Full-process tracing system of radio frequency coaxial connector production line

Abstract

The invention relates to the technical field of data processing and tracing in the production process, and discloses a full-process tracing system for a radio frequency coaxial connector production line, which is characterized in that a golden rhythm template containing physical feasible time constraint is preset, and a checking gateway is used for carrying out prepositive admission review on an acquired timestamp sequence; according to the method and the system, the traceability information is analyzed, individual rhythm fingerprints are generated only for the sequence passing the examination, the difference RDS is calculated, meanwhile, an unknown anomaly recognition rule for performing rejection classification and isolation evidence storage on a low-confidence-coefficient matching result is established, the traceability information is improved from discrete static event logs to quantitative characterization of the dynamic stability of the production process, and the method and the system have the advantages that the method and the system are easy to implement. A manager can perceive and distinguish internal process quality differences among different qualified products, and a decision basis is provided for diagnosing a known deviation mode and discovering unknown abnormal risks.

Description

Technical Field

[0001] This invention relates to a full-process traceability system for an RF coaxial connector production line, belonging to the field of production process data processing and traceability technology. Background Technology

[0002] In the management of RF coaxial connector production lines, Manufacturing Execution Systems (MES) are commonly used to trace the entire production process. These systems provide basic data logs for production control by recording discrete events of products at various preset checkpoints, which constitutes the current common practice for data processing in this field.

[0003] However, for application scenarios with higher requirements for long-term product reliability, the limitations of the aforementioned data management method based on discrete event recording become apparent. Specifically, the system only records whether the parameters of the product are qualified at the checkpoint, but cannot reflect the dynamic stability of the product during its flow between checkpoints. For example, even if the event logs of two products are exactly the same, one may have experienced a smooth production process, while the other may have encountered minor mechanical disturbances or brief material waiting without triggering alarms. The existing system will identify these two different production histories as equivalent, resulting in the technical problem of false process equivalence.

[0004] To obtain more detailed process information, one approach is to increase the deployment density of physical sensors, but this increases the system's hardware overhead and data processing load. Another approach is to apply complex statistical algorithm models for defect prediction, but the conclusions often lack intuitive correspondence with the process and are difficult to use as a direct basis for process supervision. Analysis shows that existing data processing methods have the following shortcomings: traceability information loses dynamic rhythm information characterizing process stability during the conversion from continuous production processes to discrete event records; management lacks an effective method to intuitively quantify the process quality differences between different qualified products using existing data. Therefore, how to establish a data processing method based on process timestamps, by establishing a standardized process rhythm benchmark and calculating the difference between the actual process rhythm of any production object and the benchmark, to achieve a quantitative characterization of the dynamic stability of the production process, and to provide managers with a decision-making basis for distinguishing deep-seated process quality differences between different qualified products, becomes the technical problem to be solved by this invention. Summary of the Invention

[0005] This invention provides a full-process traceability system for an RF coaxial connector production line. Its main purpose is to solve the problem that existing traceability systems, due to deficiencies in their information models, cannot quantify and trace the dynamic rhythmic characteristics of the production process, thus failing to effectively identify deep-seated process quality differences among qualified products.

[0006] To achieve the above objectives, the present invention provides a full-process traceability system for an RF coaxial connector production line, the system comprising:

[0007] A template building module is configured to generate a golden rhythm template containing multiple process time increments based on timestamp data of a reference batch, and to pre-define a rationality boundary consisting of the minimum physical feasible time (MPT) and the maximum physical feasible time (XPT) for each process time increment in the golden rhythm template.

[0008] A verification gateway module is configured to collect the timestamp sequence of the traceable object, calculate the measured value of each process time increment based on the timestamp sequence, and then compare the measured value of each process time increment with its corresponding reasonableness boundary. The timestamp sequence is determined to be a verified timestamp sequence only when the measured values ​​of all process time increments in the timestamp sequence are within their respective reasonableness boundaries.

[0009] A fingerprint and difference calculation module is configured to generate an individual rhythm fingerprint from the verified timestamp sequence and calculate the difference RDS between the individual rhythm fingerprint and the golden rhythm template only when the verification gateway module determines that the timestamp sequence is a verified timestamp sequence.

[0010] A data association module is configured to associate and store the difference RDS as traceability metadata with the unique identifier of the traceable object;

[0011] An anomaly detection module is configured to perform similarity matching between an individual's rhythm fingerprint and any prototype in the prototype library to obtain the highest similarity score. When the highest similarity score is lower than a preset confidence threshold, an exclusive label representing the unclassified anomaly pattern is generated and associated with a unique identifier, and the verified timestamp sequence used to generate the individual's rhythm fingerprint is stored.

[0012] Preferably, the template creation module is configured to generate both the golden rhythm template and the individual rhythm fingerprint as multi-dimensional vectors, where each dimension of the multi-dimensional vector corresponds to a process time increment determined by the difference in timestamps between two adjacent predefined process nodes in the production process.

[0013] Preferably, the fingerprint and difference calculation module is configured to use a weighted Euclidean distance calculation method or a dynamic time warping (DTW) algorithm to calculate the difference RDS between an individual rhythm fingerprint and a golden rhythm template.

[0014] Preferably, the verification gateway module is further configured to: when the measured value of any process time increment in the timestamp sequence is not within its corresponding reasonable boundary, stop generating individual rhythm fingerprints for the timestamp sequence, and generate a management label representing the abnormality of the process time series data and store it together with a unique identity identifier.

