embodiments are part of the embodiments of the application, not all the embodiments, based on the embodiments in the application, all other embodiments obtained by those skilled in the art without creative labor belong to the protection scope of the application.

[0029] The application provides a packaging monitoring system based on an intelligent factory, which is deployed on a server or an edge computing node of the intelligent factory, utilizes existing scanning equipment on a production line, and converts a binary result judgment originally used for data collection into a business process optimization system for continuous quantitative supervision and management of packaging process stability. The system mainly includes a metadata harvesting module, an entropy source decoupling module, a self-calibration module, a batch health index generation module, and a decision support module in logic. The data processing process of the system starts from the metadata harvesting module acquiring quality metadata representing the physical quality of an identification code from the scanning equipment, then the entropy source decoupling module calibrates the metadata to separate the background influence introduced by the packaging substrate, further, the batch health index generation module calculates a batch health index representing a management dimension based on the calibrated index, while the self-calibration module periodically confirms the working state of the scanning equipment to ensure the credibility of the data, finally, the decision support module generates decision support information for administrative and management purposes according to the index and in combination with the data credibility state.

[0030] In engineering applications, a typical application scenario of the system is a large-scale, high-tact automatic packaging production line, for example, in the production of food, medicine or daily chemical products, there is a risk that the scanning fails in the downstream circulation link, thereby causing commercial losses, because the identification code printing quality is in a critical state between qualified and unqualified at the time of shipment; the existing management information system cannot supervise the continuous quality deterioration process in a forward-looking manner because it only handles the binary results of scanning success or scanning failure; the metadata harvesting module is the data acquisition unit of the system, when interacting with the scanning device in the existing manufacturing execution system or warehouse management system of the factory, it usually only receives the decoded string result, ignoring the process information associated with the scanning process which can represent the physical printing quality of the identification code; in order to obtain this part of information, the metadata harvesting module is configured as a software interface, which establishes an information channel with at least one scanning device used to read the identification code on the packaging through the standard industrial Ethernet protocol, the module sends instructions to the scanning device to switch its working mode from the default fast decoding mode to the detailed scanning result output mode, in this mode, the scanning device's output data packet contains not only the string of product unique identifier, but also a preset data structure which stores the quality metadata associated with this reading operation and representing the physical quality of the identification code, the quality metadata specifically includes one or more of the decoding degree, symbol contrast, axial non-uniformity, grid non-uniformity and comprehensive quality level defined according to the ISO / IEC15415 or ISO / IECTR29158 standard; after receiving the data packet, the metadata harvesting module parses and stores the quality metadata in association with the unique identifier of the packaging corresponding to the identification code and the current timestamp, in this way, a discrete business operation data point is converted into a management data record which can be used for subsequent quantitative statistics and trend analysis.

[0031] The entropy source decoupling module is used to solve the problem that when the quality of the identification code is monitored to decrease, it is unable to distinguish whether the root source is from the fluctuation of the printing, conveying and other process, or from the inherent defects or uneven optical properties of the surface of the packaging substrate itself; in order to realize the separation of the two influence sources, the entropy source decoupling module includes a photosensor installed upstream of the identification code printing device, which is used to measure the surface reflectivity of the identification code to be printed area before the packaging substrate enters the printing unit, and take the measurement value as the reference value of the background optical properties of the substrate, accordingly, the entropy source decoupling module is configured as a data processing unit, which receives two inputs, one is the reference value measured by the photosensor, and the other is the quality metadata obtained by the metadata harvesting module downstream after printing, the module internally stores a calibration model established through offline calibration experiment, which describes the functional relationship between different substrate background reference values and the highest theoretically achievable quality level of the identification code, when the module receives a set of real-time data, it first calculates a theoretical quality attenuation expected value from the calibration model according to the reference value, then subtracts the theoretical quality attenuation expected value from the actually measured quality metadata, and the difference is taken as a calibrated quality index which has compensated the influence of the background optical properties of the substrate; this index more accurately reflects the quality fluctuation introduced by the device process, thereby avoiding the monitoring reference drift caused by replacing different batches of packaging materials in management.

