Cable production traceability method and system based on big data

Abstract

The invention relates to the technical field of production traceability, and particularly discloses a cable production traceability method and system based on big data, and the method comprises the steps: collecting core traceability data of extrusion and vulcanization processes in cable production, dividing a causal disturbance interval through constructing a disturbance conduction chain, and carrying out the dynamic processing of the data into a variable-granularity traceability unit set; based on the traceability unit set, granularity matching analysis is carried out under the scenes of quality traceability, process optimization and responsibility definition, and an adaptive fault of granularity imbalance is positioned; extracting causal features of missing node pairs in the fault, and defining an applicable boundary of the traceability model; and establishing a coupling mapping relation by combining boundary features and granularity-fault features, and dynamically calibrating a granularity adjustment coefficient. The system comprises a unit construction module, a fault extraction module, a boundary defining module and a granularity calibration module. According to the invention, through variable granularity processing and dynamic calibration, the cable production traceability precision and scene adaptability are improved.

Description

A method and system for cable production traceability based on big data Technical Field

[0001] This invention relates to the field of production traceability technology, specifically to a method and system for cable production traceability based on big data. Background Technology

[0002] In the cable manufacturing process, the process parameters of core steps such as extrusion and vulcanization are closely related to product quality through disturbance transmission. Even slight fluctuations in these parameters can affect key quality indicators of the final product, such as wire diameter and insulation dielectric loss value, through the transmission chain. However, traditional cable production traceability methods often use fixed-granularity data unit division, that is, cutting and storing parameter data in the production process according to uniform time intervals or process segments.

[0003] The fixed-granularity processing method of existing technology is difficult to adapt to the dynamic characteristics of parameter disturbances in different processes: when the parameters are in a stable low-frequency disturbance range, the fine granularity will lead to data redundancy and increase the computational burden of traceability analysis; while when the parameters enter the high-disturbance range of violent fluctuations, the coarse granularity will miss key fluctuation details.

[0004] In quality traceability, fixed granularity makes it impossible to pinpoint the root cause of quality defects due to the lack of details regarding fluctuations in key parameters. In process optimization, the mismatch between data granularity and the requirements for parameter correlation analysis makes it difficult to uncover parameter coupling patterns between processes. Furthermore, in responsibility delineation, the mismatch between team information and the time granularity of parameter fluctuations leads to ambiguous responsibility assignments. These problems ultimately result in low accuracy and efficiency in cable production traceability, failing to meet the demands of modern cable production for full-process, high-precision quality traceability.

[0005] Therefore, this invention provides a method and system for tracing the production of cables based on big data. Summary of the Invention

[0006] The purpose of this invention is to provide a method and system for tracing the production of cables based on big data, so as to solve the aforementioned background problems.

[0007] The objective of this invention can be achieved through the following technical solution: a cable production traceability method based on big data, comprising the following steps: collecting cable production traceability data, establishing a traceability model, performing variable granularity processing on the cable traceability data, and constructing a variable granularity traceability unit set; based on the traceability unit set, performing granularity matching analysis on the traceability scenarios of the traceability unit set, constructing a granularity fit matrix and extracting fit equilibrium values ​​of different dimensions, determining whether the matching degree of different dimensions is balanced, and if not balanced, locating the fit fault of granularity imbalance; extracting missing node pairs covering the imbalanced units in the fit fault, establishing a fault causal vector by extracting the fault causal features of the missing node pairs, and performing failure candidate analysis on the fault causal vector to define the applicable boundary of the traceability model; extracting the core features of the applicable boundary of the traceability model, establishing a coupling mapping relationship by combining the core features of granularity and fault, and dynamically calibrating the granularity adjustment coefficient based on the coupling mapping relationship to achieve dynamic calibration of granularity.

[0008] As a further aspect of the present invention, the variable granularity processing is performed by: constructing a source tracing model, extracting the disturbance transmission chain of different processes in the source tracing model and the causal disturbance interval of the disturbance transmission chain; and dynamically dividing the granularity unit duration based on the causal disturbance interval.