[0015] Preferably, the anomaly detection module is further configured to: before performing similarity matching, first determine a drift vector based on the difference between the individual rhythm fingerprint and the golden rhythm template; the prototype library contains multiple drift prototypes, each of which corresponds to a predefined management label.

[0016] Preferably, the prototype library contains management tags, including global sluggish tags that characterize global time deviations in the production process, and single-point blockage tags that characterize time deviations in the time increments of individual processes in the production process.

[0017] Preferably, the anomaly detection module is configured to determine the highest similarity score as the classification confidence level for this match. Furthermore, the rules for determining the generation of unique tags are limited to meeting certain conditions. ,in, The confidence threshold is a preset value used to determine whether an individual's rhythm fingerprint belongs to an unknown abnormal pattern.

[0018] Preferably, the system further includes: a baseline evolution analysis module, configured to periodically perform cluster analysis on multiple individual rhythm fingerprints generated within a specified time range, and based on the results of the cluster analysis, identify whether there is a superior new rhythm paradigm that differs from the current golden rhythm template and meets preset management conditions; and a decision proposal module, configured to generate a management decision proposal to update the golden rhythm template when the baseline evolution analysis module identifies a superior new rhythm paradigm.

[0019] Preferably, the baseline evolution analysis module is configured to define the preset management conditions as follows: the vector distance between the center of a new cluster and the current golden rhythm template exceeds a preset drift threshold, and the proportion of the number of individual rhythm fingerprints contained in the new cluster to the total number within a specified time range exceeds a preset coverage threshold.

[0020] Preferably, the system also includes a downstream process triggering module, which is configured to trigger differentiated quality inspection processes or automated work order assignment processes based on the value of the difference degree RDS, or based on the management label obtained after matching the drift prototype, or based on the exclusive label.

[0021] Compared with the prior art, the beneficial effects of the present invention are:

[0022] 1. By establishing a standardized golden rhythm template and acquiring the individual rhythm fingerprint of any traceable object in real time, and then determining the rhythm drift by calculating the difference between the two, the drift is finally associated with the object as a process quality metadata. This information processing method transforms the data records of the traceability system from discrete, static workstation event logs into a quantitative representation of the continuous dynamic characteristics of the production process in the time dimension. This provides managers with a management dimension to examine and compare the process stability differences between different qualified products, in addition to the binary judgment of qualified and unqualified.

[0023] 2. By determining the drift vector and matching it with a pre-defined drift prototype library, a management label representing the drift pattern of the traced object is assigned. This mechanism structurally couples the aforementioned quantified drift value with the qualitative label representing the nature of the abnormal pattern, so that the traceability information not only reflects the degree of process deviation, but also reveals the specific form of deviation. This provides a decision-making basis for subsequent process diagnosis or automated assignment of management processes, avoiding the inherent pattern ambiguity when troubleshooting based solely on a single drift value.

[0024] 3. During similarity matching, the highest similarity score is determined as the classification confidence score and compared with a preset confidence threshold. When the classification confidence score is lower than the threshold, the system replaces the management label with a unique label representing the unclassified abnormal pattern and stores the drift vector that caused the match. This approach utilizes the accompanying information in the similarity calculation process to establish a mechanism for identifying and representing unknown abnormal patterns. This enables the system to proactively reject incorrect classifications when faced with novel drift patterns outside its knowledge base and to isolate and preserve the original data of the unknown pattern, providing a reliable data input source for the continuous evolution and organizational learning of the process knowledge base. Attached Figure Description

[0025] Figure 1 This is a diagram illustrating the architecture and core data flow of the rhythm tracing system of the present invention.

[0026] Figure 2 This is a quantitative comparison chart of RDS for the stability of different production processes in this invention;

[0027] Figure 3 This is a schematic diagram of the hierarchical information processing and interaction architecture of the system of the present invention. Detailed Implementation

[0028] To make the objectives, technical solutions, and advantages of the present invention clearer, the technical solutions of the present invention will be described in detail below. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0029] This application provides a full-process traceability system for an RF coaxial connector production line. At the information processing level, its overall architecture mainly consists of a template creation module, a verification gateway module, a fingerprint and difference calculation module, a data association module, and an anomaly determination module. At the management process optimization level, it may also include a benchmark evolution analysis module and a decision proposal module. These modules work together to transform discrete, event-based timestamp data from the Manufacturing Execution System (MES) into management information that can quantitatively characterize the dynamic stability of the production process. The top-level data flow is as follows: the verification gateway module performs a preliminary data quality access review on the collected raw timestamp sequence; the reviewed data sequence is processed by the fingerprint and difference calculation module to generate process quality metadata; subsequently, the data association module binds it to the unique identifier of the traceable object; finally, the anomaly determination module and the benchmark evolution analysis module utilize this continuously generated metadata to perform process pattern diagnosis and long-term benchmark validity maintenance, respectively, thereby providing new decision-making basis for production management and supervision.

