Methods and systems thereof for tracing production of goods

Abstract

Computer-implemented methods and systems for tracing production of goods on a production line (10) includes capturing, by a vision component (30), image data of products (20) on the production line (10), processing the image data using an object detection algorithm to detect the products (20), assigning virtual markers (32) to the detected products (20), and tracking the products (20) by maintaining an association between the products (20) and the assigned virtual markers (32) as the products (20) move along the production line (10). The vision component (30) is positioned to monitor the products (20) as the products (20) travel along a conveyor (15) through the production line (10). The virtual markers (32) provide identification labels that enable the system to distinguish between individual products (20) and to track each product (20) as the product (20) progresses through the production line (10).

Description

METHODS AND SYSTEMS THEREOF FOR TRACING PRODUCTION OF GOODSFIELD OF INVENTION

[0001] The present disclosure relates to systems and methods for tracing the production of goods, and more particularly to traceability solutions for tracking products throughout a manufacturing process, such as in the food and beverage industry, utilizing mathematical models, intelligent vision systems, and digital twin technology.BACKGROUND

[0002] Traceability in manufacturing refers to the ability to track products throughout their production lifecycle, from raw materials to finished goods delivered to consumers. In the food and beverage industry, traceability systems enable manufacturers to monitor products as they move through various stages of production, including container formation, filling, sealing, labeling, aggregation, packaging, and palletizing. Such systems may employ various technologies including vision systems, sensors, mathematical models, and digital representations of production processes to monitor and record information about products as they traverse production lines.

[0003] Existing approaches to production traceability face several limitations. Conventional tracking methods may struggle to maintain accurate identification of products as they move through complex production environments, particularly when products accumulate, mix, or are regrouped at various stages of the production line. Vision-based systems may encounter difficulties in challenging lighting conditions or when tracking large numbers of products simultaneously. Mathematical prediction models used in isolation may lack real-time validation, potentially leading to inaccuracies in tracking product positions and timing. Furthermore, systems that rely on a single tracking methodology may be unable to adapt to varying production configurations or customer requirements.

[0004] Additional challenges arise in correlating products across different levels of aggregation, such as tracking individual containers within boxes and boxes within pallets. Production line dynamics, including variable conveyor speeds, accumulation table behavior, rejection and reinsertion rates, and machine stoppages, can introduce complexity that existing systems may not adequately address. The absence of mechanisms for transferring tracking data between monitoring devices as products move along a production line can result in loss of continuity in product identification.

[0005] It has been appreciated that a system and method is needed that overcomes one or more of these problems.SUMMARY

[0006] This Summary is provided to introduce a selection of concepts in a simplified form that are further described below in the Detailed Description.

[0007] In a first aspect, a computer-implemented method for tracing production of goods on a production line is provided. The method comprises capturing, by a vision component, image data of bottles on the production line. The method further comprises processing the image data using an object detection algorithm to detect the bottles. The method further comprises assigning virtual markers to the detected bottles. The method further comprises tracking the bottles by maintaining an association between the bottles and the assigned virtual markers as the bottles move along the production line.

[0008] The vision component captures image data and processes the image data using an object detection algorithm to detect and assign virtual markers to bottles, enabling continuous identification and tracking of individual bottles throughout the production process. This provides manufacturers with real-time visibility into product flow and enables accurate traceability records to be maintained from initial detection through final palletization.

[0009] The method may further comprise generating, by a mathematical model executed by one or more processors of a data processing system, predictions of positions of the bottles on the production line based on production parameters; wherein the production parameters comprise at least one of: a speed of an accumulation belt, a production volume, a time minimum between stations, and a carry-over between production periods.

[0010] The mathematical model generates predictions of bottle positions based on production parameters including accumulation belt speed, production volume, time minimum between stations, and carry-over between production periods. This enables manufacturers to predict bottle locations with statistical confidence and to correlate bottles across different stages of production without requiring physical intervention or manual tracking.

[0011] The method may further comprise identifying, by a first vision component, a group of the bottles moving along the production line. The method may further comprise executing a handshake between the first vision component and a second vision component; wherein the handshake comprises transferring tracking data including the virtual markers assigned to the group of the bottles from the first vision component to the second vision component. The method may further comprise continuing tracking of the group of the bottles by the second vision component using the transferred tracking data.

[0012] The handshake mechanism transfers tracking data including virtual markers between vision components positioned at different locations along the production line. This maintains continuous product identification as bottles transition between coverage areas of different vision components, eliminating gaps in traceability data that would otherwise occur at transitions between monitoring device fields of view.

[0013] The method may further comprise tracking the bottles as the bottles accumulate on an accumulation table; wherein the tracking comprises maintaining the association between each of the bottles and the corresponding virtual markers as the bottles mix and redistribute in a non-linear arrangement on the accumulation table.

[0014] Tracking bottles during accumulation maintains the association between each bottle and the corresponding virtual marker as bottles mix and redistribute in non-linear arrangements on the accumulation table. This preserves product identification throughout mixing processes and enables accurate correlation of bottles with subsequent packaging and palletizing operations.

[0015] The method may further comprise capturing serialization codes of the bottles on the production line. The method may further comprise associating the captured serialization codes with the virtual markers assigned to the respective bottles.

[0016] Capturing serialization codes and associating the captured serialization codes with virtual markers enables dual identification capability wherein a bottle can be identified and its complete production history retrieved using either the physical serialization code visible on the bottle or the virtual marker maintained in the tracking system. This supports traceability operations including recall management, quality control investigations, and regulatory compliance reporting.

[0017] The method may further comprise organizing the bottles into a hierarchical aggregate structure comprising a primary aggregate, a secondary aggregate containing a plurality of primary aggregates, and a tertiary aggregate containing a plurality of secondary aggregates. The method may further comprise recording timestamp data with minute granularity for each aggregate level to enable correlation of the bottles across the hierarchical aggregate structure.

[0018] Organizing bottles into a hierarchical aggregate structure with minute-level timestamp granularity enables precise correlation across all aggregate levels. When a quality issue is identified at the tertiary aggregate level, the system traces the issue to specific secondary aggregates and further to specific primary aggregates based on the recorded timestamp data and quantity information, enabling manufacturers to identify affected products with precision and to limit the scope of corrective actions.

[0019] The method may further comprise diverting one or more of the bottles to a rejection / reinsertion table. The method may further comprise maintaining identification of the divertedbottles through the assigned virtual markers while the bottles are on the rejection / reinsertion table. The method may further comprise reintegrating the diverted bottles into the production line while preserving the association between the reintegrated bottles and the corresponding virtual markers.

[0020] Maintaining identification of diverted bottles through assigned virtual markers while the bottles are on the rejection / reinsertion table and preserving the association upon reintegration ensures accurate traceability even for bottles that deviate from the standard production flow. This provides comprehensive traceability coverage that accounts for all bottles processed through the production line, including bottles that are temporarily removed from and subsequently returned to the production flow.

[0021] In a second aspect, a production line monitoring system is provided. The system comprises a production line comprising a conveyor configured to transport bottles. The system further comprises a vision component positioned to capture image data of the bottles on the production line. The system further comprises control means configured for capturing, by the vision component, image data of bottles on the production line; processing the image data using an object detection algorithm to detect the bottles; assigning virtual markers to the detected bottles; and tracking the bottles by maintaining an association between the bottles and the assigned virtual markers as the bottles move along the production line.

[0022] The production line monitoring system with a vision component and control means enables automated detection, identification, and tracking of bottles throughout the production process without requiring manual intervention for each individual bottle.

[0023] The system may further comprise control means configured for generating, by a mathematical model executed by one or more processors of a data processing system, predictions of positions of the bottles on the production line based on production parameters; wherein the production parameters comprise at least one of: a speed of an accumulation belt, a production volume, a time minimum between stations, and a carry-over between production periods; whereinthe control means comprises a mathematical model processing unit configured to generate the predictions of positions of the bottles.

[0024] The mathematical model processing unit generates predictions of bottle positions based on production parameters, enabling statistical prediction of bottle locations that supports traceability operations and enables manufacturers to determine the scope of affected products with precision during recall operations.

[0025] The system may further comprise a first vision component and a second vision component positioned at different locations along the production line; wherein the first vision component and the second vision component are configured to execute the handshake to transfer the tracking data. The system may further comprise control means configured for identifying, by the first vision component, a group of the bottles moving along the production line; executing the handshake between the first vision component and the second vision component; wherein the handshake comprises transferring tracking data including the virtual markers assigned to the group of the bottles from the first vision component to the second vision component; and continuing tracking of the group of the bottles by the second vision component using the transferred tracking data.

[0026] The first and second vision components positioned at different locations along the production line and configured to execute the handshake provide seamless data transfer that maintains continuous product identification throughout the entire production line.

[0027] The system may further comprise an accumulation table configured to receive the bottles from the conveyor. The system may further comprise a vision component positioned to capture image data of the bottles on the accumulation table. The system may further comprise control means configured for tracking the bottles as the bottles accumulate on the accumulation table; wherein the tracking comprises maintaining the association between each of the bottles and the corresponding virtual markers as the bottles mix and redistribute in a non-linear arrangement on the accumulation table.

[0028] The accumulation table with a vision component positioned to capture image data enables continuous tracking of bottles during accumulation operations where bottles mix and redistribute, preserving product identification throughout the mixing process.

[0029] The system may further comprise a serialization code reader configured to capture serialization codes of the bottles. The system may further comprise control means configured for capturing serialization codes of the bottles on the production line; and associating the captured serialization codes with the virtual markers assigned to the respective bottles.

[0030] The serialization code reader captures serialization codes and enables association with virtual markers, providing dual identification capability that supports comprehensive traceability operations.

[0031] The system may further comprise a rejection / reinsertion table configured to receive bottles diverted from the production line. The system may further comprise a vision component positioned to capture image data of the bottles on the rejection / reinsertion table. The system may further comprise control means configured for diverting one or more of the bottles to the rejection / reinsertion table; maintaining identification of the diverted bottles through the assigned virtual markers while the bottles are on the rejection / reinsertion table; and reintegrating the diverted bottles into the production line while preserving the association between the reintegrated bottles and the corresponding virtual markers.

[0032] The rejection / reinsertion table with a vision component enables tracking of bottles that are temporarily removed from the main production flow, ensuring comprehensive traceability coverage for all bottles including those that undergo inspection or quality control operations.

[0033] The system may further comprise a data processing system communicatively coupled to the vision component and configured to receive virtual marker data and serialization code data; wherein the data processing system is configured to create and store correlation records linking the virtual markers assigned to the bottles with the serialization codes of the respective bottles.The system may further comprise control means configured for capturing serialization codes of the bottles on the production line; and associating the captured serialization codes with the virtual markers assigned to the respective bottles.

[0034] The data processing system creates and stores correlation records linking virtual markers with serialization codes, enabling comprehensive traceability queries wherein a bottle can be identified and its complete production history retrieved using any of the associated identifiers.

[0035] In a third aspect, a computer program is provided. The computer program comprises instructions which, when executed by a computer, cause the computer to carry out capturing, by a vision component, image data of bottles on a production line; processing the image data using an object detection algorithm to detect the bottles; assigning virtual markers to the detected bottles; and tracking the bottles by maintaining an association between the bottles and the assigned virtual markers as the bottles move along the production line.

[0036] The computer program enables implementation of the tracing method on computing devices, providing flexibility in deployment across different production environments and enabling software updates to enhance traceability capabilities.

[0037] In a fourth aspect, a computer-readable medium is provided. The computer-readable medium comprises instructions which, when executed by a computer, cause the computer to carry out capturing, by a vision component, image data of bottles on a production line; processing the image data using an object detection algorithm to detect the bottles; assigning virtual markers to the detected bottles; and tracking the bottles by maintaining an association between the bottles and the assigned virtual markers as the bottles move along the production line.

[0038] The computer-readable medium stores instructions for the tracing method, enabling distribution and installation of the traceability system across multiple production facilities.

[0039] The system may further comprise a digital twin model configured to create a virtual representation of the production line and to simulate scenarios for predicting outcomes based on changes in production parameters.

[0040] The digital twin model creates a virtual representation of the production line and simulates scenarios for predicting outcomes based on changes in production parameters. This enables manufacturers to test different configurations and assess their impact on bottle movement and mixing levels, supporting informed decision-making and optimization of production processes.