[0009] As a further aspect of this invention, the method for determining whether the matching degree of different dimensions is balanced is as follows: the traceability scenario is divided into quality traceability scenario, process optimization scenario, and responsibility definition scenario; from the traceability unit set, the granularity matching degree of each granular unit in the three traceability scenarios is obtained, and a granularity fit degree matrix with the behavior scenario column as the dimension is constructed; based on the granularity fit degree matrix, a fit balance analysis is performed to obtain the fit balance value; based on the fit balance value, it is determined whether the matching degree of each dimension is balanced, and if it is not balanced, the deviation analysis dimension is extracted; the node coverage of the deviation analysis dimension is extracted, and the coverage imbalance unit is filtered based on the node coverage; based on the coverage imbalance unit, it is determined whether the coverage imbalance unit is at a sensitive point of causal chain break.

[0010] As a further aspect of the present invention, the method for extracting the deviation analysis dimension is as follows: if the matching degree of each dimension is not balanced, the deviation ratio between the granular matching degree of each dimension and the preset ideal matching degree is calculated to obtain the matching deviation rate; negative matching deviation rates are filtered out, and the negative matching deviation rates are sorted in ascending order, and the dimension with the smallest negative matching deviation rate is selected as the deviation analysis dimension.

[0011] As a further aspect of the present invention, the method for locating the adaptation fault of the granularity imbalance is as follows: if the coverage imbalance unit is located at the break sensitive point of the causal chain, the granularity unit duration of the coverage imbalance unit is extracted, and the granularity imbalance type of the coverage imbalance unit is determined; for each tracing scenario, a tracing blocking analysis is performed to obtain the fault blocking coefficient and associate it with the granularity imbalance type, and missing node pairs are extracted and the adaptation fault is determined.

[0012] As a further aspect of the present invention, the failure candidate analysis is performed as follows: based on the cable tracing data, the fault causal features of missing node pairs are extracted, and a fault causal vector is established; based on the fault causal vector, failure candidate analysis is performed to locate the failure candidate interval; the significance of the applicable boundary of the failure candidate interval is verified to determine the final applicable boundary.

[0013] As a further aspect of the present invention, the method for verifying the significance of the applicable boundary is as follows: obtain the model failure confidence level through the model failure confidence analysis equation; calculate the statistical significance level coefficient of the failure confidence level; if the statistical significance level coefficient satisfies statistical significance, then determine the failure candidate interval as the interval of significant model failure; extract the boundary parameter values ​​of the process parameter intervals adjacent to the interval of significant failure as the final applicable boundary.

[0014] As a further aspect of the present invention, the method for establishing the coupling mapping relationship is as follows: obtain the core features of granularity imbalance - adaptation fault - applicable boundary, and establish a core feature vector; perform implicit correlation analysis on process parameters to obtain the implicit coupling strength of process parameters, and establish a coupling mapping relationship based on the core feature vector and the implicit coupling strength; and dynamically calibrate the granularity adjustment coefficient based on the coupling mapping relationship.

[0015] As a further aspect of the present invention, the method for dynamically calibrating the granularity adjustment coefficient is as follows: the implicit coupling strength and the granularity adjustment coefficient are combined to obtain a set of parameters to be adjusted; the average total matching degree of the granularity fit matrix of each set of parameters to be adjusted is obtained from historical production data as the traceability accuracy; with the goal of maximizing the traceability accuracy, a continuous mapping function is constructed through a multiple linear regression algorithm; and the granularity adjustment coefficient is calibrated based on the continuous mapping function.

[0016] The cable production traceability system based on big data includes the following modules: Unit Construction Module: Collects cable production traceability data, establishes a traceability model, performs variable granularity processing on the cable traceability data, and constructs a variable granularity traceability unit set; Fault Extraction Module: Based on the traceability unit set, performs granularity matching analysis on the traceability scenarios of the traceability unit set, constructs a granularity fit matrix, extracts the fit equilibrium value of different dimensions, determines whether the matching degree of different dimensions is balanced, and if unbalanced, locates the granularity imbalance fit fault; Boundary Delineation Module: Extracts missing node pairs covering unbalanced units in the fit fault, establishes a fault causal vector by extracting the fault causal features of the missing node pairs, performs failure candidate analysis on the fault causal vector, and defines the applicable boundary of the traceability model; Granularity Calibration Module: Extracts the core features of the applicable boundary of the traceability model, establishes a coupling mapping relationship by combining the core features of granularity and fault, and dynamically calibrates the granularity adjustment coefficient based on the coupling mapping relationship to achieve dynamic granularity calibration.