[0032] The self-calibration module is used to periodically verify the data reliability of the monitoring system itself to cope with the performance drift of the scanning device due to lens contamination or light source aging, and thus output unreliable quality metadata; for this purpose, the self-calibration module physically realizes a fixed reference identification code with a preset physical quality level, such as an integrated quality level of A grade, by pasting it on a fixed non-moving part in the optical path of the scanning device, and in software logic, the module is configured to automatically trigger the scanning device to read the fixed reference identification code at a preset business idle time point, such as before the start of each shift production, to obtain a set of reference quality metadata, and then compare the obtained reference quality metadata with the preset physical quality level to determine the current health status of the scanning device, the judgment procedure is: if the integrated level of the read reference quality metadata is A grade, it is determined that the health status of the scanning device meets the preset condition; if the level of three consecutive readings is lower than A grade, such as C grade, it is determined that it does not meet the preset condition, when it is determined that the current health status of the scanning device does not meet the preset condition, the self-calibration module determines a data reliability identifier as low reliability, and adds the identifier to all decision support information generated by the system in the subsequent period of time, at the same time, the module can automatically generate an electronic maintenance work order for the scanning device through a standard interface in the device maintenance management system, the content of which is performance drift suggestion cleaning or calibration, through this mechanism, the maintenance needs of the monitoring tool are converted into a traceable management event.

[0033] The batch health index generation module is used to convert the continuous and dispersed single packaging quality information on the production line into an overall index that can quantitatively represent the stability of the packaging process of a production batch from an administrative perspective; to achieve this purpose, the module first calculates the statistical distribution of the calibrated quality indicators of multiple packages output by the entropy source decoupling module in a production batch or a fixed time window, such as every 1 hour, specifically, it is configured to count the proportion of the number of calibrated quality indicators of multiple packages assessed as different integrated quality levels within the production batch, and generate a batch health index through a weighted summation formula The weighted summation formula is: wherein, is the batch health index; is a set of preset integrated quality levels, such as ; is a specific level in the set; is a preset weight coefficient corresponding to the level , which is set according to the influence of different levels on the scanning success rate of the downstream circulation link, an example of weight coefficient configuration is: ; and The percentage of the total number of packages that are rated as grade within the production batch; for example, in a batch containing 2000 packages, 1600 packages are rated as A ( ), 300 as B ( ), and 100 as C ( ), then the health index of the batch is calculated as , i.e. 92; the decision support module is the management value output link of the system, which converts the technical index of batch health index into specific management or supervision actions; the module is first configured to attach the current health status of the scanning device determined by the self-calibration module to the batch health index as a data reliability identifier, and then the module judges the index with the reliability identifier based on a rule engine configurable by the manager, when the batch health index meets the preset rule, generates decision support information for administrative and management purposes, a specific rule is: when the batch health index is lower than a preset first threshold, for example, from the conventional 95 to 85, and the data reliability identifier is high reliability, an automatic trigger of a non-physical control level management workflow, the management workflow includes one or more of the following: updating the quality management record associated with the production batch in the quality management system; adjusting the finished product inspection strategy for the production batch, for example, increasing the sampling rate from 1% to 5%; and generating a package process stability warning notice to the line supervisor and quality engineer.

[0034] To make the system adaptable to the dynamic changes of factory production plan, the system further comprises a context awareness module and an adaptive baseline management module. In production, product changeover production events, in which a production line switches from producing product A to product B, are common, which will lead to changes in packaging specifications and identification code types. If the monitoring baseline of the old product is continued to be used, invalid management information will be generated. Therefore, the context awareness module is configured to identify product changeover production events in the production process based on changes in product identification codes, such as SKU codes, in the identification code decoding content obtained from the scanning device. Specifically, the module compares the currently decoded product identification code with the main identification code in the previous time window. When a brand new identification code appears continuously more than a preset number of times, for example, 5 times, it is determined that a product changeover production event has occurred. Accordingly, the adaptive baseline management module is configured to automatically suspend the early warning function based on the batch health index in response to the identification of the product changeover production event, and enter a learning state of constructing or loading a historical baseline model for a production context corresponding to the new product identification code. This mechanism enables the monitoring logic to dynamically synchronize with the production plan of the factory. Further, to provide diagnostic decision support for managers when the batch health index decreases, the system further comprises a device event log interface module and an entropy increase tracing module. When the manager receives a process risk signal, the specific physical source causing the problem needs to be determined. Therefore, the device event log interface module is configured to obtain time-stamped device event logs of multiple key devices on the production line from the manufacturing execution system (MES) or device controller (PLC) of the factory, and the entropy increase tracing module is configured to perform statistical correlation analysis on the time series of low-quality events whose calibrated quality indicators are lower than a preset quality threshold and the time series of device event logs from different devices. When it is detected that there is a persistent positive correlation between a specific device event and a low-quality event, which is higher than a preset correlation threshold, for example, the correlation strength value reaches 75%, the entropy increase tracing module will include the name of the specific device event and the quantified strength value of the positive correlation in the diagnostic information, thereby guiding the attention of the manager from the batch-level quality problem to the specific device-associated executable maintenance task.