[0030] In engineering practice, setting an objective evaluation benchmark for the rhythm characterizing process stability is a fundamental requirement. To address this requirement, the template creation module in this invention is configured to execute an offline creation procedure for a standardized golden rhythm template. The input to this procedure is production data from one or more golden batches, reviewed and confirmed by process engineers. A golden batch refers to a benchmark production batch produced under optimal conditions in terms of equipment status, material supply, and personnel operation. The system retrieves the process timestamp data recorded at preset process nodes for all products within these batches from the MES database. Based on this, a multi-dimensional vector of process time increment is calculated for each product. Each dimension of this multi-dimensional vector corresponds to the difference in timestamps between two adjacent predefined process nodes in the production process. For example, a partial vector of a traceable object can be represented as [station 1 dwell time, station 1 to 2 transfer time, station 2 processing time, ...]. Furthermore, the template creation module performs a dimension-by-dimensional statistical averaging of the process time increment vectors of all selected golden batch samples, ultimately generating a standardized golden rhythm template representing the ideal working rhythm of the production line. This is the golden rhythm template. Its data structure is a multi-dimensional floating-point vector, for example Once established, the template is stored in the system's core database as a unified benchmark for all subsequent online real-time comparisons. This approach provides the entire traceability system with a data-driven, interpretable, and management-consensus-based ideal benchmark, giving a logical starting point for the quantitative assessment of process stability. The template creation module includes a specific offline creation procedure, referencing the objective screening and template generation of batches or "golden batches," comprising the following procedural steps: First, based on historical production data, several candidate batches with a process capability index (Cpk) higher than 1.67 and a first-pass yield rate higher than 99.5% are initially selected. Second, for each candidate batch, the timestamp sequence of all products at each process node is extracted, and their respective process time increment multidimensional vectors are calculated. Then, for all vectors within each candidate batch, the coefficient of variation (CV), i.e., the ratio of the standard deviation to the mean of that dimension, is calculated for each dimension. Finally, the candidate batch with the smallest average CV across all dimensions is selected as the "golden batch," and the initial Golden Rhythm template (GRT) is generated by taking the arithmetic mean of all process time increment vectors within that batch dimension.

[0031] In production information systems, the reliability of any timestamp-based analysis depends entirely on the quality of the data source. Network latency or instantaneous load fluctuations in data acquisition software can generate distorted timestamp records that are irrelevant to the actual process, leading to incorrect judgments by the management system. To prevent such data contamination from interfering with business logic at the source, this invention defines a golden rhythm template in the template creation module. At the same time, it was also configured to require process engineers to For each process time increment dimension, an additional pre-defined minimum physical feasible time is included. With maximum physical feasible time The reasonable boundaries of the structure; the setting of these two parameters is based on the physical and engineering constraints of the production line, rather than statistical data. For example, for the time increment of the robotic arm transfer process from workstation 1 to 2, its This can be set to the time required for the robotic arm to operate at its maximum speed, such as 1.5 seconds. Any data below this value can be considered erroneous due to clock asynchrony or other reasons. This can be set to a management tolerance limit much larger than the normal processing time, such as 30 seconds. Any data exceeding this value can be judged as equipment lag or abnormal process waiting. Parameter pairs, and Together, these elements constitute the complete definition of the template. They work together to enable the subsequent verification gateway module to perform a preliminary data access review. Specifically, when the verification gateway module collects the original timestamp sequence of a traceable object from the MES system in real time, it does not immediately send it to the analysis engine. Instead, it first calculates the measured value of each process time increment based on the timestamp sequence, and then compares each measured value with its corresponding pre-stored reasonableness boundary in the template library. The system compares the measured values ​​of all process time increments in the sequence and determines that the timestamp sequence is valid only if all measured values ​​fall within their respective reasonable boundaries. The system then allows the sequence to proceed to downstream modules for processing. Conversely, if any measured value of a process time increment exceeds its reasonable boundary, the verification gateway module will halt all subsequent calculations for generating an individual rhythm fingerprint for that timestamp sequence. It will also associate a unique identifier for the traceable object with a management tag representing an anomaly in the traceability database, and trigger an alarm message to the system administrator, containing a detailed record of which specific stage's timestamp data violated physical constraints. By establishing this verification gateway, the system protected by this invention effectively separates data noise stemming from the inherent uncertainty of the information system from process signals reflecting fluctuations in the production process. This improves the effectiveness of the data upon which all upper-level management analysis and decision-making are based, and transforms a potential system failure point into a management dimension that can be used to monitor the health of the production information system itself.

[0032] For timestamp sequences that have passed the verification gateway module's review, the fingerprint and difference calculation module will generate an individual rhythm fingerprint for it online and in real time; this is the fingerprint. Its generation logic is similar to the golden rhythm template. The calculation process is consistent, also converting its timestamp sequence into a multidimensional vector containing multiple process time increments, and its data structure is the same as... Maintaining dimensional consistency; subsequently, the module's task is to compute the newly captured individual rhythmic fingerprint. With the golden rhythm template pre-existing in the database The degree of difference between them is defined as a single numerical value, namely, rhythm drift. In a preferred embodiment, this difference degree The calculation can be performed using the weighted Euclidean distance method, and its formula can be defined as follows: ,in, It is the dimension of the vector. and They are and In the The components on the dimension, and It is a preset term used to characterize the first The weighting coefficients for the importance of each process time increment in the overall process flow are determined by process engineers based on domain knowledge. For example, key processes that have a decisive impact on the final product performance can be assigned higher weight values. To illustrate this numerically, consider a simplified three-dimensional rhythm vector with its golden rhythm template. Its weight vector There is now a verified individual rhythm fingerprint. Then its rhythmic drift degree The calculation process is as follows The calculation result is In the absence of conflict, to address scenarios where the duration of each process in certain production flows may exhibit non-linear scaling, the degree of difference... The calculation can also be performed using the Dynamic Time Warping (DTW) algorithm or its simplified form; the calculated result The numerical value, as traceability metadata, is associated with the unique identifier of the traceable object through the data association module and stored in the traceability database. Thus, the present invention completes an expansion of traditional traceability information, so that when querying any product, the manager can not only see its static event log, but also obtain a dynamic quality score that quantifies the smoothness of its process.