[0041] Reintegrating the diverted bottles may comprise recognizing a code of each reintegrated bottle during reinsertion; wherein a mathematical model dynamically adjusts predictions to incorporate timing and position of the reintegrated bottles relative to bottles that remained in a primary production flow.

[0042] Recognizing a code of each reintegrated bottle during reinsertion and dynamically adjusting mathematical model predictions maintains real-time traceability while accounting for reinsertion timing and position, enabling accurate prediction of mixing levels and aggregate composition.

[0043] Reintegrating the diverted bottles may comprise processing the diverted bottles as a separate batch after completion of a production run.

[0044] Processing diverted bottles as a separate batch after completion of a production run provides cleaner data separation enabling higher confidence levels in pallet composition predictions by eliminating the complexity introduced by intermixing of reinserted bottles with primary production bottles.

[0045] The mathematical model may provide an approximation of a confidence level and an associated maximum delay of a pallet; wherein the maximum delay represents a maximum time difference between production of bottles found on a same pallet.

[0046] The mathematical model provides an approximation of a confidence level and an associated maximum delay of a pallet, enabling manufacturers to determine that bottles found on the same pallet are produced within a known time window with statistical confidence, supporting targeted recall operations that affect a minimal number of pallets when quality issues are identified.

[0047] The bottles may be identical in shape, size, and appearance. Tracking the bottles may comprise identifying distinguishing features, characteristics, and relative orientations of the bottles to maintain identification and distinguish between the identical bottles. The virtual markers assigned to the bottles may remain consistent through continuous capturing of the distinguishing features, characteristics, and relative orientations of the bottles.

[0048] Identifying distinguishing features, characteristics, and relative orientations of identical bottles enables the system to maintain distinct identities for bottles that are otherwise identical in general appearance. This capability enables persistent tracking even when bottles are visually identical in shape, size, and general appearance, as the system utilizes the distinguishing features and spatial relationships between bottles to preserve individual identities throughout the production process.BRIEF DESCRIPTION OF FIGURES

[0049] Embodiments of the invention will be described, by way of example, with reference to the following drawings, in which:

[0050] FIGURE 1 depicts a plan view of a production line for manufacturing goods, according to aspects of the present disclosure.

[0051] FIGURE 2 depicts a side perspective view of a plurality of bottles moving along a portion of the production line of FIGURE 1, according to an embodiment.

[0052] FIGURE 3 depicts a top plan view of a plurality of bottles gathered at an accumulation table of the production line of FIGURE 1, according to aspects of the present disclosure.

[0053] Common reference numerals are used throughout the figures to indicate similar features.DETAILED DESCRIPTION

[0054] The present disclosure relates to a production line monitoring and traceability system that integrates data collection, predictive modeling, and automated processing capabilities. The system is configured to trace goods, for example products, such as bottles, throughout a manufacturing process. The system utilizes mathematical models, vision systems, and digital twin technology to provide comprehensive tracking of products as the products move through various stages of production.

[0055] The system further provides mechanisms for tracking products that are temporarily removed from the main production flow for inspection or quality control purposes and subsequently reinserted, maintaining continuous identification throughout the rejection and reinsertion process.

[0056] A production line is configured for beverage production, including blowing containers, filling and sealing containers, aggregating containers, packaging containers, and palletizing containers. The production line monitoring and traceability system is configured to track products through each of these stages, from initial container formation through final palletization.

[0057] The present system addresses limitations in existing traceability approaches. Existing approaches present challenges with maintaining product identification through complex production environments where products mix, accumulate, and are regrouped during processing. Single-methodology tracking systems exhibit difficulties in providing accurate traceability data across diverse production conditions and equipment configurations. Existing approaches also lack mechanisms for continuous data transfer between monitoring devices positioned at different locations along a production line, resulting in gaps in product tracking information.

[0058] The production line monitoring and traceability system combines multiple complementary technologies to address these limitations. Mathematical models provide statistical predictions of product positions based on production parameters. Vision systems equipped with object detection and tracking algorithms provide real-time identification of individual products. Digital twin models create virtual representations of production line operations for simulation and analysis. The integration of these technologies provides a framework for maintaining accurate traceability records throughout the production process.

[0059] Figure 1 shows a plan view of a production line 10 according to an example embodiment. The production line 10 is configured for manufacturing goods such as beverages and includes multiple interconnected components that facilitate the flow of products from initial container formation through final palletization.

[0060] The production line 10 includes a conveyor 15 configured to transport products between various processing stations. The conveyor 15 extends through multiple sections of the production line 10 and includes curved and straight segments that route products through different stages of production. The conveyor 15 connects a facility section containing container forming and filling equipment to downstream processing areas including accumulation, packaging, and palletizing stations.

[0061] According to some example embodiments, a sensor 17 is positioned along the production line 10. The sensor 17 is configured to detect products as the products move along the conveyor 15 and to collect data relating to product flow and production parameters. The sensor 17 provides input data for the mathematical models and tracking algorithms described herein.

[0062] A vision component 30 is positioned to monitor products as the products travel through the production line 10. The vision component 30 is configured to capture images of products and to execute object detection and tracking algorithms for identifying and tracking individual products. The vision component 30 provides real-time data regarding product positions and movements along the production line 10.

[0063] As used herein, the term "vision component 30" refers to any device, system, or combination of devices configured to capture image data of bottles or other products on the production line. Vision components 30 include, but are not limited to: cameras, video cameras, image capture devices, machine vision cameras, smart cameras, industrial cameras, area scan cameras, line scan cameras, 3D cameras, depth cameras, infrared cameras, thermal cameras, highspeed cameras, and any other optical or imaging device capable of capturing visual data of products on the production line. The vision component 30 may be a standalone device or may be integrated with processing capabilities for executing object detection and tracking algorithms. According to some example embodiments, the vision component 30 comprises a camera communicatively coupled to a processor configured to execute object detection algorithms. According to other example embodiments, the vision component 30 comprises a smart camera with integrated processing capabilities. According to some example embodiments, the vision component 30 comprises a mesh of multiple devices working in coordination, wherein the mesh includes multiple cameras, processors, and communication interfaces that collectively capture image data, process the image data, and share tracking information across the mesh. The mesh architecture enables distributed processing, expanded coverage areas, and redundancy in the event of individual device failure. According to some example embodiments, the mesh of multiple devices operates as a unified vision component 30 that provides seamless tracking coverage across extended sections of the production line 10.

[0064] An accumulation table 40 is positioned along the production line 10. The accumulation table 40 is configured to receive products from the conveyor 15 and to temporarily hold products before the products proceed to subsequent processing stages. The accumulation table 40 provides a buffer zone where products collect and mix during normal production operations. The filling and emptying cycles of the accumulation table 40 create discretization of produced products, which provides beneficial data for predictive modeling.

[0065] A rejection / reinsertion table 50 is positioned along the production line 10. The rejection / reinsertion table 50 is configured to receive products that are removed from the main production flow for inspection or quality control purposes and to reinsert products back into the production flow after inspection.

[0066] The production line 10 includes a facility section on the left side containing container forming equipment, filling stations with tanks or vessels, and associated processing equipment. A corridor connects the facility section to the main production line area. The production line 10 extends to include multiple parallel conveyor sections with accumulation tables 40 shown as elongated areas containing grouped products. Packaging and palletizing machinery is positioned at the downstream end of the production line 10. Multiple connection points and routing paths allow products to flow between different processing stations throughout the production line 10.

[0067] The production line monitoring and traceability system includes a mathematical model configured to predict the position of bottles on the production line. The mathematical model employs statistical methods to analyze data points including speed of accumulation belts and production metrics. The mathematical model processes these data points to generate predictions regarding the probable locations of bottles at any given time during production.

[0068] The mathematical model provides an approximation of a confidence level and an associated maximum delay of a pallet. In one example, the mathematical model determines at approximately 95% confidence that the maximum delay of a pallet is approximately 28 minutes. This determination indicates that bottles found on the same pallet, even if packaged in separate boxes, are produced at most 28 minutes apart. The confidence level and time delay vary depending on the particular configuration of the production line, the speeds of accumulation belts, and various other production metrics.

[0069] The mathematical model is non-intrusive in that data is collected from existing hardware to make predictions about the approximate position of primary aggregate within secondary and tertiary aggregates. The primary aggregate comprises a filled, capped, and labeledbottle. The secondary aggregate comprises a box or wrapped group of multiple primary aggregates. The tertiary aggregate comprises a pallet of multiple secondary aggregates. A timestamp with minute granularity is provided for each of the aggregates. The date is provided for the primary aggregate, the number of items per box or group is provided for the secondary aggregate, and the number of boxes per pallet is provided for the tertiary aggregate.

[0070] The mathematical model generates predictions of bottle positions based on production parameters including accumulation belt speed, production volume, time minimum between stations, and carry-over between production periods. The data processing system executes the mathematical model through one or more processors and stores the generated predictions in one or more databases. The predictions enable manufacturers to determine the probable locations of bottles at any given time during production and to correlate bottles across different stages of production. The mathematical model is configurable to accept additional production parameters and to generate predictions tailored to specific production line configurations and customer requirements.

[0071] Production line data provided to the mathematical model includes speed of production, an approximation of loss or waste, a time minimum between stations, spacing and mixability parameters, and carry-over between days or periods of production. The mathematical model processes these parameters to generate predictions that are customizable based on specific production requirements including production volume and risk factors.

[0072] The normal behavior of machines on the production line, which are often stopped during operation, leads to filling and emptying of the accumulation table. This filling and emptying creates discretization of the produced bottles. The discretization of produced bottles is beneficial to the accuracy of the predictive model because empty space and time are created among different sublots or parts of production. These gaps in production flow provide reference points that improve the precision of position predictions generated by the mathematical model.

[0073] Unlike conventional mathematical prediction models used in isolation that lack realtime validation and lead to inaccuracies, the present mathematical model is designed to receive supplemental real-time data from vision systems and other sensors. This integration enables continuous validation and refinement of predictions. The vision systems track the production line to confirm production volume and report accurate volume data for inclusion in the mathematical model. The vision systems also provide real-time production metrics that would otherwise go unnoticed and cause inaccuracies in the mathematical model. This combination of statistical prediction with real-time data validation provides a framework for maintaining accurate traceability throughout the production process.

[0074] The production line monitoring and traceability system organizes products into a hierarchical aggregate structure comprising three levels. A primary aggregate comprises a filled, capped, and labeled bottle representing an individual finished product unit. A secondary aggregate comprises a box or wrapped group containing multiple primary aggregates. A tertiary aggregate comprises a pallet containing multiple secondary aggregates. This hierarchical organization enables the system to track products at each level of aggregation and to maintain relationships between individual products and their containing groups throughout the production and packaging process.

[0075] The system provides timestamp data with minute granularity for each aggregate level. For the primary aggregate, the system records a date indicating when the individual bottle was produced. For the secondary aggregate, the system records the number of items per box or group, enabling correlation between the secondary aggregate and the primary aggregates contained therein. For the tertiary aggregate, the system records the number of boxes per pallet, enabling correlation between the tertiary aggregate and the secondary aggregates contained therein. This timestamp and quantity data provides a framework for tracing products through each level of the aggregate hierarchy.

[0076] Conventional traceability systems face challenges in correlating products across different levels of aggregation. Tracking individual containers within boxes and tracking boxes within pallets presents difficulties when timestamp data lacks sufficient granularity or when hierarchical relationships between aggregate levels are not maintained. These challenges result in gaps in traceability data that prevent accurate identification of which primary aggregates are contained within specific secondary aggregates and which secondary aggregates are contained within specific tertiary aggregates.

[0077] The present system addresses these challenges by maintaining clear hierarchical relationships with granular timestamp data. The minute-level timestamp granularity enables precise correlation across all aggregate levels. When a quality issue is identified at the tertiary aggregate level, the system traces the issue to specific secondary aggregates based on the recorded number of boxes per pallet and timestamp data. The system further traces the issue to specific primary aggregates based on the recorded number of items per box and the date of production. This hierarchical correlation capability enables manufacturers to identify affected products with precision and to limit the scope of any corrective actions to the specific products involved.