[0017] The beneficial effects of this invention are as follows: By collecting core traceability data from cable production and processing it with variable granularity, and dynamically dividing the granularity units based on the process disturbance transmission chain and causal disturbance interval, the constructed variable granularity traceability unit set can flexibly adapt to the disturbance fluctuations at different production stages. This makes the unit division of traceability data more consistent with the parameter change patterns in the actual production process, providing targeted basic data support for various subsequent traceability scenarios. Based on the traceability unit set, granularity matching analysis is performed in various traceability scenarios. By constructing an adaptation degree matrix and calculating the adaptation balance value to locate the adaptation faults of granularity imbalance, the mismatch between data granularity and traceability requirements in different scenarios can be identified. Breakpoints and imbalances in the traceability process can be discovered in a timely manner, allowing for targeted optimization of traceability granularity and improving the effectiveness and relevance of the traceability process. By extracting the fault causal features of missing node pairs in the adapted fault, a causal vector is established and the applicable boundaries of the traceability model are defined. This clarifies the effective application range of the traceability model in different process parameter ranges, reduces invalid traceability in inapplicable ranges, and improves the reliability of the traceability model application. The traceability results are made more consistent with the actual production process characteristics and quality rules. A coupling mapping relationship is established between the applicable boundary features of the traceability model and the granularity-fault core features, and the granularity adjustment coefficient is dynamically calibrated, achieving dynamic optimization of granularity processing. This dynamic calibration mechanism enables granularity adjustment to continuously adapt to the implicit coupling rules of the production system, facilitating the synchronization of the granularity division of the traceability unit set with changes in production parameters and traceability needs. This further enhances the flexibility and accuracy of the entire traceability method and helps ensure the stability of long-term traceability results. Attached Figure Description

[0018] The invention will now be further described with reference to the accompanying drawings.

[0019] Figure 1 is a flowchart of the cable production traceability method based on big data according to the present invention; Figure 2 is a flowchart of the present invention for determining whether the matching of each dimension is balanced; Figure 3 is a block diagram of the cable production traceability system based on big data according to the present invention. Detailed Implementation

[0020] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0021] Example 1 (See Figure 1) This invention is a cable production traceability method based on big data, including the following steps: S1. Collect cable production traceability data, establish a traceability model, perform variable granularity processing on the cable traceability data, and construct a variable granularity traceability unit set; wherein, the method of collecting cable production traceability data is: preferably, extracting core cable traceability data from the two major processes of extrusion and vulcanization in cable production from a big data platform for cable traceability; wherein, the core cable traceability data includes: extrusion temperature, extrusion pressure change curve, vulcanization time, wire diameter deviation, dielectric loss value deviation of cable insulation layer, and production team information corresponding to the two major processes; wherein, for cable... The variable granularity processing method for traceability data is as follows: S101, construct a traceability model, extract the disturbance transmission chain of different processes in the traceability model, and the causal disturbance interval of the disturbance transmission chain; preferably, construct the disturbance transmission chain of the extrusion process (extrusion temperature fluctuation - extrusion pressure change - wire diameter deviation) and the disturbance transmission chain of the vulcanization process (vulcanization time - dielectric loss value deviation) respectively, and use temperature fluctuation, extrusion pressure change, and wire diameter deviation as causal nodes of the traceability model of the extrusion process; combine adjacent causal nodes to establish causal node pairs; calculate the disturbance transmission intensity of adjacent causal nodes within the monitoring period using a Bayesian network, extract the rate of change of disturbance transmission intensity within adjacent monitoring periods, and label it as the causal perturbation rate. Those skilled in the art will understand that when calculating the disturbance transmission intensity of adjacent causal nodes within a monitoring period using a Bayesian network, a Bayesian network structure is first constructed based on the production process logic, with adjacent causal nodes set as network nodes and connection relationships defined. Then, historical monitoring data is used to learn the conditional probability table of the nodes (e.g., the probability distribution of pressure when temperature takes different values). When a disturbance occurs at the causal node within the monitoring period, the probability distribution change of the result node is calculated through Bayesian inference. Finally, the quantified value of the probability change is used as the disturbance transmission intensity to measure the strength of the disturbance transmitted from the causal node to the result node. Different causal disturbance intervals are divided based on the causal perturbation rate. For example, the low-frequency disturbance interval is: ≤20%, disturbance propagation is stable; medium disturbance range: 20% < ≤50%, disturbance propagation exhibits moderate fluctuations; high disturbance range: >50%, disturbance propagation changes drastically; S102, dynamically divide the granularity unit duration based on the causal disturbance interval; preferably, through the formula: Time to obtain granularity unit Where n is the number of consecutive stable periods, which refers to the total number of periods that are in the same disturbance interval consecutively, including the current period and the previous periods. n≥1. If a period enters a new interval, n is reset to 1. The initial granularity unit duration; To integrate the causal perturbation rate and the granularity adjustment coefficient, dynamic values ​​are assigned according to intervals. For example, the granularity adjustment coefficients for the low-frequency perturbation interval, the medium-frequency perturbation interval, and the high-frequency perturbation interval are 1.0, 0.6, and 0.3, respectively. The mean perturbation conduction intensity and granularity confidence of each granularity unit are obtained as granularity unit attributes. For each granularity unit, process parameter fragments, work group association information, and quality association tags of the cable traceability data are supplemented within the granularity unit duration. A traceability unit set containing process parameter fragments, work group association information, and quality association tags of the cable traceability data is constructed.