[0033] A single While scalar values ​​can reflect the degree of process deviation, they cannot reveal the pattern of this deviation, which is valuable for accurate fault diagnosis. Therefore, this invention further includes an anomaly determination module, the mechanism of which involves calculating... Simultaneously, the deviation pattern is qualitatively attributed; this module is first based on individual rhythm fingerprints. With the Golden Rhythm Template The difference determines a drift vector, which is... This vector depicts the specific time deviation at each stage of the production process. Simultaneously, the system pre-configures a drift prototype library defined by process engineers based on typical failure modes. This library contains multiple drift prototypes, each a standardized vector corresponding to a label with clear managerial meaning. For example, the library might contain a global hysteresis label, whose corresponding prototype vector is... This is used to characterize minor delays that are common in the production process. This may indicate problems such as unstable air pressure or voltage fluctuations throughout the workshop. The library may also contain one or more single-point blockages. workstation The label, whose corresponding prototype vector is a vector that exists only in the first... A sparse vector with large positive values ​​in each dimension, such as This is used to characterize localized equipment stalls or material waits at specific workstations; the workflow of the anomaly detection module involves using the drift vector calculated in real time. The system performs similarity matching with any prototype in the prototype library, for example, using a cosine similarity algorithm, to obtain a highest similarity score. The management tag corresponding to this highest score is then used as the second process quality metadata and stored in association with the unique identifier of the traceable object. This approach couples the quantified drift value with the qualitative tag that characterizes the nature of the abnormal pattern, providing managers with higher-quality decision-making information. This enables subsequent management processes, such as automated work order assignment, to be distributed to the correct responsible unit based on the specific nature of the problem.

[0034] In real industrial environments, besides known, predefined anomaly patterns, the potential risks lie in unknown, undefined anomaly patterns. Incorrect classification of these patterns can mislead managers' diagnostic approach. To establish a mechanism for identifying and representing unknown anomalies, the anomaly determination module of this invention incorporates a set of decision bifurcation rules based on confidence assessment within the similarity matching logic. Specifically, the system determines the highest similarity score obtained during the matching process as the classification confidence level for this match. At the same time, a management parameter defined by the administrator is added to the system, namely the confidence threshold. It is a value between 0 and 1, used to define the boundary between known and unknown information; the judgment rules of the anomaly detection module are limited to: when the condition is met... When this happens, the system will determine that the drift is an unknown anomalous pattern. Under this condition, the module will refuse to associate any known management labels with the object, and instead generate a unique label representing the unclassified anomalous pattern, which will be associated with and stored as the object's unique identifier. Simultaneously, the system will also store the original drift vector that caused this low-confidence match. Alternatively, it can be used to generate a fully isolated and documented data sequence through a verified timestamp, for example, by storing it in a new pattern candidate pool database specifically for in-depth post-hoc analysis by process experts; the introduction of this mechanism enables the entire traceability system to distinguish between known problems and newly discovered problems, avoiding the risk of misleading diagnoses.

[0035] Furthermore, the system claimed in this invention also includes a downstream process triggering module, which is configured to convert traceability metadata generated by upstream modules into automated management instructions. In one specific embodiment, this module has a pre-set set of rule-based decision logic matrices. For example, rule one is set so that when the rhythm drift (RDS) value of any traceable object exceeds a preset first-level threshold T1 (e.g., 0.8) but is lower than a second-level threshold T2 (e.g., 1.5), an enhanced inspection label is automatically associated with the object. This label is read by the quality management system through a data interface, and the object is automatically guided to a supplementary high-precision dimensional re-inspection station. Rule two is set so that when the management label output by the anomaly judgment module is a single-point blockage... At workstation 3, the system not only executes the enhanced inspection process of Rule 1, but also automatically generates an equipment maintenance work order. This work order contains a timestamp, product batch number, and diagnostic information indicating that workstation 3 may have experienced a localized equipment jam or material waiting. Based on the preset responsibility matrix, the work order is pushed to the work order pool of the equipment maintenance department. Rule 3 is set so that when the anomaly judgment module outputs an unclassified anomaly mode exclusive label, the system triggers the highest priority management alarm. In addition to executing all the aforementioned processes, the system also sends the original drift vector DV and related data used to generate this judgment to the process engineer's designated email address to indicate that a novel event requiring in-depth manual analysis has occurred.