[0078] The minute-granularity timestamp data enables manufacturers to reduce the scope of recall operations from an entire production day to a specific time window. For example, when a quality issue is identified, the hierarchical correlation capability enables manufacturers to identify that affected bottles were produced within a specific 28 -minute window rather than requiring recall of all bottles produced during an entire production shift or day. This targeted recall capability reduces the number of affected pallets and minimizes the commercial impact of quality issues.

[0079] Figure 2 shows a side perspective view of a plurality of bottles 20 moving along a portion of the production line 10. The bottles 20 are transported by the conveyor 15 along a curved path through the production line 10. Each bottle 20 includes a cap and body visible from the side perspective. The vision component 30 is positioned to monitor the bottles 20 as the bottles 20 travel along the conveyor 15.

[0080] The vision component 30 is equipped with an object detection algorithm configured to identify and track individual bottles 20 as the bottles 20 move along the production line 10. The object detection algorithm processes image data captured by the vision component 30 to detect the presence of bottles 20 and to assign identification data to each detected bottle 20.

[0081] The vision component 30 is configured to track bottles (20) that are identical in shape, size, and appearance, such as bottles of the same product type. To maintain identification and distinguish between identical bottles (20), the object detection algorithm identifies distinguishing features, characteristics, and relative orientations of the bottles (20). Example distinguishing features include features of the cap, label characteristics, and bottle features along the shoulder, base, or body of each bottle (20). In some embodiments, the vision component 30 logs for each bottle (20) the surrounding bottles (20) and their distinguishing parameters and relative orientations, enabling the system to maintain distinct identities for bottles (20) that are otherwise identical in general appearance.

[0082] According to one example embodiment, alternative methods for identifying distinguishing features include detecting micro -variations in slight variations in cap orientation, surface imperfections, and / or reflective properties of each bottle (20), and, for example label placement according to an example embodiment. According to another example embodiment, the vision component 30 utilizes pattern recognition to detect unique combinations of features that differentiate one bottle (20) from another. In some embodiments, the system creates a feature signature or fingerprint for each bottle (20) based on the combination of detected features. The feature signature enables re-identification of the bottle (20) even if the bottle (20) temporarily exits and re-enters the field of view of the vision component 30.

[0083] According to some example embodiments, when the vision component 30 temporarily loses tracking of a bottle 20, the system utilizes the stored feature signature to re-identify the bottle 20 when the bottle 20 re-enters the field of view. The re-identification operation compares detected features of bottles 20 in the field of view against stored feature signatures to match the re-enteredbottle 20 with the corresponding virtual marker 32. This failure recovery capability maintains tracking continuity even when temporary occlusions, lighting variations, or other factors cause momentary loss of tracking.

[0084] Visual markers 32 are assigned to individual bottles 20 by the vision component 30. The virtual markers 32 comprise identification labels such as id: 1 , id:2, and id : 3 that are associated with respective bottles 20. Each virtual marker 32 provides a persistent identifier that enables the system to distinguish between individual bottles 20 and to track each bottle 20 as the bottle 20 progresses through the production line 10. The virtual markers 32 are shown beneath three of the bottles 20 in Figure 2, indicating that the vision component 30 has identified and is tracking these bottles 20.

[0085] The virtual markers 32 are non-physical, software-generated identifiers that exist within the tracking system and are not physically applied to the bottles 20. Unlike physical labels or codes that are printed or etched onto the bottle surface, the virtual markers 32 are maintained in the data processing system and associated with each bottle 20 through the object detection and tracking algorithms executed by the vision components 30.

[0086] The vision component 30 employs deep learning algorithms to analyze video frames and to identify and track bottles 20 with accuracy across varying lighting conditions. The vision component 30 is configured to evaluate over 15,000 bottles 20 simultaneously, providing real-time data regarding the position and movement of bottles 20 along the production line 10. This realtime data supports operational efficiency by enabling continuous monitoring of product flow through the production line 10.

[0087] The deep learning algorithms are configured to detect subtle distinguishing features among identical bottles (20). In some aspects, the system maintains identification by tracking relative positions and orientations of each bottle (20) relative to neighboring bottles (20). This capability enables persistent tracking even when bottles (20) are visually identical in shape, size,and general appearance, as the system utilizes the distinguishing features and spatial relationships between bottles (20) to preserve individual identities throughout the production process.

[0088] According to one example embodiment, the deep learning algorithms are trained on datasets of identical bottles (20) to learn subtle variations that distinguish one bottle (20) from another. In some aspects, the system employs convolutional neural networks to extract feature vectors representing each bottle (20). The feature vectors enable comparison and matching of bottles (20) across successive frames captured by the vision component 30. According to another example embodiment, tracking algorithms such as object re-identification techniques are utilized to maintain persistent identification of bottles (20) as the bottles (20) move through different sections of the production line 10.

[0089] Conventional tracking methods present challenges in maintaining accurate identification of products as the products move through complex production environments. When products accumulate, mix, or are regrouped during processing, conventional methods exhibit difficulties in preserving the association between individual products and their assigned identifiers. These difficulties result in gaps in traceability data and reduced accuracy in tracking products through the production process.

[0090] The present system addresses these challenges by maintaining persistent identification through the virtual markers 32. The virtual markers 32 remain associated with each bottle 20 throughout the production process, including during accumulation, mixing, and regrouping operations. The vision component 30 continuously tracks the virtual markers 32 and updates the position data for each bottle 20 as the bottle 20 moves through different stages of the production line 10.

[0091] Figure 3 shows a top plan view of a plurality of bottles 20 gathered at the accumulation table 40. The bottles 20 are distributed across the accumulation table 40 in a scattered pattern, with some bottles 20 positioned closer together and others more spread apart. Each bottle 20 is represented by a square with a circular cap visible from above.

[0092] The virtual markers 32 are associated with each bottle 20 on the accumulation table 40. The virtual markers 32 display identification labels including idl, id2, id3, id4, id5, id6, id8, id9, idlO, idl 1, idl3, idl4, idl6, idl8, idl9, id20, id21, id22, id23, id24, id25, id26, id27, id28, id29, id30, id40, and id48. These identification labels demonstrate that the vision component 30 maintains tracking of each bottle 20 even as the bottles 20 mix and accumulate in a non-linear arrangement on the accumulation table 40.

[0093] The vision component 30 is equipped with an object detection and tracking algorithm configured to identify and track the caps of the bottles 20 from the top plan view. The object detection and tracking algorithm processes image data to maintain the association between each bottle 20 and the corresponding visual marker 32 as the bottles 20 move and reposition on the accumulation table 40. Curved lines in the lower left portion of Figure 3 represent portions of the production line 10 and the conveyor 15 that feed into or connect with the accumulation table 40.

[0094] The persistent identification provided by the virtual markers 32 enables the system to maintain accurate traceability data even in complex production environments where bottles 20 accumulate and mix. The virtual markers 32 provide a continuous record of each bottle 20 from initial identification on the conveyor 15 through accumulation on the accumulation table 40 and subsequent processing stages.

[0095] The production line monitoring and traceability system includes an intelligent system that leverages artificial intelligence to recognize and monitor bottles as the bottles move along the production line. The intelligent system includes one or more vision components (30) or image capture devices positioned to capture image data of bottles during production. The vision components (30) are configured to continuously monitor the production line and to transmit image data to processing components for analysis.

[0096] The intelligent system comprises an object detection algorithm supported by smart vision technology. The object detection algorithm is configured to perform two sequential operations: first detecting bottles within captured image frames, and second tracking the detectedbottles as the bottles move through the production line. The detection operation identifies the presence and location of bottles within each image frame. The tracking operation maintains the association between detected bottles across successive image frames, enabling continuous monitoring of bottle positions and movements.

[0097] According to some example embodiments, the object detection algorithm comprises a convolutional neural network architecture such as YOLO (You Only Look Once), Faster R-CNN (Region-based Convolutional Neural Network), or SSD (Single Shot MultiBox Detector). These architectures enable real-time detection and localization of bottles 20 within image frames captured by the vision components 30. The selection of object detection algorithm architecture depends on production line speed, required detection accuracy, and available computational resources. According to another example embodiment, the object detection algorithm comprises a desired architecture, or for example, a new architecture is developed and implemented, which is suitable for the particular production line speed, required detection accuracy, and available computational resources.

[0098] The intelligent system employs deep learning algorithms to analyze video frames captured by the vision components (30). The deep learning algorithms process the video frames to identify bottles and to track the identified bottles with accuracy across varying lighting conditions. The deep learning algorithms are trained to recognize bottle characteristics and to maintain accurate identification even when lighting conditions change during production operations. This capability enables the intelligent system to operate reliably in production environments where lighting conditions vary due to equipment operation, time of day, or other factors.

[0099] The intelligent system is configured to evaluate over 15,000 bottles simultaneously. The processing capacity of the intelligent system enables real-time data collection regarding the position and movement of large numbers of bottles as the bottles progress through the production line. The real-time data generated by the intelligent system supports operational monitoring and provides input for the mathematical models and other traceability components of the system.

[0100] Conventional vision-based systems encounter difficulties in challenging lighting conditions and when tracking large numbers of products simultaneously. Variations in lighting intensity, shadows, and reflections cause conventional systems to lose track of products or to generate inaccurate identification data. The processing limitations of conventional systems restrict the number of products that are tracked concurrently, resulting in gaps in traceability data during high-volume production operations.

[0101] The present intelligent system utilizing deep learning algorithms overcomes these limitations. The deep learning algorithms maintain accurate identification and tracking regardless of lighting variability by learning to recognize bottle characteristics across a range of lighting conditions. The processing architecture of the intelligent system enables concurrent tracking of over 15,000 bottles, providing comprehensive coverage during high -volume production operations without gaps in traceability data.

[0102] The intelligent system supports the mathematical model by providing real-time information that enhances the accuracy of predictions generated by the mathematical model. The intelligent system tracks the production line to confirm production volume and reports accurate volume data for inclusion in the mathematical model. The production volume confirmation provided by the intelligent system enables the mathematical model to operate with validated input data rather than estimated or historical data. The accurate volume data reported by the intelligent system is incorporated into the mathematical model calculations, thereby improving the precision of position predictions and timing estimates generated by the mathematical model.

[0103] The intelligent system provides real-time production metrics to the mathematical model. These real-time production metrics include data that would otherwise go unnoticed during production operations and that would cause inaccuracies in the mathematical model if not captured. The continuous flow of real-time data from the intelligent system to the mathematical model enables ongoing validation and refinement of the predictions generated by the mathematical model.

[0104] Observations are directly communicated to the object detection and tracking algorithm. The direct communication of observations enables the object detection and tracking algorithm to receive updated information regarding production conditions and product flow. The observations are also uploaded to modules or data -receiving devices configured to receive and process observation data. The modules and data-receiving devices process the uploaded observations and provide the processed data to the object detection and tracking algorithm and the mathematical model. This communication of observations influences the accuracy of detection and tracking operations performed by the object detection and tracking algorithm.

[0105] One or more vision components (30) or other vision systems are implemented to provide remote observations of the production line. The vision components (30) and vision systems capture image data and other observation data from positions along the production line without requiring an operator to be physically present at each observation location. The remote observation capability provided by the vision components (30) and vision systems enables continuous monitoring of production parameters and product flow throughout the production line. The remote observations generated by the vision components (30) and vision systems replace or supplement observations that would otherwise be provided by an operator, thereby reducing reliance on manual observation and enabling automated data collection.

[0106] Mathematical prediction models used in isolation lack real-time validation of the predictions generated by the models. The absence of real-time validation leads to inaccuracies in tracking product positions and timing because the isolated prediction models operate on static or historical data that does not reflect current production conditions. Changes in production parameters, equipment operation, or product flow that occur during production are not captured by isolated prediction models, resulting in divergence between predicted positions and actual positions of products on the production line.

[0107] The present integration of vision systems with the mathematical model provides continuous real-time validation of predictions. The vision systems capture current production dataand transmit the current production data to the mathematical model for comparison with predicted values. When discrepancies between predicted values and observed values are detected, the mathematical model adjusts the predictions based on the real-time data provided by the vision systems. This continuous validation and adjustment process overcomes the inaccuracies inherent in isolated prediction approaches by maintaining alignment between the mathematical model predictions and actual production conditions throughout the production process.