[0022] S2. Based on the traceability unit set, perform granularity matching analysis on the traceability scenarios of the traceability unit set, construct a granularity fit matrix, and extract the fit equilibrium value of different dimensions to determine whether the matching degree of different dimensions is balanced. If it is unbalanced, locate the fit discontinuity of granularity imbalance. The method of granularity matching analysis based on the traceability unit set in the traceability scenario is as follows: the traceability scenario is divided into quality traceability scenario, process optimization scenario, and responsibility definition scenario; from the traceability unit set, obtain the granularity matching degree of each granular unit in the three traceability scenarios, and construct a granularity fit matrix with the behavior scenario column as the dimension. Where s represents the scenario, for example, S=1, 2, and 3 represent quality traceability, process optimization, and responsibility delineation scenarios, respectively; d represents the dimension, d=1, 2, and 3 represent causal integrity, time precision, and data efficiency, respectively; calculate the mean granularity matching degree across scenarios for each dimension, and the total matching degree for each dimension; perform a ratio calculation between the mean granularity matching degree and the total user matching degree to obtain the dimension matching ratio Pd; and apply the adaptation equilibrium equation: Obtain the adaptation balance value H of the granularity adaptation matrix. It can be understood that the adaptation balance value, based on the principle of information entropy, comprehensively quantifies the degree of balance in the matching degree distribution of each dimension (causal integrity, time precision, data efficiency) in the granularity adaptation matrix under three scenarios: quality traceability, process optimization, and responsibility definition. The value reflects the balance of the dimension matching degree: the higher the H, the greater the difference in matching degree among the dimensions, and the more unbalanced the distribution, meaning that the adaptability of some dimensions to the scenario deviates significantly from the overall level; the lower the H, the more balanced the matching degree distribution among the dimensions, and the more the overall granularity meets the comprehensive requirements of the traceability scenario. Based on the adaptation balance value, determine whether the matching degree of each dimension is balanced. As shown in Figure 2, if the adaptation balance value is higher than or equal to the preset adaptation balance threshold, it is determined that the matching degree of each dimension is unbalanced; if it is lower, it is determined that the matching degree of each dimension is balanced and meets expectations. The method for extracting the adaptation gaps of granularity imbalance is as follows: if the matching degree of each dimension is unbalanced, calculate the granularity matching degree of each dimension and... The deviation percentage of the preset ideal matching degree is used to obtain the matching deviation rate; negative matching deviation rates are filtered out and sorted in ascending order of numerical value, and the dimension with the smallest negative matching deviation rate is selected as the deviation analysis dimension; based on the deviation analysis dimension being the causal integrity dimension, the sensitive points of breakage in the causal chain are identified in conjunction with the disturbance transmission chain; the method for identifying sensitive points of breakage in the causal chain is as follows: S201, extract the node coverage of the deviation analysis dimension, and filter the coverage imbalance units based on the node coverage; preferably, obtain the number of causal node pairs actually covered by each granular unit in the traceability unit set and the total number of causal node pairs in the complete causal chain of the process; the ratio of the number of causal node pairs actually covered by each granular unit to the total number of causal node pairs in the complete causal chain of the process is processed to obtain the node coverage; for example, if the complete causal chain of the extrusion process contains 2 causal node pairs, and the granular unit actually covers one complete causal node pair, then the node coverage is 0.5; Determine the granular unit of coverage imbalance based on node coverage; For example, if the node coverage is less than 30%, the granular unit is marked as a coverage imbalance unit; S202, Based on the coverage imbalance unit, determine whether the coverage imbalance unit is at a sensitive point of causal chain break; Obtain the correlation probability between the coverage imbalance unit and the process quality defect traceability, and determine whether the coverage imbalance unit is at a sensitive point of causal chain break based on the correlation probability; Those skilled in the art will understand that the correlation probability is calculated based on Bayesian network, that is, the conditional probability that the missing causal node pair of the unit leads to the corresponding quality defect in historical data, such as the probability of the occurrence of wire diameter deviation defect when the pressure change-wire diameter deviation node pair is missing; For example, If the association probability of a coverage imbalance unit is higher than 60%, the corresponding coverage imbalance unit is placed at a break-sensitive point. If the coverage imbalance unit is at a break-sensitive point in the causal chain, then an adaptation fault analysis is performed on the coverage imbalance unit to obtain the adaptation fault of granularity imbalance. The method for performing adaptation fault analysis to obtain the adaptation fault of granularity imbalance is as follows: S211, extract the granularity unit duration of the coverage imbalance unit to determine the granularity imbalance type of the coverage imbalance unit. Preferably, the granularity imbalance type is divided into: incomplete coverage due to excessively coarse granularity, node fragmentation due to excessively fine granularity, and coverage loss due to insufficient continuity. It can be