[0036] Any static management benchmark faces the law of diminishing applicability in a dynamically evolving production environment, rendering the initially defined golden rhythm template obsolete. The benchmark used in the traceability system may become unsuitable due to continuous process improvements or the benign evolution of equipment status. To ensure the long-term effectiveness of the benchmark used in the traceability system, the present invention may further include a benchmark evolution analysis module and a decision proposal module. The benchmark evolution analysis module is configured to automatically perform a retrospective clustering analysis task in the background at a preset management cycle. It collects all individual rhythm fingerprints generated during this cycle and stored in the database. The data was analyzed, and an unsupervised clustering algorithm was used to analyze these data. The purpose of vector partitioning is to automatically discover whether a new high-density rhythmic cluster exists in the data; after clustering is completed, the module analyzes the clustering results to identify whether there is a cluster that differs from the current one. Furthermore, it satisfies the advantages of a new rhythmic paradigm that meets preset management conditions. These preset management conditions are a set of explicit and quantifiable business rules, such as the center of a new cluster being similar to the current cluster. The vector distance exceeds a preset drift threshold, and the new cluster contains The proportion of the number of occurrences within the period exceeds the preset coverage threshold; once such a dominant new rhythmic paradigm that has become the mainstream is identified, the decision proposal module will not be automatically updated. Instead of generating a structured rhythm baseline evolution suggestion report, this invention pushes the report as a management decision proposal to the workflow approval interface of the designated process engineer or manager. The manager then makes the final decision on whether to approve this new model as the golden rhythm template for the next stage. In this way, the present invention constructs a closed loop of self-examination and dynamic evolution for the main system, which helps the long-term applicability of its core management tools and transforms implicit process improvement results into solidifiable management assets.

[0037] Example 1: In a specific industrial application, two automated production lines, A and B, for producing coaxial connectors operate in parallel. Their equipment configurations, material supplies, and process flow settings are consistent. The existing Manufacturing Execution System (MES) provides management data, including output, first-pass yield, and single-point process capability index for each workstation. For several consecutive production cycles, no statistically significant differences were observed between the two production lines. However, during accelerated aging tests for long-term reliability of the finished products, the quality management department discovered that products from production line B had a higher probability of early failure within their batches compared to those from production line A, even though all products from production line B passed all electrical performance and dimensional tests upon leaving the factory. This phenomenon reveals a blind spot in traditional data traceability methods regarding management information: the inability to distinguish the stability differences experienced during the production process among qualified products that passed all tests. To address this situation, the full-process traceability system of this invention was deployed in this scenario. First, the template establishment module uses all product batches produced by production line A within a complete production cycle as reference batches, extracts the full-process timestamp data recorded in the MES, and calculates the process time increment vector for each product. By statistically averaging these vectors, a standardized golden rhythm template is established. Subsequently, the system uses the verification gateway module to conduct data access review on the timestamp sequences collected in real time from production lines A and B. After confirming that the data has not been contaminated due to factors such as network latency, the fingerprint and difference calculation modules generate individual rhythm fingerprints online for each traceable object passing through the two production lines. And calculate its relationship with Difference between Data shows that all products on production line A... The values ​​are all stably distributed within a low-level range close to zero, while the products of production line B... The values ​​show a consistently high level, and the data dispersion is relatively large.

[0038] In response to the persistent high [risk] on production line B The numerical anomaly detection module automatically initiates anomaly attribution analysis. This module first bases its analysis on each high-value anomaly. Individual and The difference determines a drift vector. Accordingly, the system will Similarity matching was performed with multiple prototypes pre-installed in the drift prototype library. The matching results showed that the vast majority of prototypes in production line B were high-quality. Individual All of these are related to a single point of blockage in the prototype library, which is defined by the management tag. The drift prototype at workstation 3 has a cosine similarity higher than a preset threshold; this single-point blockage... The prototype of workstation 3 has a vector feature that shows a positive deviation only in the dimension corresponding to the processing time of workstation 3, while the other dimensions are all positive. The results largely matched, pointing to a localized and repetitive anomaly at workstation 3 as the overall process instability of production line B. Based on this data, the equipment maintenance department inspected the material clamps at workstation 3 of production line B and found a slight, intermittent pressure deficiency in an internal pneumatic component. This caused the clamps to occasionally vibrate slightly for several hundred milliseconds during positioning. This vibration did not cause products to be scrapped due to out-of-tolerance issues and was therefore never detected by the traditional SPC system. However, it was this repetitive, slight disturbance that caused a continuous deviation in the process time increment of this step, which was then identified and located by the system of this invention. After the pneumatic component was replaced, subsequent batches of products produced on production line B... The values ​​quickly fell back to the same low level range as production line A. At the same time, the early failure probability of its products in the long-term reliability accelerated aging test also returned to the same level as production line A.

[0039] Example 2: To objectively verify the effectiveness of the full-process traceability system claimed in this invention in identifying different types of process anomalies, this example constructs an experimental environment based on discrete event simulation. This simulation platform can generate timestamp logs that are consistent with the data structure of the real production line MES system and is configured to accurately inject preset time disturbances. Its core functional specification is that the timestamp generation accuracy is better than 1ms. The purpose of this experiment is to compare the identification and classification capabilities of the traditional process control method based on parameter upper and lower limits with the system of this invention when processing data containing different disturbance modes.