[0035] Example 1: In a highly automated pharmaceutical packaging line, the objective management problem is that, although the first scan success rate reported by the online identification code scanning device deployed at the end of the line has been maintained at more than 99.5% for a long time, complaints and returns about the failure to identify product traceability codes continue to occur in the downstream distribution and logistics links, resulting in an unexplained deviation between the factory's internal quality control data and the commercial costs caused by the far end of the supply chain; after deploying the packaging monitoring system of the present application in the production line, the metadata harvesting module obtains the quality metadata associated with each reading operation and containing the comprehensive quality level from the scanning device used for the final verification after product packaging through industrial Ethernet, and stores it in the quality database after associating it with the unique identifier of the packaging and the timestamp; in the initial stage of system operation, the batch health index generation module calculates the health index of the first batch based on the quality metadata of about 20,000 packagings in a production shift according to the weighted summation formula in the specific implementation , which is 78, while the first scan success rate of this shift is 99.6%, the simultaneous presentation of these two data provides managers with a quantitative evaluation of the internal process stability of the batch, showing that there are a large number of identification codes at level C in the batch of products.

[0036] In a subsequent production batch, due to the replacement of a new corrugated box supplier, the batch health index monitored by the system drops to 65, at this time, the decision support module pushes a packaging process stability warning notice to the supervisor's board according to the index being lower than the preset first threshold 80 and the self-calibration module confirming that the scanning device is in a high-confidence state; to diagnose the problem, the manager first reviews the data of the entropy source decoupling module, which obtains the reference values of the optical properties of the base materials of the new and old cartons through the photoelectric sensor deployed upstream of the printing device before printing, and calculates that the reference value of the new batch of cartons itself will cause the theoretical quality degradation expectation value of the identification code to increase by 10%, which makes the manager adjust the priority direction of troubleshooting from the printing device itself to the matching of materials and processes, and this judgment process is based on the data confidence guaranteed by the self-calibration module and the background noise separation capability provided by the entropy source decoupling module; to further accurately locate the process link that causes the index to drop, the entropy increase tracing module starts to perform statistical correlation analysis on the time series of the calibrated quality indicators and the time series of the device event logs obtained from the production line device controller, and after calculating thousands of low-quality events, the module reports that the probability of the closing action of the upper flap arm of the carton sealer occurring simultaneously within 0-2 seconds before all identification code events judged as grade D is 82% higher than the random occurrence probability of the event in the entire production process, showing strong positive correlation; based on this diagnostic information, the device maintenance personnel check the corresponding parts of the carton sealer and find that the pressure setting value of a pneumatic element is too high, causing the flap arm to cause excessive impact on the surface of the carton at the moment of closing, thereby causing the identification code with ink that has not yet fully solidified after printing to be slightly damaged, and after adjusting the pressure value, the batch health index of the new production batch returns to 96 and remains stable at this level; through the operation of the above-mentioned manner, the deviation in the management data is eliminated, and the quality management activity is changed from the original passive handling of downstream customer complaints to the upstream production process based on the batch health index and entropy increase tracing diagnostic information for prospective and quantitative supervision and intervention.