understood that if the granularity unit duration is higher than the average propagation delay of the causal chain, the granularity unit duration is 10. If the average propagation delay of pressure change-wire diameter deviation is 5 minutes, then due to excessively wide granular unit merging, some causal node pairs (such as the peak moment of wire diameter deviation) are not fully included, which is judged as an over-merging imbalance, i.e., incomplete coverage caused by excessively coarse granularity; if the granular unit duration is less than the duration of a single causal node pair (e.g., the node pair's pressure change due to temperature fluctuation needs to last 3 minutes, while the granular unit duration is 1 minute), then the node pair is split into multiple units, causing a single unit to be unable to cover the complete node pair, which is judged as an over-splitting imbalance, i.e., node fragmentation caused by excessively fine granularity; if the number of consecutive stable periods of the granular unit n=1 (1 period is within the current perturbation range), and the causal perturbation rate is >2 0% (disturbance fluctuation) indicates that the granularity division did not dynamically adjust with the disturbance, resulting in node pairs being missed during the fluctuation period. This is judged as an imbalance due to insufficient dynamic response, i.e., coverage gaps caused by insufficient continuity. S212: Perform traceability blocking analysis on each traceability scenario to obtain the fault blocking coefficient and associate it with the granularity imbalance type. Extract missing node pairs and determine the appropriate fault. Obtain the node coverage in each traceability scenario in the deviation analysis dimension, as well as the correlation probability L between the coverage imbalance unit and the quality defect. Obtain the coverage imbalance unit of the causal chain that is not actually covered by each granularity unit, and mark the causal node pairs in the coverage imbalance unit of the causal chain that is not actually covered as missing node pairs. Calculate the average disturbance propagation intensity of the missing node pairs. The ratio of the number of consecutive stable cycles to the total number of monitoring cycles contained in the granular unit for each traceability scenario is obtained and calculated to obtain the cycle stability ratio S; the correlation probability L and the average disturbance propagation intensity are also considered. , After standardization, the formula is used: Obtain the fault blocking coefficient for each traceability scenario. It should be explained that the fault blocking coefficient measures the ability of the adapted fault to block the production traceability process. Specifically, the correlation probability L anchors the historical correlation between the fault and quality defects. For example, if defects occur frequently when a node pair is missing, a high L indicates that the fault is more likely to hide the root cause of the defect. The average disturbance transmission intensity W reflects the causal hub status of the node pair in the production parameter chain. If the temperature node has a very strong transmission effect on pressure fluctuations, a high W indicates that the node pair is more core, making it more difficult to reconstruct the cause of parameter fluctuations after its absence. The production interval instability 1-S (S is the periodic stability ratio) reflects the interference characteristics of the environment in which the fault is located. For example, when parameter fluctuations are severe during the production cycle, 1-S... High, at this point the missing key node pairs will further amplify the difficulty of tracing due to the loss of fluctuation details. The combination of these three factors, whether the fault hits the defect logic, the irreplaceable function of the node pairs, and the tracing complexity of the production scenario, integrates to quantify the degree to which the fault hinders the location of the root cause, restoration of the process, and definition of responsibility in scenarios such as quality traceability and process optimization. The higher the value, the more the fault can jointly inhibit the effectiveness of the tracing process from the perspectives of correlation depth, transmission importance, and environmental interference. If the fault blocking coefficient of each tracing scenario is higher than the preset fault blocking coefficient threshold, the tracing scenario is marked as a fault scenario. The continuous occurrence of overlapping imbalance units with the same granularity imbalance type within the fault scenario is merged to obtain the adapted fault. It can be understood that the role of obtaining the adapted fault is: first, to provide a key sample set for the failure boundary analysis of the tracing model. The missing node pairs covering imbalanced units in the adaptation fault contain the fracture mode and parameter deviation characteristics of the causal chain transmission. This is the core basis for defining the applicable domain of the disturbance transmission chain in the process parameter space, and is helpful in locating the critical interval of the model's traceability effectiveness decay. Secondly, it provides targeted correction basis for dynamic granularity adjustment. The granularity imbalance type associated with the adaptation fault directly maps the degree of mismatch between the current granularity division and the process disturbance transmission characteristics. By extracting the fault blocking coefficient and causal node coverage defects, the correction magnitude of the granularity adjustment coefficient can be quantified, achieving dynamic adaptation between the traceability unit and the production disturbance. Thirdly, it provides clues for causal chain completion in production quality traceability. High-probability fracture sensitive points in the adaptation fault correspond to missing segments of the key transmission path for quality defect formation. By reverse-engineering the process parameter fragments of the missing node pairs, a complete quality fluctuation transmission chain can be reconstructed, providing a complete causal evidence chain for defect root cause location and responsibility tracing.