[0040] The experiment included a control group and an experimental group using the method of this invention. The control group was treated by simply determining whether the time consumed in each process segment exceeded the traditional SPC control limits, which were set according to the standard Six Sigma principle based on baseline data. The experimental group, on the other hand, applied the complete method of this invention, including establishing a golden rhythm template. Calculating individual rhythm fingerprints With rhythmic drift And perform anomaly detection; in the experiment, the confidence threshold of the key parameter in the anomaly detection module. The value was set to 0.7. This value was determined through receiver operating characteristic curve analysis of historical data, balancing the sensitivity to novelty anomaly detection with the need to avoid overreacting to normal process fluctuations. The experiment included five operating conditions, each generating 100 independent product traceability data samples: Condition 1 (baseline state): The simulation platform operated according to a standard rhythm without any human disturbance; the generated data served as sample group A of this invention. Condition 2 (global delay): The time consumption of all process segments was increased by a fixed 5% delay based on the standard value, which remained within the single-point SPC control limits; the data served as... Sample group B of this invention; Case 3 (single-point blockage): An additional delay of 100ms to 300ms is randomly injected only in the third process segment, and its data is used as sample group C of this invention; Case 4 (composite disturbance): The time consumption of the first and second process segments is reduced by 5%, while the time consumption of the fourth process segment is increased by 15%, simulating an atypical abnormal mode, and its data is used as sample group D of this invention; After the experiment was carried out, the analysis results of the control group showed that for all 400 samples under all four cases, the time consumption of each process segment did not exceed the preset SPC control limit, and the detection rate was 0%; while the experimental group using the method of this invention obtained quantitative results with discrimination. Table 1 shows the key experimental data records.

[0041] Table 1: Recording table of key experimental data.

[0042] Sample group name Injected exception types Average RDS RDS standard deviation Anomaly detection module attribution results Sample A of the present invention Baseline state 0.012 0.005 None (RDS is below the trigger threshold) Sample B of the present invention Global lag 0.855 0.021 Global lag Sample C of the present invention Single point of blockage 1.231 0.453 Single point of blockage @ workstation 3 Sample D of the present invention Composite disturbance 1.057 0.158 Unclassified anomaly pattern

[0043] Referring to Table 1, the present invention sample group A A value close to zero indicates the stability of the system under baseline conditions; the value of sample group B in this invention... When the numerical value increases and the standard deviation is small, its drift vector Matched with the global hysteresis prototype in the prototype library, and accurately attributed; the present invention sample C The highest value and the largest fluctuation reflect the characteristics of random blocking events and are accurately located to specific workstations; for sample group D of this invention, its The numerical values ​​are also relatively high, but the anomaly detection module calculates the highest similarity score when performing similarity matching. The value is 0.61, which is lower than the preset confidence threshold. That is, 0.7. Therefore, the system refuses to classify it into any known prototype, but instead associates it with a specific label for an unclassified abnormal pattern and its drift vector. The data is stored in a new pattern candidate pool; experimental results show that the system claimed in this invention, by calculating the rhythm drift degree... It can quantitatively detect process anomalies that traditional SPC methods cannot identify; moreover, its anomaly determination module can qualitatively attribute known anomaly patterns and has the ability to identify unknown anomaly patterns, providing objective data basis for in-depth monitoring and management of production processes.

[0044] Example 3: This example combines Figures 1 to 3 This paper describes a full-process traceability system for an RF coaxial connector production line, such as... Figure 1 As shown, the first step is to enter the data verification module, using the preset templates in the Golden Rhythm Template Library. The boundaries are reviewed, and sequences that fail are marked as abnormal time-series data, while timestamp sequences that pass the verification are sent to the fingerprint generation and difference calculation module, which uses the golden rhythm template. Generate individual rhythmic fingerprints for sequences And calculate the degree of difference Subsequently, the abnormal mode determination module receives... and The value is calculated by comparing the drift vector (DV) with drift prototypes in the drift prototype library. If a match is found, a management label is output; if the match confidence is too low, an unclassified anomaly label is output and the data is stored in the new pattern candidate pool. These analysis results are finally integrated by the associated traceability metadata module to form associated traceability metadata, which is then stored in the traceability database. This database not only supports traceability metadata queries but also allows for the storage of historical data. The data is provided to the analysis benchmark evolution module. The analysis results of this module are submitted to process engineers / managers in the form of management decision proposals. This is used to update the golden rhythm template library and drift prototype library by establishing standard template processes, forming a closed-loop optimization process. At the same time, the system can also trigger downstream processes based on traceability data and issue automated instructions to downstream management systems such as quality and equipment.

[0045] like Figure 2 As shown, the graph uses the production date as the horizontal axis and the... The vertical axis clearly shows the quantitative difference in process stability between production line A and production line B. The solid line representing production line A consistently operates stably at a low level close to 0, indicating that its production process is highly consistent with the golden rhythm. In contrast, the dashed line representing production line B fluctuates continuously within a higher value range, deviating from the ideal state. This intuitively reveals that its production process has deep-seated dynamic instabilities that are difficult to detect using traditional methods. In addition, the graph also includes a control limit to indicate the management warning threshold for process deviations.

[0046] like Figure 3 As shown in the diagram, the system is logically divided into four layers. The bottom layer is the production physical layer, which provides the system with massive amounts of real-time timestamp data. The middle layer is the system's real-time computing core and data storage core. The real-time computing core includes dynamic processing units such as a verification gateway module, fingerprint computing module, anomaly judgment module, data association module, and process triggering module. The data storage core contains static data assets such as a traceability database, a golden rhythm template library, a drift prototype library, and a new pattern candidate pool. It also has a self-learning and benchmark evolution closed-loop mechanism. The outermost layer is the interaction layer. The left side is the human-computer interaction layer that interacts with managers and is responsible for model calibration, strategy configuration, and outputting decision analysis and reports. The right side is the external collaboration layer that interacts with external systems and is responsible for issuing precise automated commands.