[0108] The production line monitoring and traceability system includes monitoring capabilities for the accumulation table 40 that enable continuous tracking of bottles 20 during accumulation operations. As shown in Figure 3, bottles 20 gathered at the accumulation table 40 are identified and tracked using vision components 30 that assign virtual markers 32 to each bottle 20. The vision components 30 maintain tracking of each bottle 20 even as bottles 20 mix and accumulate in a non-linear arrangement on the accumulation table 40.

[0109] The accumulation table 40 receives bottles 20 from the conveyor 15 and temporarily holds the bottles 20 before the bottles 20 proceed to subsequent processing stages. During accumulation, bottles 20 distribute across the accumulation table 40 in scattered patterns where some bottles 20 position closer together and others spread apart. The vision components 30 equipped with object detection and tracking algorithms process image data to maintain the association between each bottle 20 and the corresponding virtual marker 32 as bottles 20 move and reposition on the accumulation table 40.

[0110] During accumulation where identical bottles (20) mix, the vision components 30 track distinguishing parameters and relative orientations of each bottle (20) relative to surrounding bottles (20). In some embodiments, the vision components 30 log for each bottle (20) the neighboring bottles (20) and the distinguishing features thereof, including cap features, label characteristics, and bottle features along the shoulder, base, or body. This tracking of distinguishing parameters and relative orientations enables the system to maintain distinctidentities for bottles (20) that are otherwise identical in appearance as the bottles (20) mix and redistribute on the accumulation table 40.

[0111] According to one example embodiment, during accumulation, the system predicts bottle (20) movements based on physics models and validates predictions against observed positions of the bottles (20). In some embodiments, the system maintains a spatial map of bottles (20) on the accumulation table 40 with associated feature signatures for each bottle (20). When bottles (20) move or reposition on the accumulation table 40, the system matches observed features against stored feature signatures to maintain identification. According to another example embodiment, temporal tracking is utilized where the system continuously monitors bottle (20) movements to preserve identity associations as the bottles (20) mix and redistribute on the accumulation table 40.

[0112] One or more operators are provided for monitoring certain aspects of the production line 10. An operator is positioned by the accumulation table 40 and monitors fullness of the accumulation table 40. The operator observes and reports parameters including whether the accumulation table 40 is full, empty, one-third full, or at other fullness levels. Observations from operators are communicated to the object detection and tracking algorithm or uploaded to modules or data-receiving devices to influence the accuracy of detection and tracking operations.

[0113] According to some example embodiments, the quantity of bottles 20 present on the accumulation table 40 is automatically monitored and communicated with an algorithm by accumulation sensors. The accumulation sensors are physical or optical sensors placed on the accumulation table 40. The accumulation sensors detect the presence and quantity of bottles 20 on the accumulation table 40 and transmit this data to the algorithm. In one example, the algorithm comprises the object detection and tracking algorithm. The automatic monitoring provided by the accumulation sensors supplements or replaces manual observation by operators.

[0114] Existing traceability systems do not adequately address the complexity introduced by accumulation table behavior where products mix and regroup. When products accumulate in non-linear paterns and reposition relative to one another, existing systems lose track of individual product identities and fail to maintain accurate associations between products and their assigned identifiers. This loss of tracking during accumulation results in gaps in traceability data that prevent accurate identification of products after the products exit the accumulation table.

[0115] The present system maintains continuous tracking of each botle 20 through virtual markers 32 even during non-linear accumulation patterns. The virtual markers 32 remain associated with each bottle 20 throughout the accumulation process, preserving product identification as botles 20 mix and regroup on the accumulation table 40. The vision components 30 continuously update position data for each bottle 20 based on the virtual markers 32, enabling the system to maintain accurate traceability records regardless of the complexity of bottle movements during accumulation. This continuous tracking capability preserves product identification throughout the mixing process and enables accurate correlation of botles 20 with subsequent packaging and palletizing operations.

[0116] The production line monitoring and traceability system includes a vision component linking configuration in which one or more vision components (30) are intelligently linked together such that objects moving along the production line 10 are identified by a first vision component (30) and then become identified by a second vision component (30) while remaining identified by the first vision component (30). The first vision component (30) and the second vision component (30) are positioned at different locations along the production line 10 and are configured to communicate with each other to transfer tracking data as botles 20 move between the respective fields of view of the vision components (30).

[0117] The first vision component (30) and the second vision component (30) execute a handshake upon which intelligent data is transferred therebetween. The handshake comprises a data exchange operation in which tracking information collected by the first vision component (30) regarding identified botles 20 is transmitted to the second vision component (30). The intelligent data transferred during the handshake includes the virtual markers 32 assigned to eachbottle 20, position data, timestamp data, and other tracking information collected by the first vision component (30). Upon completion of the handshake, the second vision component (30) becomes engaged with the bottles 20 and begins collecting intelligent data on the bottles 20 as the bottles 20 continue moving along the production line 10.

[0118] According to some example embodiments, the handshake operation is triggered when a bottle 20 enters an overlapping field of view shared by the first vision component (30) and the second vision component (30). The overlapping field of view provides a transition zone in which both vision components (30) simultaneously track the bottle 20, enabling verification of the virtual marker 32 transfer before the bottle 20 exits the field of view of the first vision component (30). According to other example embodiments, the handshake operation is triggered based on position coordinates indicating that the bottle 20 has reached a predefined handshake zone along the production line 10.

[0119] The handshake mechanism enables stitching of tracking data across multiple vision components (30) positioned along the production line 10. The stitching operation combines tracking information from successive vision components (30) to create a continuous tracking record for each bottle 20 as the bottle 20 progresses through the entire production line 10. The stitching of tracking data preserves the virtual markers 32 assigned to each bottle 20 and maintains the association between each bottle 20 and the corresponding virtual marker 32 across transitions between vision component (30) coverage areas. The continuous tracking record created through the stitching operation is transmitted to the data processing system for storage and enables retrieval of the complete tracking history for any bottle 20 from initial detection through final palletization.

[0120] After the handshake is executed, the first vision component (30) identifies other bottles 20 upstream on the production line 10 and remains engaged with these upstream bottles 20. The first vision component (30) continues collecting intelligent data on the upstream bottles 20 until the collected data is sent to the second vision component (30) downstream through a subsequenthandshake operation. This sequential handshake process enables continuous tracking of bottles 20 as the bottles 20 progress through different sections of the production line 10.

[0121] The tracking data transferred during each handshake operation includes the virtual markers 32 assigned to each bottle 20, position coordinates indicating the location of each bottle 20 within the field of view of the vision component (30), velocity data indicating the speed and direction of movement of each bottle 20, timestamp data indicating when each bottle 20 was detected and tracked, and any additional metadata collected by the vision component (30) during tracking. The comprehensive transfer of tracking data during the handshake operation ensures that the receiving vision component (30) has complete information to continue tracking each bottle 20 without loss of identification or tracking continuity.

[0122] Two or more vision components (30) are provided to allow for full integration with one or more production lines. The vision components (30) are strategically positioned throughout the production line 10 such that handshaking between the vision components (30) permits identifying and tracking of groups of bottles 20 throughout the production line 10. The number and positioning of vision components (30) is selected based on the layout and configuration of the production line 10 to provide comprehensive coverage of product flow through all processing stages.

[0123] Conventional systems lack mechanisms for transferring tracking data between monitoring devices as products move along a production line. The absence of data transfer mechanisms results in loss of continuity in product identification when products exit the field of view of one monitoring device and enter the field of view of another monitoring device. This loss of continuity creates gaps in traceability data and prevents accurate tracking of products through the entire production process.

[0124] The present handshake mechanism ensures seamless data transfer between vision components (30), thereby maintaining continuous product identification throughout the entire production line 10. The handshake operation preserves the association between each bottle 20 andthe corresponding virtual marker 32 as bottles 20 transition between vision component (30) coverage areas. The continuous data transfer provided by the handshake mechanism eliminates gaps in traceability data that would otherwise occur at transitions between vision component (30) fields of view.

[0125] According to some example embodiments, the sensor 17 of the production line 10 permits passage of data between vision components 30. The vision component 30 permits the virtual marker 32 assigned to a bottle 20 to be transferred to data collected by the sensor 17. The sensor 17 receives the virtual marker 32 data and associates the virtual marker 32 with sensor data collected regarding the bottle 20 as the bottle 20 passes the sensor 17. The virtual marker 32 data is then transferred from the sensor 17 to a vision component 30 positioned further along the production line 10 and after the sensor 17. This data transfer through the sensor 17 enables the vision component 30 downstream of the sensor 17 to receive and utilize the virtual marker 32 data, maintaining continuous identification of the bottle 20 across different monitoring devices positioned along the production line 10.

[0126] The vision component 30 is mounted in various configurations to capture the bottles 20 moving along the production line 10. In one configuration, the vision component 30 is mounted in a stationary manner above the production line 10. In another configuration, the vision component 30 is mounted along a side of the production line 10. In further configurations, the vision component 30 is positioned alongside the conveyor 15, the accumulation table 40, the rejection / reinsertion table 50, or other components of the production line 10 to capture the bottles 20 as the bottles 20 move through respective sections of the production line 10.

[0127] The vision component 30 is configured to articulate and move in any desired direction to follow or track the bottles 20 or other objects moving along the production line 10. The articulation capability enables the vision component 30 to adjust orientation and viewing angle in response to the movement of the bottles 20. The movement capability enables the vision component 30 to translate position to maintain the bottles 20 within the field of view as the bottles20 progress through the production line 10. This articulation and movement capability enables the vision component 30 to track the entirety of the bottles 20 moving along the production line 10 such that intelligent data is collected thereon.

[0128] The vision component 30 is movably mounted to one or more tracks suspended above a floor of the production line 10. The tracks extend along sections of the production line 10 and provide a guided path along which the vision component 30 translates. The suspended mounting configuration positions the vision component 30 above the bottles 20 to provide a viewing angle that captures the bottles 20 from above as the bottles 20 move along the conveyor 15 or accumulate on the accumulation table 40. The movable mounting to the tracks enables the vision component 30 to follow groups of the bottles 20 as the bottles 20 progress through the production line 10, maintaining continuous visual coverage of the bottles 20 during movement.

[0129] In an alternative configuration, the vision component 30 is mounted in a stationary manner relative to the production line 10 but permits at least some amount of articulation or rotation. The stationary mounting fixes the position of the vision component 30 at a location along the production line 10. The articulation or rotation capability enables the vision component 30 to adjust the viewing angle and orientation while remaining at the fixed position. This configuration enables the vision component 30 to track the bottles 20 as the bottles 20 pass through the field of view by rotating or articulating to follow the movement of the bottles 20 without translating the position of the vision component 30.

[0130] The data set transferred from one vision component (30) to another as a group of bottles moves along the production line enables the group to remain identified throughout travel of the group for wrapping and palletizing. When a first vision component (30) identifies a group of bottles and collects tracking data including virtual markers and position information, the data set is transferred to a second vision component (30) through the handshake operation as the group progresses downstream. The second vision component (30) receives the data set and continues tracking the group using the transferred identification data. As the group proceeds throughsubsequent vision components (30) positioned along the production line, each vision component (30) receives the data set from the preceding vision component (30) and maintains the group identification. This sequential transfer of the data set preserves the identity of the group as the group moves through wrapping operations where bottles are grouped into secondary aggregates and through palletizing operations where secondary aggregates are assembled into tertiary aggregates.

[0131] All production parameters and components thereof are attained and logged during the production process. The production parameters include cap data identifying the cap applied to each bottle, preform data identifying the preform from which each bottle is formed, and label data identifying the label applied to each bottle. A time stamp and a date stamp are recorded indicating when each bottle is produced. Additional parameters at the time of production are captured including production line speed, equipment settings, and environmental conditions. Parameters relating to secondary packaging are recorded including the grouping configuration, wrapping specifications, and the time at which secondary packaging occurs. Parameters relating to palletizing are recorded including the pallet configuration, the number of secondary aggregates per pallet, and the time at which palletizing occurs.