[0037] Example 2: To verify the batch health index The correlation between the packaging and the actual scanning success rate in the downstream circulation link, and the effectiveness of the entropy source decoupling module, are tested as follows. The test platform includes a packaging production line equipped with a thermal transfer printer and an online scanning device, and a scanning environment simulating the downstream circulation link, which is composed of multiple fixed and handheld scanning devices of different brands and models, and is set at different reading angles and light intensities. In the test, a control group data acquisition method using the prior art and a test group data acquisition method using the technical solution of the application are set. The control group only records the first scanning success rate of the online scanning device on the packaging production line. The test group uses the packaging monitoring system of the application to monitor the same batch of packaging passing through the same production line, collects quality metadata and calculates the batch health index . Five production batches are set in the test, each batch containing 1000 packaging samples. Different levels of identification code printing quality are generated by adjusting the temperature parameters of the printer head, and the material background influence is introduced by replacing the substrate type of the packaging carton. The test process and data recording are as follows. In batches 1, 2 and 3, A-type standard white cardboard is used as the packaging substrate, and the temperature parameters of the printer head are set to 100%, 85% and 70% respectively. In batches 4 and 5, B-type recycled corrugated paper cartons are used as the packaging substrate, which has lower surface roughness and optical reflection uniformity than A-type substrate, and the temperature parameters of the printer head are set to 85%. When monitoring batch 4, the packaging monitoring system of the test group turns on the entropy source decoupling module, while when monitoring batch 5, the module is disabled. After the samples of all five batches are produced and data are collected on the production line, they are sent to the downstream scanning environment, where they are scanned once at each of the 10 different workstations, and the overall scanning success rate is recorded. The test data are recorded in Table 1.

[0038] Table 1: Comparison of test data for different batches of samples.

[0039] Batch No. Print Parameters / Substrate In-plant first scan success rate (%) Entropy source decoupling module status Lot health index ) Downstream supply chain comprehensive scan success rate (%) 1 100% / Type A substrate 99.9 On 98 99.7 2 85% / Type A substrate 99.8 On 82 92.1 3 70% / Type A substrate 99.6 On 65 75.3 4 85% / Type B substrate 99.7 On 81 91.5 5 85% / Type B substrate 99.7 Off 55 91.5

[0040] According to the test data in Table 1, the first scanning success rates of batches 1, 2 and 3 in the factory are all above 99.6% and have little difference, but the batch health indexes calculated using the technical solution of the application show a gradient decline from 98 to 65. The change trend of this index is related to the change trend of the overall scanning success rate of the downstream supply chain, which decreases from 99.7% to 75.3%. This data shows that the batch health index This can serve as a management indicator, quantitatively reflecting potential downstream risks caused by decreased stability in the packaging process, which cannot be captured by traditional binary scanning results. Comparing the results of batch 4 and batch 5, both had the same physical production conditions and the same downstream scanning success rate of 91.5%. However, in batch 5, due to the disabling of the entropy source decoupling module, the calculated batch health index was lower. The index is 55, lower than the index 81 calculated after calibration by this module in batch 4. The index 55 deviates from the index 82 (corresponding to batch 2, with a similar downstream success rate), while the index 81 is close to the value of index 82. This data indicates that the entropy source decoupling module can separate the background influence introduced by the packaging substrate, so that the batch health index can more stably reflect the process quality determined by the production process itself. The test results confirm that the batch health index generated by the packaging monitoring system of this invention has a quantifiable correlation with the comprehensive readability of packaging in a variable application environment. By collecting and analyzing process data during the scanning process, the system provides a decision support basis that can be used for proactive monitoring of production process stability and supply chain risk management.

[0041] To further verify the substantial effect of the technical solution of the present invention compared with the conventional technical path in the background art from the reverse perspective, the following comparative example 1 is set up.