[0047] Example 4: In a specific system deployment scenario, an RF coaxial connector production line is about to go into operation. Because this production line uses new processing and conveying equipment, there is a lack of readily available historical production data and process management experience. In the initial stages of this project, to ensure that the full-process traceability system claimed in this invention can be objectively and effectively initialized and configured, a standardized offline calibration procedure needs to be executed to determine two core system parameters, namely, those used to calculate rhythm drift. weight vector And initial entries for the drift prototype library used for anomaly attribution; for calibrating the weight vector First, run a gold batch to establish the initial gold rhythm template. Subsequently, a process sensitivity analysis method was employed. This method quantifies the importance of a specific stage in the overall process flow by injecting a small, standardized time perturbation into a particular stage of the production process and measuring the impact of this perturbation on the key performance indicators of the final product. Specifically, the voltage standing wave ratio (VSWR) of the final product is used as the metric. As a key performance indicator, the gold batch of products was measured first. Benchmark values; then, while ensuring that other process stages strictly adhere to them. Under the premise that a fixed delay of 100ms is artificially added only to the first process time increment (station 1 dwell time), and a batch of products is produced, and the batch of products is measured. Average degradation ; in turn All of them Repeat this operation for each process time increment to obtain a set of... Deterioration Weight vector Each weight component The value is determined to be proportional to the corresponding performance degradation, obtained through normalization calculation, i.e. ,in, Indicates the relationship with the first The time increment of each process corresponds to Deterioration amount, This indicates that the value used to iterate through the degradation values ​​in this group is the 1st. This procedure transforms the setting of weighting coefficients from relying on subjective experience to a data-driven calibration process that can be reproduced through controlled experiments.

[0048] In the weight vector Once determined, the initial construction of the drift prototype library is carried out. This construction process utilizes a fault injection experiment method, which maps physically meaningful abnormal operating conditions to a standardized drift vector. To establish a global hysteresis prototype, engineers reduced the overall drive air pressure of the production line by 5%, ran a batch under these conditions, and collected individual rhythmic fingerprints for all products. Calculate the relative value of each product to drift vector Subsequently, through the analysis of all samples within this batch... Vector averaging is performed to obtain a stable average drift vector that characterizes the global hysteresis condition, eliminating random noise. This vector is then stored in the prototype library and associated with the management tag for global hysteresis. Similarly, by setting a tiny physical limit on the material handling robot arm at station 3 to simulate a single-point blockage condition, the same process can be used to generate and calibrate the single-point blockage. The prototype of workstation 3; through this series of calibration experiments based on fault injection, the initial construction of a drift prototype library containing several core abnormal modes with clear physical root causes was completed; after completing the above two calibration procedures, the full-process traceability system claimed by this invention has all the initial configuration parameters for effective deployment on a new production line.

[0049] Example 5: In a specific system commissioning scenario, to ensure that the physical feasible time boundary parameters of the verification gateway module are objectively set, a set of targeted on-site calibration procedures needs to be executed; to determine the minimum physical feasible time corresponding to any process time increment. The production equipment involved in the increment is then independently operated at its rated speed under no material load conditions, and this operation is repeated 10 times. The time taken for each operation is recorded, and the shortest recorded time is selected as the time increment for this process. To determine the maximum physically feasible time Then it is set as a template related to management strategy and based on the golden rhythm. Multiples of, for the first The time increment of the process, The calculation formula is: ,in, For the first The maximum physical feasible time for each process time increment This is the standard value corresponding to the time increment of this process in the golden rhythm template. It is a global management tolerance coefficient, which is set to 5 in this embodiment. This provides a unified benchmark for all process segments to identify extreme abnormal waiting or equipment stall events.

[0050] Accordingly, in order to calibrate the key trigger thresholds in the baseline evolution analysis module, after the production line has been running stably for one full month, the system first collects all individual rhythm fingerprints generated during this period. Data was collected, and the rhythmic drift of all corresponding products was calculated. Numerical value; based on this initial value The dataset is used to calculate the overall standard deviation. Therefore, the drift threshold used to determine whether new cluster centers have drifted is set to 6 times the initial standard deviation, i.e. This setting links the threshold to the inherent process fluctuation level of the production line at this stage; at the same time, based on the management's requirements for the process change verification cycle, the coverage threshold for judging whether a new cluster has become mainstream is set at 30%, that is, a potential advantageous new rhythmic paradigm must account for at least 30% of the total output in the entire analysis cycle before it can be identified by the system and generate corresponding management decision proposals.

[0051] Example 6: In a specific production management scenario, the production line needs to temporarily switch to a batch of certified but slightly different process characteristics of backup supplier raw materials. To avoid this known planned process adjustment being incorrectly identified as a process anomaly by the system, a pre-set management procedure needs to be implemented. Before this batch of materials is put into use, the system administrator pre-registers a temporary process change event in the traceability system and sets the product batch number range associated with the event. When the first product belonging to this batch number range on the production line passes through all processes, the system automatically captures its individual rhythm fingerprint. This is then set as a temporary golden rhythm template for that specific batch; during the validity period of this temporary template, the rhythm drift of all subsequent products belonging to that batch number range will be... The calculations all involve taking their respective... Compare with this temporary golden rhythm template, rather than with the production line standard. In comparison, under these temporary process conditions, the system continues to effectively monitor the dynamic stability of the production process until all products in that batch are produced, at which point the system automatically reverts to the standard operating conditions. The default state for comparison.