[0132] The production parameters and component data are logged to a computer, a database, or a cloud server. The logging operation stores the production parameters in association with the identification data for each bottle and each group of bottles. The computer, database, or cloud server maintains records that correlate the production parameters with the virtual markers assigned to individual bottles and with the group identification data transferred between vision components (30). The logged data provides a comprehensive record of production conditions and component information for each bottle from initial production through secondary packaging and palletizing.

[0133] Prior systems lose continuity in product identification as products move through production stages. When products transition between different monitoring devices or processing stations in prior systems, the association between products and their identification data isinterrupted. This interruption results in incomplete traceability records where production parameters captured at one stage are not correlated with the same products at subsequent stages. The loss of continuity in prior systems prevents accurate reconstruction of the production history for individual products and groups of products.

[0134] The present system maintains persistent group identification with comprehensive parameter logging throughout the production process. The data set transferred between vision components (30) preserves the group identification as the group moves through each production stage. The production parameters captured at each stage are associated with the persistent group identification, enabling correlation of all parameters from production through palletizing. The comprehensive parameter logging ensures that complete traceability data is preserved, including cap, preform, label, time and date stamp, and all other parameters recorded during production, secondary packaging, and palletizing operations. This persistent identification and comprehensive logging provides a complete production history for each group of bottles that is accessible from the computer, database, or cloud server for traceability purposes.

[0135] The intelligent system and the traditional vision system are configured to utilize one or more vision components (30) to generate virtual markers 32 for bottles or packages on the production line. The serialization codes comprise unique identifiers printed, etched, or otherwise applied to individual bottles or packages during production. According to some example embodiments, the serialization code is applied to each bottle after filling or after passing through the filler. In some embodiments, the serialization code is applied after capping or after labeling. The timing of serialization code application may vary depending on production line configuration and customer requirements. The sensors 17 are positioned along the production line at locations where the serialization codes are visible and accessible for capture. The sensors 17 read or capture the serialization codes and transmit the serialization code data to processing components that extract and record the serialization code information.

[0136] According to some example embodiments, an operator interacts with a dedicated system periodically to permit serialized codes to be read. The dedicated system comprises an interface through which the operator initiates or authorizes serialization code reading operations. The operator accesses the dedicated system at intervals during production to enable the reading of serialization codes from bottles or packages passing through the production line. According to some example embodiments, operators are placed at relevant areas of the production line where serialization code reading is performed. The positioning of operators at these relevant areas enables the operators to access bottles or packages for serialization code reading and to interact with the dedicated system to authorize reading operations.

[0137] According to some example embodiments, a system is implemented such that all serialization codes of each bottle or package are read and captured automatically. According to some example embodiments, one or more sensors 17 are configured to read and capture the serialization codes of bottles as the bottles move along the production line 10. The sensors 17 continuously monitor bottles or packages as the bottles or packages move along the production line 10 and capture the serialization code from each bottle or package without requiring manual intervention for each individual serialization code. The sensors 17 communicate the reading and capturing of serialization codes directly with the object detection and tracking algorithm. The direct communication enables the object detection and tracking algorithm to receive serialization code data in real time and to associate the serialization code data with the corresponding bottle or package being tracked. According to some example embodiments, the sensors 17 connect the captured serialization code to the virtual marker 32 assigned to the corresponding bottle by the vision component 30, thereby establishing a link between the physical serialization code and the virtual identification code. The association between serialization codes and tracked bottles or packages enables correlation of the unique serialization identifiers with the virtual markers 32 and other tracking data maintained by the object detection and tracking algorithm. The data captured by the sensors 17 is transmitted to the data processing system for storage and subsequent retrieval,enabling the data processing system to maintain comprehensive records linking serialization codes with virtual markers 32. In example embodiments, each virtual marker 32 is associated with a respective bottle 20, wherein the bottle 20 is nearly identical in shape, size, and appearance to other bottles 20 on the production line 10 but comprises a unique serialization code that distinguishes the bottle 20 from other bottles 20.

[0138] Each serialization code is obtained while bottles are contained in a secondary aggregate. The secondary aggregate comprises a box or wrapped group containing multiple bottles. According to some example embodiments, the sensors 17 capture the serialization codes after bottles have been grouped into the secondary aggregate and before the secondary aggregate proceeds to subsequent processing stages. In some examples, an operator periodically accesses one or more bottles of the secondary aggregate and reads and captures the serialization code of one or more bottles within the secondary aggregate. The capture of serialization codes while bottles are contained in the secondary aggregate enables correlation between individual bottle identifiers and the secondary aggregate in which the bottles are packaged.

[0139] The serialization code printed or etched onto the physical bottle is interconnected with a virtual identification code generated by one or more vision components 30 through a data processing system. The data processing system comprises a server, database, cloud server, or other data processing system configured to receive serialization code data and virtual marker data and to create and maintain associations therebetween. According to some example embodiments, the data processing system permits connection of the virtual marker 32 with the serialization code without requiring the sensors 17, for example by utilizing production sequence data, timestamp correlations, or other data available to the data processing system.

[0140] According to one example embodiment, the data processing system receives the serialization code captured from the physical bottle and the virtual marker 32 assigned to the same bottle by the vision component 30, and creates a correlation record linking the serialization code to the virtual marker 32. The correlation record is stored in the data processing system and enablesretrieval of all tracking data associated with a bottle using either the serialization code or the virtual marker 32.

[0141] According to another example embodiment, the vision component 30 simultaneously captures the serialization code and associates the serialization code with the virtual marker 32 already assigned to that bottle. The vision component 30 transmits both the serialization code and the associated virtual marker 32 to the data processing system for storage and subsequent retrieval.

[0142] According to a further example embodiment, the data processing system pre-assigns or predicts which serialization code will be applied to a bottle based on production sequence data. When the vision component 30 identifies the bottle and assigns a virtual marker 32, the data processing system correlates the pre-assigned serialization code with the virtual marker 32 based on the position and timing of the bottle in the production sequence.

[0143] According to yet another example embodiment, the data processing system performs post-production correlation of serialization codes with virtual markers 32. After bottles have been grouped into secondary aggregates, the data processing system utilizes time stamp data, position data, and production sequence information to correlate serialization codes captured from bottles within the secondary aggregate with the virtual markers 32 previously assigned to those bottles by the vision components 30.

[0144] The interconnection between serialization codes and virtual identification codes enables comprehensive traceability wherein a bottle is identified and its complete production history is retrieved using either the physical serialization code visible on the bottle or the virtual marker 32 maintained in the tracking system. This dual identification capability supports traceability operations including recall management, quality control investigations, and regulatory compliance reporting.

[0145] The rejection / reinsertion table 50 operates within the production line 10 to handle bottles 20 that are rejected during production. As shown in Figure 1, the rejection / reinsertiontable 50 is positioned along the production line 10 and is configured to receive bottles 20 that are removed from the main production flow for inspection, quality control, or other purposes. The rejection / reinsertion table 50 temporarily holds rejected bottles 20 and provides a mechanism for reinserting bottles 20 back into the production flow after inspection or corrective action.

[0146] The vision components 30 capture and identify rejected bottles 20 as the bottles 20 enter the rejection / reinsertion table 50. The vision components 30 maintain identification of rejected bottles 20 through the virtual markers 32 assigned to each bottle 20 during initial tracking on the production line 10. When a bottle 20 is diverted to the rejection / reinsertion table 50, the vision components 30 continue to track the bottle 20 and preserve the association between the bottle 20 and the corresponding virtual marker 32. The serialization codes of rejected bottles 20 are also captured and maintained, enabling correlation between the rejected bottle 20 and all previously collected tracking data.

[0147] The details of a bottle 20 being reintroduced into production are obtained through one or more methods. In a first method, a vision virtual code assigned by the vision components 30 is utilized, wherein the virtual identification is maintained or reassigned upon reinsertion of the bottle 20 into the production flow. The vision components 30 recognize the bottle 20 as the bottle 20 exits the rejection / reinsertion table 50 and either maintain the existing virtual marker 32 or assign a new virtual marker 32 that is correlated with the previous identification data. In a second method, the sensor 17 reads or detects the bottle 20 as the bottle 20 is reinserted and communicates bottle data to the tracking system. The sensor 17 captures data regarding the reinserted bottle 20 and transmits this data to the object detection and tracking algorithm for integration with existing tracking records. In a third method, the vision components 30 read the serial number printed or encoded on the bottle 20, thereby enabling the system to retrieve and correlate all previously captured data for that bottle 20. The serial number provides a persistent identifier that links the reinserted bottle 20 to the complete tracking history collected prior to rejection.31

[0148] Two processing options are available for reinserted bottles 20. In a first processing option, bottles 20 are reinserted during production while recognizing the bottle code during reinsertion. The vision components 30 identify the reinserted bottle's code as the bottle 20 re-enters the production flow. The mathematical model accounts for the reinserted bottle's position within the current production flow, enabling accurate prediction of mixing levels and maintaining traceability continuity without production interruption. The mathematical model dynamically adjusts predictions to incorporate the timing and position of the reinserted bottle 20 relative to bottles 20 that remained in the primary production flow.

[0149] In a second processing option, reinserted bottles 20 are run after the end of the production run. The reinserted bottles 20 are processed as a separate batch, allowing the mathematical model to treat the reinserted bottles 20 as a distinct production segment with known time boundaries. This separation simplifies mixing calculations and provides clearer aggregate correlation since the reinserted bottles 20 do not intermingle with the primary production flow. The time boundaries of the separate batch enable the mathematical model to generate predictions for the reinserted bottles 20 independently from predictions for the primary production flow.

[0150] Each processing scenario offers distinct benefits to the mathematical model. The first scenario maintains real-time traceability but requires the mathematical model to dynamically adjust predictions to account for reinsertion timing and position. The dynamic adjustment incorporates the reinserted bottle 20 into ongoing calculations of mixing levels and aggregate composition. The second scenario provides cleaner data separation enabling higher confidence levels in pallet composition predictions. The separation of reinserted bottles 20 into a distinct batch eliminates the complexity introduced by intermixing of reinserted bottles 20 with primary production bottles 20, thereby simplifying the statistical calculations performed by the mathematical model.

[0151] Existing systems do not adequately address rejection and reinsertion rates as part of production line dynamics. Existing systems treat rejected bottles as removed from the productionflow without mechanisms for reintegrating tracking data when bottles are reinserted. This treatment results in gaps in traceability data for reinserted bottles and inaccuracies in predictions generated by mathematical models that do not account for reinsertion events.

[0152] The present system integrates rejection and reinsertion data directly into the mathematical model and the vision tracking system. The integration ensures accurate traceability even for bottles 20 that deviate from the standard production flow. The mathematical model receives data regarding rejection events and reinsertion events and incorporates this data into predictions of bottle positions and mixing levels. The vision components 30 maintain or reestablish identification of bottles 20 through the rejection and reinsertion process, preserving the association between each bottle 20 and the corresponding tracking data. This integration of rejection and reinsertion data provides comprehensive traceability coverage that accounts for all bottles 20 processed through the production line 10, including bottles 20 that are temporarily removed from and subsequently returned to the production flow.

[0153] The production line monitoring and traceability system includes a digital twin model that creates a digital replica of the production line. The digital twin model comprises a virtual representation of the physical production line that replicates the configuration, equipment, and operational characteristics of the actual production environment. The digital twin model enables manufacturers to simulate various scenarios and parameters by modeling the flow of bottles and predicting outcomes based on changes in the system.

[0154] The digital twin model incorporates physics formulas and associated data to accurately model the physical behavior of the production line. The physics formulas include equations governing conveyor dynamics that describe the movement of bottles along conveyor surfaces at varying speeds and through curved and straight sections. The digital twin model incorporates bottle momentum calculations that account for the mass and velocity of bottles as the bottles transition between different sections of the production line. Friction coefficients are modeled to represent the interaction between bottles and conveyor surfaces, accumulation table surfaces, and othercontact points throughout the production line. The digital twin model includes accumulation table flow patterns that describe how bottles distribute, collect, and mix when the bottles enter and exit accumulation areas. Gravity effects during transfers between stations are modeled to account for elevation changes and the influence of gravitational forces on bottle movement as bottles transition between different processing stages.