[0042] Comparative Example 1: This comparative example aims to simulate the conventional operating mode in the background technology that treats the identification code scanning process as a discrete binary result (i.e., success or failure), and to directly compare it with the experimental group in Example 2. This comparative example uses the exact same experimental platform, production batches (batch 2 and batch 3), and physical production conditions as Example 2. The only difference is that the data acquisition system in this comparative example strictly follows the conventional mode of the prior art and is only configured to record the number of successful and failed first scans of the online scanning equipment on the packaging production line. It does not collect or analyze any metadata characterizing the physical quality of the identification code (such as decoding degree, symbol contrast, comprehensive quality level, etc.), and therefore cannot calculate the batch health index. The experimental process and data recording are as follows: Under the same conditions as in Example 2, production batches 2 (printhead temperature 85% / Type A substrate) and 3 (printhead temperature 70% / Type A substrate) were produced. The online scanning equipment at the end of the production line scanned 1,000 packaging samples of each batch. The data acquisition system only recorded the number of samples that were successfully decoded and calculated the first scan success rate in the factory. Subsequently, the samples of these two batches were sent to the same downstream simulated scanning environment as in Example 2, and their comprehensive scan success rate at 10 different workstations was recorded. The experimental data are recorded in Table 2.

[0043] Table 2: Data recording table for Comparative Example 1.

[0044] Batch No. Print Parameters / Substrate In-plant first scan success rate (%) Management system status Downstream supply chain comprehensive scan success rate (%) Actual business impact 2 85% / Type A substrate 99.8 No anomaly detected, no early warning triggered 92.1 Small batch (about 7.9%) of downstream scan identification difficulties occurred 3 70% / Type A substrate 99.6 No anomaly detected, no early warning triggered 75.3 Significant (about 24.7%) downstream scan failure events occurred

[0045] According to the test results of Table 2, the first scan success rate recorded by the conventional data acquisition system in the factory is maintained at a high level of 99.6% in batches 2 and 3, with only a slight difference of 0.2% between the two batches. According to this single management indicator, the system fails to identify the significant deterioration of packaging process stability between the two batches, thus determining that the production process is stable and not triggering any early warning or management intervention workflow. However, after entering the downstream circulation link, the downstream supply chain comprehensive scan success rate of these samples is 92.1% and 75.3% respectively, showing a significant deviation from the monitoring data in the factory. In particular, in batch 3, nearly a quarter of the packages failed to scan downstream, indicating that the conventional technical path relying solely on binary scan results cannot quantify and warn potential downstream business risks caused by deteriorating print quality, and quality management activities significantly lag behind the actual occurrence of physical problems.

[0046] Embodiment 3: This embodiment combines Figures 1 to 3 , a packaging monitoring system based on an intelligent factory is described as follows: as shown in Figure 1 , at a preset calibration time, such as the beginning of each shift production, the self-calibration module sends a trigger calibration command to the scanning device, which then reads a fixed reference identification code preset to be of A quality level, and returns the reading result containing the reference quality metadata to the self-calibration module. After comparing the actual level with the preset level, if the health status meets the conditions, the decision support module is set a high credibility identifier to confirm the high credibility of subsequent monitoring data, otherwise, if the health status does not meet the conditions, a low credibility identifier is set and an electronic maintenance work order suggesting to clean or calibrate the device is generated in the device maintenance management system.

[0047] As shown in Figure 2 , the horizontal axis represents time and the vertical axis represents the batch health index BHI value. The batch health index BHI curve shown in solid line in the figure shows the fluctuation of its value within 24 hours. In the figure, a dashed line representing the statistical mean of 97.2 and a dotted line representing the warning threshold of 93.6 are also shown. By comparing the real-time index curve with these two reference lines, the manager can intuitively judge the stability state of the current production process. As shown in Figure 3 , the line supervisor / quality engineer as the core user can perform operations such as monitoring the batch health index, receiving process stability warnings, and analyzing entropy increase source diagnosis information. The manager can use the system to evaluate the efficiency of line change operation, and when the system generates an electronic maintenance work order, the device maintenance personnel will be the recipient and executor of the work order.