[0052] To ensure that the unsupervised clustering algorithm used in the baseline evolution analysis module can be deterministically configured, this embodiment discloses a method for addressing the key hyperparameter of the K-Means algorithm: the number of clusters. The system employs a systematic optimization procedure; it first extracts all individual rhythm fingerprints generated within the past quarter from the traceability database. The data forms a historical dataset for calibration; subsequently, the system iteratively executes commands on this dataset. K-Means clustering operations with values ​​from 1 to 10 are performed, and the sum of squared intra-cluster errors is calculated for each operation. ; by drawing Follow The system analyzes the curve of value changes and automatically finds the inflection point of the curve's slope using the elbow rule. This allows the system to identify the optimal number of clusters within the historical dataset. Finally, the optimization procedure outputs... The value is set as the clustering quantity parameter used by the baseline evolution analysis module when performing periodic retrospective analysis. This transforms the selection of key algorithm hyperparameters from relying on trial and error by the operator into an automatically executed optimization process based on the distribution characteristics of the data itself.

[0053] It will be apparent to those skilled in the art that the present invention is not limited to the details of the exemplary embodiments described above, and that the present invention can be implemented in other specific forms without departing from the spirit or essential characteristics of the present invention.

[0054] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention.

Claims

1. A full-process traceability system for an RF coaxial connector production line, characterized in that, The system includes: A template building module is configured to generate a golden rhythm template containing multiple process time increments based on timestamp data of a reference batch, and to pre-define a rationality boundary consisting of the minimum physical feasible time (MPT) and the maximum physical feasible time (XPT) for each process time increment in the golden rhythm template. A verification gateway module is configured to collect the timestamp sequence of the traceable object, calculate the measured value of each process time increment based on the timestamp sequence, and then compare the measured value of each process time increment with its corresponding reasonableness boundary. The timestamp sequence is determined to be a verified timestamp sequence only when the measured values ​​of all process time increments in the timestamp sequence are within their respective reasonableness boundaries. A fingerprint and difference calculation module is configured to generate an individual rhythm fingerprint from the verified timestamp sequence and calculate the difference RDS between the individual rhythm fingerprint and the golden rhythm template only when the verification gateway module determines that the timestamp sequence is a verified timestamp sequence. A data association module is configured to associate and store the difference RDS as traceability metadata with the unique identifier of the traceable object; An anomaly detection module is configured to perform similarity matching between an individual's rhythm fingerprint and any prototype in the prototype library to obtain the highest similarity score. When the highest similarity score is lower than a preset confidence threshold, an exclusive label representing the unclassified anomaly pattern is generated and associated with a unique identifier, and the verified timestamp sequence used to generate the individual's rhythm fingerprint is stored.

2. The full-process traceability system for an RF coaxial connector production line according to claim 1, characterized in that, The template creation module is configured to generate both the golden rhythm template and the individual rhythm fingerprint as multi-dimensional vectors. Each dimension of the multi-dimensional vector corresponds to a process time increment determined by the difference in timestamps between two adjacent predefined process nodes in the production process.

3. The full-process traceability system for an RF coaxial connector production line according to claim 1, characterized in that, The fingerprint and difference calculation module is configured to use either a weighted Euclidean distance calculation method or a dynamic time warping (DTW) algorithm to calculate the difference RDS between an individual rhythm fingerprint and a golden rhythm template.

4. The full-process traceability system for an RF coaxial connector production line according to claim 1, characterized in that, The verification gateway module is also configured to: when the measured value of any process time increment in the timestamp sequence is not within its corresponding reasonable boundary, stop generating individual rhythm fingerprints for the timestamp sequence, and generate a management label representing the abnormality of the process time series data and associate it with a unique identity identifier for storage.

5. The full-process traceability system for an RF coaxial connector production line according to claim 1, characterized in that, The anomaly detection module is also configured to: determine a drift vector based on the difference between the individual rhythm fingerprint and the golden rhythm template before performing similarity matching; the prototype library contains multiple drift prototypes, each of which corresponds to a predefined management label.

6. The full-process traceability system for an RF coaxial connector production line according to claim 5, characterized in that, The prototype library contains management tags, including global sluggish tags that characterize global time deviations in the production process, and single-point blockage tags that characterize time deviations in the time increments of individual processes in the production process.

7. The full-process traceability system for an RF coaxial connector production line according to claim 1, characterized in that, The anomaly detection module is configured to determine the highest similarity score as the classification confidence level for this match. Furthermore, the rules for determining the generation of unique tags are limited to meeting certain conditions. ,in, This is the confidence threshold.

8. The full-process traceability system for an RF coaxial connector production line according to claim 1, characterized in that, The system also includes: a baseline evolution analysis module, configured to periodically perform cluster analysis on multiple individual rhythm fingerprints generated within a specified time range, and based on the results of the cluster analysis, identify whether there is a superior new rhythm paradigm that is different from the current golden rhythm template and meets preset management conditions; and a decision proposal module, configured to generate a management decision proposal to update the golden rhythm template when the baseline evolution analysis module identifies a superior new rhythm paradigm.

9. The full-process traceability system for an RF coaxial connector production line according to claim 8, characterized in that, The baseline evolution analysis module is configured to define the preset management conditions as follows: the vector distance between the center of a new cluster and the current golden rhythm template exceeds a preset drift threshold, and the proportion of the number of individual rhythm fingerprints contained in the new cluster to the total number within a specified time range exceeds a preset coverage threshold.

Images