[0155] The digital twin model processes the physics formulas in combination with production line configuration data to generate predictions regarding bottle positions, mixing levels, and timing throughout the production process. The configuration data includes conveyor lengths, speeds, and orientations, accumulation table dimensions and surface characteristics, transfer point geometries, and equipment spacing. The digital twin model applies the physics formulas to this configuration data to simulate the movement of bottles through each section of the production line and to predict the resulting distribution of bottles at each processing stage.

[0156] Conventional systems rely on mathematical prediction models in isolation without real-time validation. The isolated mathematical prediction models operate on static parameters and historical averages without receiving feedback from actual production operations. This isolation results in divergence between predicted bottle positions and actual bottle positions as production conditions vary from the assumptions embedded in the isolated models. The absence of real-time validation in conventional systems leads to inaccuracies in tracking product positions and timing that accumulate over the course of production operations.

[0157] The digital twin model continuously validates predictions against actual production data. The digital twin model receives real-time data from sensors, vision components, and other monitoring devices positioned along the production line. The real-time data includes actual bottle positions, conveyor speeds, accumulation table fill levels, and timing measurements from production operations. The digital twin model compares the predictions generated by the physics -based simulation with the actual production data received from the monitoring devices. When discrepancies between predicted values and actual values are detected, the digital twin modeladjusts the simulation parameters to align the predictions with observed production conditions. This continuous validation process maintains accuracy in the predictions generated by the digital twin model throughout production operations.

[0158] Prior systems employ a single tracking methodology and struggle to adapt to varying production configurations. The single tracking methodology of prior systems is calibrated for a specific production line configuration and does not accommodate changes in equipment layout, conveyor speeds, or processing sequences. When production configurations change, prior systems require recalibration or replacement of the tracking methodology, resulting in downtime and reduced traceability accuracy during transition periods.

[0159] The digital twin model integrates multiple data inputs and is reconfigurable to match different customer requirements. The multiple data inputs include sensor data, vision system data, production metrics, equipment parameters, and operator observations. The digital twin model processes these multiple data inputs concurrently to generate predictions that account for the full range of factors influencing bottle movement and mixing. The reconfigurability of the digital twin model enables manufacturers to modify the virtual representation to match changes in production line configuration, equipment settings, or processing sequences. The digital twin model is updated with new configuration data and adjusted physics parameters to reflect the modified production environment, enabling accurate predictions without requiring replacement of the underlying simulation framework.

[0160] The digital twin model is trained with historical data such that the model becomes more accurate, precise, reliable, and representative of the production line. The historical data includes records of past production operations including bottle positions, timing measurements, mixing patterns, and equipment performance data collected over extended periods of production. The digital twin model processes the historical data to identify patterns and correlations between production parameters and bottle behavior. The training process adjusts the physics parameters and simulation algorithms of the digital twin model based on the patterns identified in the historicaldata. As the digital twin model is trained with increasing amounts of historical data, the predictions generated by the digital twin model converge toward the actual behavior observed in the production line, thereby improving the accuracy, precision, and reliability of the predictions.

[0161] The digital twin model enables predictions at the accumulation table, at other areas of the machine, and in secondary or tertiary aggregates. At the accumulation table, the digital twin model predicts how bottles distribute across the accumulation surface, how bottles mix as the bottles enter and exit the accumulation area, and how the fill level of the accumulation table changes over time. At other areas of the machine, the digital twin model predicts bottle positions along conveyors, timing of bottle arrivals at processing stations, and the effects of equipment stoppages on bottle flow. In secondary aggregates, the digital twin model predicts which bottles are grouped together in boxes or wrapped groups based on the timing and position of bottles as the bottles enter packaging operations. In tertiary aggregates, the digital twin model predicts which secondary aggregates are assembled onto each pallet based on the timing and sequence of packaging operations. These predictions at each level of the production process and aggregate hierarchy improve overall traceability of bottles throughout the production process by providing position and grouping information that correlates individual bottles with their containing aggregates.

[0162] According to some example embodiments, the digital twin model receives real-time feedback from the vision components 30 to update simulation parameters. The real-time feedback includes actual bottle positions, movement velocities, and accumulation patterns observed by the vision components 30. The digital twin model compares the real-time feedback with simulated predictions and adjusts simulation parameters including friction coefficients, flow rates, and timing estimates to improve alignment between simulated behavior and actual production line behavior.

[0163] The production line monitoring and traceability system includes a synergistic combination of the mathematical model with vision systems to provide enhanced prediction capabilities. The mathematical model is combined with either the intelligent system or thetraditional vision system to enhance prediction of mixing levels on accumulation tables and mass flows throughout the production line.

[0164] The combination of the mathematical model with the intelligent system integrates statistical prediction methods with real-time visual tracking data generated by deep learning algorithms. The mathematical model receives production parameters including accumulation belt speeds, production metrics, and timing data to generate statistical predictions of bottle positions and mixing levels. The intelligent system provides real-time visual tracking data including actual bottle positions, movement patterns, and accumulation table fill levels captured through object detection and tracking algorithms. The mathematical model processes the real-time visual tracking data from the intelligent system to validate and refine the statistical predictions. When the intelligent system detects bottle positions or mixing patterns that differ from the predictions generated by the mathematical model, the mathematical model adjusts the predictions based on the observed data. This integration enables the mathematical model to maintain accuracy by continuously incorporating actual production conditions captured by the intelligent system.

[0165] The combination of the mathematical model with the traditional vision system integrates statistical prediction methods with image processing data generated by grayscale image capture and tracking models. The traditional vision system captures grayscale images of bottles on the production line and processes the images using tracking models to identify and track bottles. The mathematical model receives the tracking data from the traditional vision system and incorporates this data into the statistical calculations used to predict mixing levels and mass flows. The traditional vision system provides position data and movement data that the mathematical model uses to validate predictions and to adjust calculations based on observed bottle behavior. This integration enables the mathematical model to operate with validated input data derived from actual visual observations of the production line.

[0166] The synergistic combination of the mathematical model with vision systems provides a framework for traceability that integrates statistical prediction with real-time visual tracking data.The statistical prediction capabilities of the mathematical model provide baseline estimates of bottle positions, mixing levels, and timing based on production parameters and historical patterns. The real-time visual tracking data provided by the vision systems validates these baseline estimates and provides correction data when actual production conditions deviate from predicted conditions. The integration of statistical prediction with real-time visual tracking enables the system to maintain accurate traceability records even when production conditions vary from expected parameters.

[0167] The prediction of mixing levels on accumulation tables is enhanced through the synergistic combination. The mathematical model generates predictions of how bottles mix as the bottles enter, accumulate, and exit accumulation tables based on accumulation belt speeds, fill levels, and timing parameters. The vision systems capture actual mixing patterns by tracking individual bottles through virtual markers as the bottles move across accumulation table surfaces. The mathematical model compares the predicted mixing patterns with the actual mixing patterns observed by the vision systems and adjusts the prediction algorithms based on the comparison. This continuous comparison and adjustment process improves the accuracy of mixing level predictions over time as the mathematical model incorporates observed mixing behavior into the statistical calculations.

[0168] The prediction of mass flows throughout the production line is enhanced through the synergistic combination. The mathematical model generates predictions of bottle flow rates, timing, and distribution based on conveyor speeds, production volumes, and equipment configurations. The vision systems capture actual flow data by tracking bottles as the bottles move through different sections of the production line. The mathematical model receives the actual flow data from the vision systems and validates the predicted flow rates and timing against the observed values. When discrepancies between predicted and observed flow data are detected, the mathematical model adjusts the flow predictions to align with actual production conditions. Thisvalidation and adjustment process enables accurate prediction of mass flows that reflects current production operations rather than static assumptions.

[0169] Systems that rely on a single tracking methodology encounter difficulties in adapting to varying production configurations or customer requirements. Single methodology systems are calibrated for specific production line configurations and operate with fixed parameters that do not accommodate changes in equipment layout, conveyor speeds, accumulation table configurations, or processing sequences. When production configurations change or when different customer requirements are introduced, single methodology systems require recalibration or replacement of the tracking approach, resulting in reduced traceability accuracy during transition periods and increased implementation costs for different production environments.

[0170] The synergistic combination of mathematical modeling with vision systems enables adaptive tracking that is tailored to specific production needs and customer requirements. The mathematical model is configured with parameters that reflect the specific production line configuration, including conveyor speeds, accumulation table dimensions, equipment spacing, and processing sequences. The vision systems are positioned and configured to capture visual data from the specific production environment. The combination of the mathematical model with the vision systems enables the system to adapt to changes in production configuration by updating the mathematical model parameters and adjusting the vision system positioning. The flexibility of the combined approach enables manufacturers to tailor the tracking system to specific production needs by selecting appropriate mathematical model parameters and vision system configurations. The combined approach also enables customization based on customer requirements including production volume targets, risk tolerance levels, and traceability precision specifications. This adaptive capability enables the synergistic combination to provide accurate traceability across diverse production environments and varying customer requirements without requiring replacement of the underlying tracking methodology.

[0171] The production line monitoring and traceability system includes a synergistic combination of the mathematical model with the digital twin model to provide enhanced prediction of mixing levels. The mathematical model is combined with the digital twin model to simulate real-world scenarios and to analyze data from the production line, thereby enabling manufacturers to gain deeper insights into operations and to make data-driven decisions to improve efficiency.

[0172] The combination of the mathematical model with the digital twin model integrates statistical prediction methods with physics-based simulation capabilities. The mathematical model employs statistical methods to predict bottle positions based on production parameters including accumulation belt speeds, production metrics, and timing data. The digital twin model creates a virtual representation of the production line that incorporates physics formulas governing conveyor dynamics, bottle momentum, friction coefficients, accumulation table flow patterns, and gravity effects during transfers. The integration of the mathematical model with the digital twin model enables predictions that combine statistical analysis of production data with physics -based simulation of bottle behavior throughout the production line.

[0173] The synergistic combination enhances prediction of mixing levels by simulating real-world scenarios within the digital twin model and analyzing the simulation results using the statistical methods of the mathematical model. The digital twin model simulates scenarios including variations in conveyor speeds, changes in accumulation table fill levels, equipment stoppages, and transitions between different production configurations. The mathematical model analyzes the simulation results generated by the digital twin model to identify patterns in mixing behavior and to generate predictions of mixing levels under different operational conditions. The combination of simulation capabilities with statistical analysis enables predictions that account for the range of scenarios encountered during actual production operations.

[0174] The digital twin model provides the mathematical model with simulated data representing production line behavior under various conditions. The simulated data includes bottle positions, timing measurements, and mixing patterns generated by the physics -based simulationfor different operational scenarios. The mathematical model processes the simulated data to refine the statistical algorithms used to predict mixing levels and bottle positions. The refinement of the statistical algorithms based on simulated data enables the mathematical model to generate predictions that are informed by the physics-based understanding of production line behavior embodied in the digital twin model.

[0175] The mathematical model provides the digital twin model with statistical parameters derived from analysis of production data. The statistical parameters include probability distributions for bottle positions, confidence intervals for timing predictions, and correlation coefficients relating production parameters to mixing levels. The digital twin model incorporates the statistical parameters into the physics-based simulation to generate predictions that reflect both the physical behavior of the production line and the statistical patterns observed in production data. The incorporation of statistical parameters into the digital twin model enables simulations that are calibrated to match actual production behavior as captured in the statistical analysis performed by the mathematical model.

[0176] The synergistic approach enables manufacturers to gain deeper insights into operations by providing predictions that integrate multiple analytical perspectives. The mathematical model provides statistical insights regarding the probability of different outcomes based on historical patterns and production parameters. The digital twin model provides physical insights regarding the mechanisms driving bottle movement and mixing based on the physics of conveyor dynamics and accumulation table behavior. The combination of statistical insights with physical insights enables manufacturers to understand both the likelihood of different outcomes and the underlying causes of the predicted behavior. This understanding supports informed decision -making regarding production line configuration, equipment settings, and operational procedures.

[0177] The synergistic approach enables manufacturers to make data-driven decisions to improve efficiency by providing predictions that support evaluation of different operational strategies. The digital twin model simulates the effects of proposed changes to production lineconfiguration or operational procedures. The mathematical model analyzes the simulation results to predict the impact of the proposed changes on mixing levels, timing, and traceability accuracy. Manufacturers evaluate the predictions generated by the combined system to assess whether proposed changes are likely to improve efficiency and to select the operational strategies that provide the greatest benefit. The data-driven evaluation enabled by the synergistic combination supports optimization of production operations based on quantitative predictions rather than assumptions or estimates.