[0048] Example 4: In a specific packaging monitoring system deployment and debugging scenario, before the system is put into daily administrative management and supervision, its core calculation parameters need to be calibrated to reflect the process characteristics and risk features of a specific production line. This scenario involves a newly built packaging production line for producing high-value biological agents; the parameters used to calculate the batch health index... Weighting coefficients The calibration procedure is as follows: First, a calibration sample set containing 500 individually packaged items is produced on the production line. The printing equipment parameters are adjusted to ensure the sample set contains identification code samples with a roughly uniform distribution from Grade A to Grade D. Then, these 500 packaged samples are sent to a test platform simulating downstream distribution. This platform integrates multiple handheld scanning devices commonly used in pharmaceutical retail terminals. Under different lighting and angle conditions, each packaged sample is scanned 100 times repeatedly, and the average scan success rate is recorded. Through this experiment, the comprehensive quality level of the identification codes is established. Compared to its average scan success rate in this specific application environment The correspondence between them was determined. Furthermore, the weighting coefficient The value is set as the normalized value of the average scan success rate corresponding to each level, i.e. This yields a set of weighting coefficients associated with the risk characteristics of this production line: .

[0049] After determining the weighting coefficients, a batch health index was set for the decision support module. The procedure for setting the warning threshold is as follows: the production line is operated continuously for 24 hours under confirmed process stability, and the packaging monitoring system calculates and records the health index of a total of 24 batches in 1-hour time windows. The baseline data points were statistically analyzed to calculate their mean. The standard deviation is Based on the principles of statistical process control, this early warning threshold... The value is set at 3 standard deviations below the mean, and its calculation procedure is as follows: Substituting the values ​​will give you the result. During subsequent production processes, the values ​​calculated by the system When the value is below 93.6, the decision support module is triggered. The entropy increase tracing module analyzes the temporal correlation between physical quality degradation events and specific equipment events according to the following procedure: this module performs a correlation calculation at a fixed time period, such as every 10 minutes, and its input is the set of timestamps of all low-quality events recorded within that period. and a set of timestamps and types of all device events obtained from the device controller. The processing steps are as follows: First, for Each low-quality event timestamp Define a forward-looking time window The second step is to identify each key equipment event type that needs to be monitored on the production line. For example, the top-folding action of the carton sealing machine C, the calculation within this cycle, Total number of occurrences Total number of low-quality events ,as well as The occurrence timestamp falls into any low-quality event tracing window Total number of times within The third step is to calculate. low-quality events Improvement The calculation formula is as follows: In the formula, For the event With low-quality events The number of times they occur together within the trace window. This represents the total number of low-quality events. For device events The total number of occurrences, This represents the total number of device events within the current period; if the calculated lift... If the value is greater than the preset correlation threshold of 1.5, the system determines that the device is involved in an event. There is a statistical correlation with the decline in product physical quality, and corresponding diagnostic information is generated. By executing the above series of standardized calibration and interpretation procedures, the core parameters and algorithm logic inside the packaging monitoring system are set with reproducible engineering settings, enabling it to be coupled with the management needs and process details of a specific production line.

[0050] Example 5: In a bottled beverage factory where the production plan involves rapid switching between multiple products, the packaging monitoring system of the present invention operates as follows when dealing with the management needs of dynamic adjustment of monitoring benchmarks caused by product line changes: When the production line switches from producing product A to product B, the packaging bottle material and the label printing process change. If the monitoring benchmarks for product A are continued, management information that is not applicable to the production process of product B will be generated.

[0051] In this scenario, when the context-aware module identifies a product changeover event by continuously monitoring changes in the product identification code within the decoded identification code content, the adaptive baseline management module is triggered and automatically switches the system's operating state to a learning state for building or loading historical baselines for new products. In this state, all processes based on the batch health index... The early warning function was suspended to avoid generating invalid management notifications during the normal adjustment phase of process parameters; the system then began accumulating quality metadata for the production process of product B, and calculated the batch health index every 15 minutes within a preset time window. This learning state will continue until two preset quantitative conditions are simultaneously met within a preset number of consecutive monitoring cycles. The two quantitative conditions are: First, within four consecutive monitoring cycles, i.e., within one hour, the batch health index calculated in each cycle... The values ​​were all higher than a preset health threshold of 90; secondly, within the 1-hour sliding time window, these four... Standard deviation of the value Below a preset stability threshold When both conditions are met simultaneously, the adaptive baseline management module determines that the production process of the new product has entered a stable state, and the system automatically exits the learning state and transfers the current stable state data. The mean and standard deviation serve as monitoring benchmarks for new products, reactivating the early warning function. Simultaneously, the system records the time consumed from the identification of a product changeover production event to the fulfillment of the aforementioned stable state conditions, and generates a changeover process stability time management index for quantifying and evaluating the efficiency of changeover operations. This index is linked to the corresponding production batch record, providing data support for subsequent production management and process optimization.