[0023] Example 2, as shown in Figure 1, is a cable production traceability method based on big data, which further includes the following steps: S3, extracting missing node pairs of the overlying imbalance unit in the adaptation fault, establishing a fault causal vector by extracting the fault causal features of the missing node pairs, and performing failure candidate analysis on the fault causal vector to define the applicable boundary of the traceability model; wherein, the applicable boundary of the traceability model for the missing node pairs is defined by the following method: It should be noted that the traceability model is a disturbance conduction chain; S301, extracting the fault causal features of the missing node pairs based on the cable traceability data and establishing a fault causal vector; preferably, by formula: Obtain and calculate the parameter deviation PD of the missing node pairs. i Where i represents the parameter number in the cable traceability data, for example, i=1 represents cable temperature, i=2 represents extrusion pressure, i=3 represents vulcanization time, i=4 represents wire diameter deviation, and i=5 represents dielectric loss deviation; x represents the parameter value in the cable traceability data, such as cable temperature and extrusion pressure. , This represents the maximum and minimum values ​​of the compliant range of parameters in the cable traceability data; it obtains and calculates the ratio of the number of missing node pairs in the high-disturbance interval to the total number of missing node pairs, thus obtaining the disturbance correlation ratio; it obtains the occurrence rate of missing node pairs being marked as quality defects in historical traceability, thus obtaining the same-deficiency occurrence rate; it obtains the occurrence rate of causal node pairs actually covered by each granularity unit in the traceability unit set being marked as quality defects, thus obtaining the single-deficiency occurrence rate; it calculates the deviation ratio between the same-deficiency occurrence rate and the single-deficiency occurrence rate, thus obtaining the quality correlation degree; it uses the quality correlation degree, disturbance correlation ratio, and parameter deviation as fault causal features to construct a fault causal vector containing fault causal features; S302, it performs failure candidate analysis based on the fault causal vector to locate failure candidate intervals; it divides the parameters in the cable traceability data into different process parameter intervals and establishes the feature mean vector corresponding to all process parameter intervals; it should be noted that the feature mean vector is the mean of the three parameters corresponding to the quality correlation degree, disturbance correlation ratio, and parameter deviation of all process parameter intervals, and the feature mean vector is established based on the mean of the three parameters; preferably, it is obtained through the feature concentration equation: Obtain the feature concentration coefficient (FCC) for different process parameter ranges; where k is the number of missing node pairs within the process parameter range. Let S1 be the Euclidean distance between the fault causal vector and the feature mean vector of the j-th missing node pair in the process parameter interval; the process parameter interval is screened based on feature concentration to obtain failure candidate intervals; S303, the significance of the applicable boundary of the failure candidate interval is verified to determine the final applicable boundary; preferably, the model failure confidence analysis equation is used: MFC is used to obtain the model failure confidence score; where, As a correction factor, the preferred one is... =0.8; An and Ac are the model traceability accuracy of process parameter intervals not marked as failure candidate intervals and the model traceability accuracy of failure candidate intervals, respectively; Sl is the number of failure candidate intervals; calculate the statistical significance level coefficient of the failure confidence MFC. If the statistical significance level coefficient meets the statistical significance, the failure candidate interval is determined to be an interval where the model is significantly failed. For example, when the statistical significance level coefficient is higher than 0.95, the failure candidate interval is determined to be an interval where the model is significantly failed. Then, the boundary parameter values ​​of the process parameter intervals adjacent to the significantly failed interval are extracted as the final applicable boundary. For example, if the significantly failed interval is the extrusion temperature 190℃-210℃, and its adjacent compliant process parameter interval is 170℃-190℃, then 190℃ is extracted as the final applicable boundary. It can be understood that the purpose of extracting the final applicable boundary is: first, to clarify the effective application scope of the traceability model. By extracting the final applicable boundary, the range within which the traceability model can stably function within the process parameter interval can be accurately defined, reducing invalid traceability analysis in intervals where the model significantly fails, thus improving the reliability and effectiveness of traceability results. Secondly, it provides boundary constraints for dynamic granular calibration. The final applicable boundary serves as the performance threshold of the traceability model; the boundary crossing frequency and failure confidence derivative can be used as key inputs to the granular calibration module, ensuring that the dynamic calibration of the granularity adjustment coefficient always occurs within the effective range of the model, improving the targeted nature of granularity adaptation. Thirdly, it supports precise control of the production process. Based on the final applicable boundary, it can be clearly identified in which intervals of process parameters are prone to traceability gaps, providing guidance for key parameter monitoring during production and helping to proactively avoid quality traceability failures and unclear responsibility definitions caused by parameters exceeding the applicable boundary.