[0178] Conventional systems employ a single tracking methodology that encounters difficulties with variable conveyor speeds and machine stoppages. Single methodology systems are calibrated for specific operational conditions and generate predictions based on fixed assumptions regarding conveyor speeds and continuous equipment operation. When conveyor speeds vary during production or when machines stop and restart, single methodology systems generate predictions that do not reflect actual production conditions. The discrepancy between predicted conditions and actual conditions results in inaccuracies in tracking bottle positions and mixing levels during periods of variable operation.

[0179] The combination of mathematical modeling with digital twin simulation accounts for production line dynamics and enables accurate predictions even under varying operational conditions. The digital twin model incorporates physics formulas that describe bottle behavior under variable conveyor speeds, including the effects of acceleration, deceleration, and speed transitions on bottle positions and mixing patterns. The digital twin model simulates the effects of machine stoppages on bottle accumulation, including the filling and emptying of accumulation tables that occurs when upstream or downstream equipment stops and restarts. The mathematical model analyzes the simulation results to generate predictions that account for the dynamics introduced by variable conveyor speeds and machine stoppages. The integration of dynamic simulation with statistical analysis enables the combined system to maintain prediction accuracyacross the range of operational conditions encountered during production, including periods of variable conveyor speeds and intermittent machine stoppages.

[0180] The production line monitoring and traceability system includes a comprehensive integrated configuration in which the mathematical model, the digital twin model, and the intelligent system or traditional vision system are combined to create a system for predicting mixing levels and ensuring accurate traceability throughout the production process. The comprehensive integrated configuration combines statistical prediction methods, physics -based simulation capabilities, and real-time visual tracking to provide a framework that addresses traceability requirements across all stages of production.

[0181] The comprehensive integrated configuration operates through coordinated data exchange between the mathematical model, the digital twin model, and the vision system. The mathematical model receives production parameters and generates statistical predictions of bottle positions and mixing levels based on accumulation belt speeds, production metrics, and timing data. The digital twin model receives the statistical predictions from the mathematical model and incorporates the predictions into physics-based simulations that model conveyor dynamics, bottle momentum, friction coefficients, and accumulation table flow patterns. The intelligent system or traditional vision system captures real-time visual data of actual bottle positions and movements and transmits this data to both the mathematical model and the digital twin model. The mathematical model and the digital twin model process the real-time visual data to validate predictions and to adjust calculations based on observed production conditions.

[0182] The comprehensive integrated configuration achieves specific mixing values and predictability metrics that enable precise traceability throughout the production process. In one example configuration, the comprehensive integrated system achieves mixing level predictions with confidence levels exceeding 95% for determining which bottles are contained within a given pallet. The system predicts that bottles found on the same pallet are produced within a maximum time window of 28 minutes with 95% confidence, enabling manufacturers to identify theproduction time range for all bottles contained in any given pallet. The comprehensive integrated configuration further enables prediction of bottle locations within a reduced number of pallets. In one example, the system predicts the location of a specific bottle within three or fewer pallets with 98% confidence, enabling targeted recall operations that affect a minimal number of pallets when quality issues are identified. In another example, the system predicts the location of a specific bottle within a single pallet with 85% confidence under standard production conditions, enabling precise identification of affected products without requiring inspection of multiple pallets.

[0183] The comprehensive integrated configuration enhances operational efficiency by providing manufacturers with real-time visibility into production operations and predictive capabilities that support proactive decision-making. The integration of statistical modeling with physics-based simulation and visual tracking enables the system to generate predictions that reflect current production conditions rather than historical averages or static assumptions. Manufacturers utilize the predictions generated by the comprehensive integrated system to optimize production line configurations, to adjust equipment settings in response to changing conditions, and to identify potential issues before the issues affect product quality or traceability accuracy. The real-time data provided by the vision system enables immediate detection of deviations from expected production behavior, and the mathematical model and digital twin model process this data to generate updated predictions that account for the detected deviations.

[0184] The comprehensive integrated configuration provides manufacturers with tools to respond swiftly to market demands while maintaining high standards of product quality and safety. When market demands require changes to production volumes, product configurations, or processing sequences, the comprehensive integrated system adapts to the changed conditions through coordinated updates to the mathematical model parameters, digital twin model configuration, and vision system positioning. The digital twin model simulates the effects of proposed changes before implementation, enabling manufacturers to evaluate the impact of changes on mixing levels and traceability accuracy. The mathematical model analyzes thesimulation results to predict the traceability metrics achievable under the proposed configuration. The vision system provides real-time validation of predictions after changes are implemented, enabling rapid confirmation that the changed configuration achieves the expected traceability performance.

[0185] The comprehensive integrated configuration maintains high standards of product quality and safety by providing complete traceability records that correlate individual products with production parameters, component data, and aggregate groupings throughout the production process. When quality issues are identified, the comprehensive integrated system enables rapid identification of affected products based on the traceability records maintained by the system. The mixing level predictions generated by the comprehensive integrated system enable manufacturers to determine the scope of affected products with precision, limiting recall operations to the specific pallets containing potentially affected bottles rather than requiring broader recalls that affect products not involved in the quality issue.

[0186] Conventional approaches face multiple limitations that reduce traceability accuracy and operational efficiency. Conventional approaches struggle with maintaining identification when products accumulate or mix during production operations. When products enter accumulation tables and redistribute across accumulation surfaces, conventional approaches lose track of individual product identities and fail to maintain associations between products and their assigned identifiers. Conventional approaches encounter difficulties correlating products across aggregation levels. Tracking individual containers within boxes and tracking boxes within pallets presents challenges for conventional approaches when hierarchical relationships between aggregate levels are not maintained with sufficient granularity. Conventional approaches lack realtime validation for mathematical models. Mathematical prediction models used in isolation in conventional approaches operate on static parameters without receiving feedback from actual production operations, resulting in divergence between predicted positions and actual positions as production conditions vary. Conventional approaches exhibit inability to adapt to varyingproduction configurations. Single methodology tracking systems in conventional approaches are calibrated for specific production line configurations and do not accommodate changes in equipment layout, conveyor speeds, or processing sequences without recali bration or replacement.

[0187] The comprehensive integrated system addresses each of these limitations through the synergistic combination of statistical modeling, physics -based digital twin simulation, and AI-powered vision tracking. The Al-powered vision tracking maintains identification when products accumulate or mix by assigning persistent virtual markers to individual products and tracking the virtual markers throughout accumulation and mixing operations. The statistical modeling correlates products across aggregation levels by maintaining hierarchical relationships with minute-level timestamp granularity that enables precise correlation between primary aggregates, secondary aggregates, and tertiary aggregates. The physics -based digital twin simulation provides real-time validation for the mathematical model by comparing predictions generated by the statistical algorithms with simulation results generated by the physics -based model and with actual production data captured by the vision system. The synergistic combination adapts to varying production configurations by enabling coordinated updates to statistical model parameters, digital twin model configuration, and vision system positioning when production configurations change. The comprehensive integrated system thereby provides accurate traceability across diverse production environments and varying operational conditions through the coordinated operation of statistical modeling, physics-based simulation, and Al-powered visual tracking.

[0188] The production line monitoring and traceability system includes a data processing system that serves as a central data processing and storage infrastructure for the traceability operations. The data processing system comprises one or more servers, processors, databases, and communication interfaces configured to receive, process, store, and transmit data from the various components of the traceability system. According to some example embodiments, the data processing system is implemented as a cloud-based server infrastructure accessible via network connections. According to other example embodiments, the data processing system is implementedas an on-premises server system located at or near the production facility. The data processing system includes one or more processors configured to execute instructions stored in memory for performing the data processing and correlation operations described herein.

[0189] The data processing system is communicatively coupled to the vision components 30, the mathematical model, and the digital twin model. According to some example embodiments, the data processing system is also communicatively coupled to one or more sensors 17. The communicative coupling comprises wired connections, wireless connections, network connections, or combinations thereof that enable bidirectional data transfer between the data processing system and the connected components. The data processing system receives image data and virtual marker data from the vision components 30, production parameter data from production line equipment, and prediction data from the mathematical model and digital twin model. According to some example embodiments, the data processing system also receives serialization code data from the sensors 17.

[0190] According to some example embodiments, the communicative coupling between the data processing system and the vision components 30 utilizes industrial communication protocols including MQTT (Message Queuing Telemetry Transport), REST API (Representational State Transfer Application Programming Interface), or OPC-UA (Open Platform Communications Unified Architecture). The selection of communication protocol depends on latency requirements, data throughput needs, and compatibility with existing production line infrastructure.

[0191] The mathematical model is implemented as a software module executed by one or more processors of the data processing system. According to some example embodiments, the mathematical model comprises algorithms stored in non-transitory computer-readable memory and executed by the processors to generate statistical predictions of bottle positions based on production parameters. The mathematical model receives input data from the data processing system including accumulation belt speeds, production volumes, timing data, loss / waste approximations, and carry-over data between production periods. The mathematical modelprocesses the input data using statistical algorithms to generate output data including predicted bottle positions, confidence levels, and maximum delay values for pallets.

[0192] The data processing system stores the output data generated by the mathematical model in one or more databases. The databases comprise relational databases, non-relational databases, or other data storage structures configured to maintain records of predictions, confidence levels, and timing data associated with production operations. The stored output data is retrievable by the data processing system for subsequent analysis, reporting, and correlation with other traceability data including virtual markers 32 and serialization codes.

[0193] According to some example embodiments, the traceability system comprises a hybrid configuration that combines vision components 30, sensors 17, and data processing system correlation to provide multiple methods for linking virtual markers 32 with serialization codes. In the hybrid configuration, the vision components 30 assign virtual markers 32 to bottles as the bottles are detected, and the correlation between virtual markers 32 and serialization codes is performed using one or more of: sensors 17 reading the physical serialization codes from the bottles, vision components 30 capturing the serialization codes from the bottles, or the data processing system correlating the virtual markers 32 with serialization codes based on production sequence data without reading the physical codes. The hybrid configuration provides redundancy and flexibility, enabling the system to maintain accurate traceability even when one correlation method is unavailable or unreliable. For example, if a serialization code on a bottle is damaged or obscured such that sensors 17 or vision components 30 cannot read the code, the data processing system correlates the virtual marker 32 with the serialization code based on production sequence timing. According to some example embodiments, the hybrid configuration enables manufacturers to select the most appropriate correlation method based on production line configuration, accuracy requirements, and available hardware.

[0194] According to one example embodiment, the serialization code is known at the time of application by the production equipment. The production equipment, such as a labeler, printer, orother serialization device, applies serialization codes to bottles in a known sequence, and this sequence information is available to the data processing system. When the vision component 30 first obtains visibility of a bottle and assigns a virtual marker 32, the data processing system correlates the virtual marker 32 with the serialization code that was applied to that bottle based on production sequence timing. This correlation is made internally by the data processing system without requiring sensors 17 or vision components 30 to read the physical serialization code from the bottle.

[0195] According to another example embodiment, the serialization code is automatically generated by the data processing system based on production parameters or a pre-set configuration. The data processing system generates serialization codes according to a defined sequence, pattern, or algorithm, and transmits the generated serialization codes to the production equipment for application to the bottles. Because the data processing system generates the serialization codes, the data processing system maintains a record of which serialization code is applied to each bottle based on production timing. When the vision component 30 assigns a virtual marker 32 to a bottle, the data processing system correlates the virtual marker 32 with the corresponding generated serialization code based on the time at which the bottle was detected by the vision component 30.

[0196] According to some example embodiments, the entire traceability system operates using only vision components 30 without requiring sensors 17 or other devices to read the serialization codes from the bottles. The vision components 30 assign virtual markers 32 to bottles as the bottles are detected, and the data processing system correlates the virtual markers 32 with the serialization codes based on production sequence data, timing information, or automatically generated serialization code records. This configuration reduces hardware requirements and enables traceability even when serialization codes are not readable due to damage, obscuration, or other factors.