[0052] Example 6: To ensure the accuracy of the entropy source decoupling module and the entropy increase tracing module in the packaging monitoring system of the present application in a specific production line environment, a set of standardized pre-calibration and model construction procedures need to be performed before the system is put into formal use. In a packaging production line that produces multiple specifications of corrugated cartons, the steps of the procedure are as follows: data is filled into the calibration model inside the entropy source decoupling module, which is used to establish the relationship between the optical properties of different packaging substrates and the theoretical maximum quality level of the identification codes that can be achieved on them; for this purpose, the engineer collects samples of all three specifications of packaging substrates used by the production line, namely substrate X, substrate Y and substrate Z. In the debugging mode of the production line, the process parameters of the printing unit are set to the optimal state, i.e. using a brand new print head, and the printing speed is reduced to 50% of the normal production speed, in order to exclude the influence of process fluctuations on printing quality to the greatest extent. Under these conditions, 100 identification code samples are printed on each of the three substrates, and the background optical property reference values of each sample before printing and the symbol contrast values in the quality metadata after printing are recorded using the photoelectric sensor and scanning equipment in the system. By calculating the average symbol contrast value of each set of 100 samples, the theoretical quality attenuation expectation value corresponding to each substrate is obtained, and this corresponding relationship is filled into a lookup table of the calibration model, with the content being {substrate X reference value: 85%, theoretical symbol contrast upper limit: 95%}, {substrate Y reference value: 72%, theoretical symbol contrast upper limit: 82%}, {substrate Z reference value: 65%, theoretical symbol contrast upper limit: 75%}.

[0053] To address the problem that the entropy increase tracing module may identify multiple device events with time correlation to low-quality events under complex working conditions, causing ambiguity in diagnostic information, a secondary sorting procedure is added to the diagnostic logic of the entropy increase tracing module. Under this procedure, the entropy increase tracing module calculates the correlation score of each candidate device event and selects all candidate device events with a correlation score greater than the threshold value of 1.5. After that, the exclusive contribution score of each candidate device event is calculated. The calculation method is as follows: in the trace window of all low-quality events, a subset of windows is selected in which only the device event occurs, without any other candidate device event occurring. On this subset, the conditional probability of the device event occurring with the low-quality event is calculated, and this probability value is recorded as the exclusive contribution score. Finally, when outputting diagnostic information to the decision support module, all candidate device events will be ranked according to their The scores are arranged in descending order, and the event with the highest score is identified as the most likely physical source; by performing the above procedure, the internal model of the entropy source decoupling module is constructed based on the field materials, and the diagnostic logic of the entropy increase tracing module has a quantitative priority ranking capability when facing multiple potential fault sources, so that the management decision support function of the monitoring system obtains a traceable data and logic basis in specific engineering applications.

[0054] It is obvious for those skilled in the art that the present application is not limited to the details of the above exemplary embodiments, and the present application can be implemented in other specific forms without departing from the spirit or essential characteristics of the present application.

[0055] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present application and are not limiting, and although the present application has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present application can be modified or replaced by equivalents without departing from the spirit and scope of the technical solutions of the present application.

Claims

1. A smart factory based packaging monitoring system, characterized in that, The system includes: A metadata harvesting module is configured to acquire quality metadata, which characterizes the physical quality of the identification code, associated with each read operation, from a scanning device used to read identification codes on packaging. An entropy source decoupling module is configured to acquire a reference value characterizing the background optical properties of the substrate in the area to be printed before the identification code is printed onto the packaging substrate, and to calibrate the printed quality metadata acquired by the metadata harvesting module based on the reference value to generate a calibrated quality index that has compensated for the influence of the background optical properties of the substrate. A self-calibration module is configured to trigger a scanning device to read a fixed reference identifier code with a preset physical quality level to obtain reference quality metadata, and to determine the current health status of the scanning device based on the comparison between the reference quality metadata and the preset physical quality level. A batch health index generation module is configured to calculate a batch health index characterizing the stability of the packaging process of a production batch based on the statistical distribution of calibrated quality indicators of multiple packages within a production batch. A decision support module is configured to attach the current health status of the scanning device as a data credibility identifier to the batch health index, and generate decision support information for administrative and management purposes when the batch health index meets preset rules.