[0024] S4. Extract the core features of the applicable boundary of the source tracing model, establish a coupling mapping relationship with the core features of granularity-fault, and dynamically calibrate the granularity adjustment coefficient based on the coupling mapping relationship to achieve dynamic calibration of granularity; wherein, the method of extracting the core features of the applicable boundary of the source tracing model and establishing a coupling mapping relationship with the core features of granularity-fault is as follows: S401. Obtain the core features of granularity imbalance-adaptive fault-applicable boundary and establish a core feature vector; preferably, obtain and calculate the ratio of the standard deviation to the mean of the granularity unit duration of all granularity imbalance units to obtain the granularity fluctuation coefficient; obtain and calculate the gradient of the change of the fault blocking coefficient in adjacent granularity units to obtain the blocking coefficient gradient, which is used to reflect the diffusion rate of fault influence. Rate; Obtain and calculate the frequency of process parameters crossing the applicable boundary per unit time in the applicable boundary, and obtain the boundary crossing frequency; Calculate the rate of change of model failure confidence with parameter deviation, and obtain the failure confidence derivative; It can be understood that the granularity fluctuation coefficient corresponds to the core feature of granularity imbalance, the blocking coefficient gradient corresponds to the core feature of adaptation fault, and the failure confidence derivative and boundary crossing frequency are the core features of the applicable boundary; Construct a core feature vector containing granularity fluctuation coefficient, blocking coefficient gradient, boundary crossing frequency, and failure confidence derivative; S402, Perform implicit correlation analysis on process parameters to obtain the implicit coupling strength of process parameters, and establish a coupling mapping relationship based on the core feature vector and implicit coupling strength; Preferably, through the mutual information entropy formula: Obtaining the implicit coupling strength ; This represents the strength of implicit coupling, quantifying the degree of hidden correlation between an unmonitored implicit parameter X and a monitored process parameter Y; where X is the monitored implicit parameter, i.e., a parameter that should be monitored by sensors in the process but is not actually recorded; Y is the monitored process parameter, i.e., an explicit parameter collected in real time by sensors deployed in production; x is the specific value of X, and y is the specific value of Y. The probability of X=x and Y=y occurring simultaneously is derived by back-calculating from historical fault data; in the historical data, Let X and Y represent the marginal probabilities of X taking only x and Y taking only y, respectively. It's important to explain that since X is an unmonitored latent parameter, and x is the specific value of X, the probability distribution of X is inferred from historical fault data. This involves collecting historical fault data (such as the values ​​of Y at the time of faults like cable core breakage or out-of-tolerance outer diameter, and corresponding fault cause analysis), and using Bayesian inference to establish the conditional probability relationship between X and Y—using the known values ​​of Y at the time of the fault (such as a sudden drop in outlet pressure) to infer the posterior probability of X taking different x values. Simultaneously, a quantitative relationship between X and Y is derived using a process mechanism model (such as a fluid dynamics model of the extrusion process), correcting the inference results, and finally obtaining the probability distribution of X (i.e., the probability of different x values ​​occurring). This is achieved through partial least squares equations. Establish a coupling mapping relationship between core feature vectors and implicit coupling