[0197] The data processing system correlates the virtual markers 32 assigned by the vision components 30 with the serialization codes captured by the sensors 17 and with the predictionsgenerated by the mathematical model. The correlation operation creates association records that link each bottle with the corresponding virtual marker 32, serialization code, predicted position data, confidence level, and timestamp data. The association records are stored in the databases of the data processing system and enable comprehensive traceability queries wherein a bottle can be identified and its complete production history retrieved using any of the associated identifiers or parameters.

[0198] The data processing system transmits real-time data from the vision components 30 to the mathematical model for validation and refinement of predictions. According to some example embodiments, the data processing system also transmits real-time data from the sensors 17 to the mathematical model. The real-time data includes actual bottle positions observed by the vision components 30 and timing measurements captured during production operations. According to some example embodiments, the real-time data also includes production volumes confirmed by the sensors 17. According to some example embodiments, the sensors 17 capture the serialization codes. The mathematical model compares the real-time data with the predicted values and adjusts the prediction algorithms based on detected discrepancies. The adjusted predictions are transmitted back to the data processing system for storage and for use in subsequent traceability operations.

[0199] The data processing system interfaces with the digital twin model to provide production data for simulation and to receive simulation results for integration with the mathematical model predictions. The digital twin model receives production line configuration data, physics parameters, and real-time operational data from the data processing system. The digital twin model processes the received data to generate simulations of bottle movement, mixing patterns, and accumulation behavior. The simulation results are transmitted to the data processing system for storage and for comparison with the predictions generated by the mathematical model.

[0200] According to some example embodiments, the data processing system includes application programming interfaces (APIs) that enable external systems to access traceability data. The APIs provide standardized interfaces through which external systems submit queries andreceive responses containing traceability information including bottle identification data, production parameters, aggregate correlations, and prediction confidence levels. The APIs support traceability operations including recall management, regulatory compliance reporting, and quality control investigations.

[0201] The data processing system maintains audit logs recording all data transactions, correlation operations, and prediction updates performed by the system. The audit logs comprise timestamped records of data received from vision components 30 and sensors 17, correlation records created between virtual markers 32 and serialization codes, predictions generated by the mathematical model, and queries submitted through the APIs. The audit logs are stored in the databases of the data processing system and provide a complete record of traceability operations for regulatory compliance and quality assurance purposes.

[0202] The systems and methods described herein may be implemented in any form of computing or electronic device. The term "computer," as used herein, encompasses any device with processing capabilities sufficient to execute instructions. This includes, but is not limited to, personal computers, servers, mobile devices, personal digital assistants, and similar devices.

[0203] Such devices may include one or more processors, such as microprocessors, controllers, or other suitable types of processors, capable of executing instructions to control the device's operation. For example, in some implementations using a system -on-a-chip architecture, the processors may include fixed-function blocks (hardware accelerators) that perform parts of the method in hardware rather than software or firmware. Platform software, such as an operating system or similar, may be installed to support the execution of application software.

[0204] The described functionality may be implemented in hardware, software, or any combination thereof. When implemented in software, the instructions or code can be stored on or transmitted via a computer-readable medium. Such media include computer-readable storage media, which may be volatile or non-volatile, removable or non-removable, and implemented using any technology for storing information such as program code, data structures, or other data.Examples include, but are not limited to, ROM, EEPROM, RAM, magnetic or optical storage, flash memory, or any other storage medium accessible by a computer. Communication media that facilitate the transfer of software, such as via coaxial cables, fiber optics, DSL, or wireless signals, may also be considered part of computer-readable media.

[0205] Alternatively, or in addition, some or all of the described functionality may be implemented using hardware logic components. Examples include, but are not limited to, application-specific integrated circuits, system-on-a-chip systems, field-programmable gate arrays, application-specific standard products, and complex programmable logic devices. In some cases, software instructions may also be implemented in dedicated circuits, such as programmable logic arrays or digital signal processors.

[0206] The computing device may operate as a standalone system or as part of a distributed system, where tasks are performed collectively by multiple devices connected via a network. Such devices may communicate over a network connection to perform the described functionality. For instance, software may be stored on a remote computer and accessed by a local device, which may download and execute portions of the software as needed. Similarly, some instructions may be processed locally, while others may execute on remote systems or networks. In some cases, the computing device may be remote and accessible via a communication interface. Storage of program instructions may also be distributed across a network or stored in a combination of local and remote locations. For example, software may reside on a remote computer and be accessed by a local terminal, or the system may execute some software locally while other components operate on remote servers.

[0207] Features of any of the examples or embodiments outlined above may be combined to create additional examples or embodiments without losing the intended effect. It should be understood that the description of an embodiment or example provided above is by way of example only, and various modifications could be made by one skilled in the art. Furthermore, one skilled in the art will recognize that numerous further modifications and combinations of various aspectsare possible. Accordingly, the described aspects are intended to encompass all such alterations, modifications, and variations that fall within the scope of the appended claims.

Claims

CLAIMS1. A computer-implemented method for tracing production of goods on a production line (10), the method comprising:capturing, by a vision component (30), image data of bottles (20) on the production line (10);processing the image data using an object detection algorithm to detect the bottles (20); assigning virtual markers (32) to the detected bottles (20); andtracking the bottles (20) by maintaining an association between the bottles (20) and the assigned virtual markers (32) as the bottles (20) move along the production line (10).

2. A method according to claim 1, further comprising:generating, by a mathematical model executed by one or more processors of a data processing system, predictions of positions of the bottles (20) on the production line (10) based on production parameters; whereinthe production parameters comprise at least one of: a speed of an accumulation belt, a production volume, a time minimum between stations, and a carry-over between production periods.

3. A method according to claim 1 or 2, further comprising:identifying, by a first vision component (30), a group of the bottles (20) moving along the production line (10);executing a handshake between the first vision component (30) and a second vision component (30); whereinthe handshake comprises transferring tracking data including the virtual markers (32) assigned to the group of the bottles (20) from the first vision component (30) to the second vision component (30); andcontinuing tracking of the group of the bottles (20) by the second vision component (30) using the transferred tracking data.

4. A method according to any of claims 1 to 3, further comprising:tracking the bottles (20) as the bottles (20) accumulate on an accumulation table (40); whereinthe tracking comprises maintaining the association between each of the bottles (20) and the corresponding virtual markers (32) as the bottles (20) mix and redistribute in a non-linear arrangement on the accumulation table (40).

5. A method according to any of claims 1 to 4, further comprising:capturing serialization codes of the bottles (20) on the production line (10); and associating the captured serialization codes with the virtual markers (32) assigned to the respective bottles (20).

6. A method according to any of claims 1 to 5, further comprising:organizing the bottles (20) into a hierarchical aggregate structure comprising a primary aggregate, a secondary aggregate containing a plurality of primary aggregates, and a tertiary aggregate containing a plurality of secondary aggregates; andrecording timestamp data with minute granularity for each aggregate level to enable correlation of the bottles (20) across the hierarchical aggregate structure.

7. A method according to any of claims 1 to 6, further comprising:diverting one or more of the bottles (20) to a rejection / reinsertion table (50);maintaining identification of the diverted bottles (20) through the assigned virtual markers (32) while the bottles (20) are on the rejection / reinsertion table (50); andreintegrating the diverted bottles (20) into the production line (10) while preserving the association between the reintegrated bottles (20) and the corresponding virtual markers (32).

8. A production line monitoring system according to claim 1, the system comprising: a production line (10) comprising a conveyor (15) configured to transport bottles (20); a vision component (30) positioned to capture image data of the bottles (20) on the production line (10); andcontrol means configured for performing the method according to claim 1.

9. A system according to claim 1 or 2, the system comprising:a production line (10) comprising a conveyor (15) configured to transport bottles (20); a vision component (30) positioned to capture image data of the bottles (20) on the production line (10); andcontrol means configured for performing the method according to claim 2; wherein the control means comprises a mathematical model processing unit configured to generate the predictions of positions of the bottles (20).

10. A system according to any of claims 1 to 3, the system comprising:a production line (10) comprising a conveyor (15) configured to transport bottles (20); a first vision component (30) and a second vision component (30) positioned at different locations along the production line (10); whereinthe first vision component (30) and the second vision component (30) are configured to execute the handshake to transfer the tracking data; andcontrol means configured for performing the method according to claim 3.

11. A system according to any of claims 1 to 4, the system comprising:a production line (10) comprising a conveyor (15) configured to transport bottles (20); an accumulation table (40) configured to receive the bottles (20) from the conveyor (15); a vision component (30) positioned to capture image data of the bottles (20) on the accumulation table (40); andcontrol means configured for performing the method according to claim 4.

12. A system according to any of claims 1 to 5, the system comprising:a production line (10) comprising a conveyor (15) configured to transport bottles (20); a vision component (30) positioned to capture image data of the bottles (20) on the production line (10);a serialization code reader configured to capture serialization codes of the bottles (20); and control means configured for performing the method according to claim 5.

13. A system according to any of claims 1 to 7, the system comprising:a production line (10) comprising a conveyor (15) configured to transport bottles (20); a rejection / reinsertion table (50) configured to receive bottles (20) diverted from the production line (10);a vision component (30) positioned to capture image data of the bottles (20) on the rejection / reinsertion table (50); andcontrol means configured for performing the method according to claim 7.

14. A system according to any of claims 1 to 5, the system comprising:a production line (10) comprising a conveyor (15) configured to transport bottles (20);a vision component (30) positioned to capture image data of the bottles (20) on the production line (10);a data processing system communicatively coupled to the vision component (30) and configured to receive virtual marker data and serialization code data; whereinthe data processing system is configured to create and store correlation records linking the virtual markers (32) assigned to the bottles (20) with the serialization codes of the respective bottles (20); andcontrol means configured for performing the method according to claim 5.

15. A computer program comprising instructions which, when executed by a computer, cause the computer to carry out the method according to any of claims 1 to 7.

16. A computer-readable medium comprising instructions which, when executed by a computer, cause the computer to carry out the method according to any of claims 1 to 7.

17. A system according to any of claims 1 to 7, the system comprising:a production line (10) comprising a conveyor (15) configured to transport bottles (20); a vision component (30) positioned to capture image data of the bottles (20) on the production line (10);a digital twin model configured to create a virtual representation of the production line (10) and to simulate scenarios for predicting outcomes based on changes in production parameters; and control means configured for performing the method according to any of claims 1 to 7.

18. A method according to claim 7, wherein reintegrating the diverted bottles (20) comprises:recognizing a code of each reintegrated bottle (20) during reinsertion; and whereina mathematical model dynamically adjusts predictions to incorporate timing and position of the reintegrated bottles (20) relative to bottles (20) that remained in a primary production flow.

19. A method according to claim 7, wherein reintegrating the diverted bottles (20) comprises processing the diverted bottles (20) as a separate batch after completion of a production run.

20. A method according to claim 2, wherein the mathematical model provides an approximation of a confidence level and an associated maximum delay of a pallet; wherein the maximum delay represents a maximum time difference between production of bottles (20) found on a same pallet.

21. A method according to claim 1, wherein the bottles (20) are identical in shape, size, and appearance; and whereintracking the bottles (20) comprises identifying distinguishing features, characteristics, and relative orientations of the bottles (20) to maintain identification and distinguish between the identical bottles (20); and whereinthe virtual markers (32) assigned to the bottles (20) remain consistent through continuous capturing of the distinguishing features, characteristics, and relative orientations of the bottles (20).

22. A method according to claim 1, wherein the vision component (30) is configured to simultaneously evaluate and track more than 15,000 bottles (20); whereinthe simultaneous evaluation and tracking enables real-time traceability during high-volume production operations.

23. A method according to claim 5, wherein the data processing system includes application programming interfaces (APIs) enabling external systems to submit queries and receive traceability information; whereinthe traceability information comprises bottle identification data, aggregate correlations, and prediction confidence levels.

24. A method according to claim 21, wherein the vision component (30) creates a feature signature for each bottle (20) based on the identified distinguishing features; whereinthe feature signature enables re-identification of the bottle (20) after temporary exit from the field of view of the vision component (30).

25. A method according to claim 5, wherein the serialization codes are correlated with the virtual markers (32) by the data processing system based on production sequence data at the time the vision component (30) first obtains visibility of the bottles (20); whereinthe correlation is performed without requiring reading of the physical serialization codes from the bottles (20).

Images