2. A smart factory based packaging monitoring system as claimed in claim 1, wherein, The quality metadata includes one or more of the following data as defined by ISO / IEC standards: decoding degree, symbol contrast, axial non-uniformity, grid non-uniformity, and overall quality level. The metadata harvesting module is also configured to associate and store the quality metadata with the unique identifier of the package corresponding to the identification code.

3. A smart factory based packaging monitoring system as claimed in claim 1, wherein, a batch health index generation module configured to count the proportion of the number of calibrated quality indicators of the plurality of packages rated at different comprehensive quality levels in the production batch, and generate a batch health index through a weighted summation formula The weighted summation formula is wherein, is a set of preset comprehensive quality levels, is a specific level in the set, is a preset weight coefficient corresponding to the level , and is the percentage of the number of packages rated at the level in the batch in the total number.

4. The smart factory based package monitoring system of claim 1, wherein, The decision support module is configured to automatically trigger a non-physical control layer management workflow when the batch health index is lower than a preset first threshold and the data credibility is marked as high credibility. The management workflow includes one or more of the following: updating the quality management records associated with the production batch, adjusting the finished product inspection strategy for the production batch, and generating a packaging process stability early warning notification.

5. The smart factory based package monitoring system of claim 1, wherein, The self-calibration module is further configured to, when it is determined that the current health status of the scanning device does not meet the preset conditions, identify the data confidence level as low confidence, and automatically generate an electronic maintenance work order in the equipment maintenance management system that suggests cleaning or calibration for the scanning device to address performance drift, thereby transforming the maintenance needs of the monitoring tool into a traceable management event.

6. The smart factory based package monitoring system of claim 1, wherein, The entropy source decoupling module includes a photoelectric sensor installed upstream of the identification code printing device. The photoelectric sensor is used to measure the reflectivity of the area to be printed in the identification code just before printing on the packaging substrate, as a reference value for the background optical properties of the substrate. The entropy source decoupling module is also configured to calculate a theoretical mass decay expectation value from a preset calibration model based on the reference value, and subtract the theoretical mass decay expectation value from the actual measured mass metadata, with the difference serving as the calibrated mass index.

7. The smart factory based package monitoring system of claim 1, wherein, The system further comprises a context-aware module configured to identify a product changeover event in the production process based on decoding a change in the product identification code in the content based on the identification code obtained from the scanning device; and an adaptive baseline management module configured to automatically suspend the early warning function based on the batch health index and enter a learning state of constructing or loading a historical baseline model for a production context corresponding to the new product identification code in response to the identification of the product changeover event.

8. A smart factory based packaging monitoring system as claimed in claim 7, wherein, The adaptive baseline management module is further configured to record the time consumed from the identification of the product changeover event to the batch health index of the new product being higher than a preset health threshold in consecutive, preset number of monitoring cycles and the fluctuation of the batch health index in a sliding time window being lower than a preset stability threshold, to generate a changeover process stability time management indicator for quantifying and evaluating the efficiency of the changeover operation.

9. The smart factory based package monitoring system of claim 1, wherein, The system further comprises a device event log interface module configured to obtain the device event logs with precise time stamps of a plurality of key devices on the production line from a manufacturing execution system or a device controller of the factory; and an entropy increase tracing module configured to perform statistical correlation analysis on the time series of low-quality events with calibrated quality indicators lower than a preset quality threshold and the time series of device event logs from different devices to determine the possible physical sources causing the decrease in the batch health index and generate corresponding diagnostic information when the batch health index decreases.

10. A smart factory based packaging monitoring system as claimed in claim 9, wherein, The entropy increase tracing module is configured to include the name of a certain device event and a quantified strength value of the positive correlation in the diagnostic information when detecting that there is a persistent positive correlation between the certain device event and the low-quality event that is higher than a preset correlation threshold, to direct the attention of the manager from the batch-level quality problem to the specific and device-associated executable maintenance task.

Images