strength; whereby, For implicit coupling strength, - These are the fitting coefficients. - Each core feature corresponds to a core feature in the core feature vector; for example For particle size fluctuation coefficient, For the gradient of the blocking coefficient, For boundary crossing frequency, For the failure confidence derivative; S403, dynamically adjust the granularity based on the coupling mapping relationship; preferably, the implicit coupling strength is... Particle size adjustment coefficient By performing combination processing, the set of parameters to be adjusted is obtained. From historical production data, obtain the average total matching degree of the granularity fit matrix for each parameter group to be adjusted, and use it as the traceability accuracy A; then, compare the traceability accuracy A with the parameter group to be adjusted. Combine them to construct the sample set to be adjusted. With the goal of maximizing the source tracing accuracy A, a continuous mapping function is constructed using a multiple linear regression algorithm. Preferably, through the mapping equation: Construct a continuous mapping function; where, This is the initial granularity adjustment factor. The influence factor of implicit coupling strength on the granularity adjustment coefficient is determined by analyzing historical samples. The positive and negative associations with A are obtained through training, such as When positively correlated with A If the value is negative, K needs to be adjusted downwards to improve accuracy; Causal perturbation rate The correction factor reflects the influence of perturbation strength on the coupling relationship, such as in the high perturbation range. Increase, weaken The adjustment range for K; after each cable production run, the newly generated sample set to be adjusted... Input to a continuous mapping function and update parameters. and This allows the mapping relationship to continuously adapt to the implicit coupling patterns of the current production system. The purpose of establishing this coupling mapping relationship is threefold: First, to achieve a quantitative correlation between implicit coupling patterns and granularity adjustments. By mapping the implicit coupling strength of process parameters, such as the correlation characteristics between unmonitored and monitored parameters, to the core characteristics of granularity-fault mapping, implicit production patterns can be transformed into calculable granularity adjustment criteria, which helps solve the problem that parameter correlation cannot directly guide granularity optimization. Second, to provide mathematical model support for dynamically calibrating granularity adjustment coefficients. The continuous mapping function constructed based on the coupling mapping relationship can correlate traceability accuracy targets with parameters such as implicit coupling strength and causal perturbation rate. The correction magnitude of the granularity adjustment coefficients is quantified through a multiple linear regression algorithm, ensuring that granularity adjustments are always guided by maximizing traceability accuracy. Third, to ensure the adaptability and long-term stability of the traceability system. The coupling mapping relationship can be continuously updated with new production data, enabling the granularity adjustment mechanism to dynamically adapt to changes in the perturbation characteristics of the production system, reducing the attenuation of traceability accuracy caused by fluctuations in production conditions.

[0025] Example 3, as shown in Figure 3, describes a cable production traceability system based on big data, comprising the following modules: Unit Construction Module: This module collects cable production traceability data, establishes a traceability model, performs variable granularity processing on the cable traceability data, and constructs a variable granularity traceability unit set; Fault Extraction Module: Based on the traceability unit set, it performs granularity matching analysis on the traceability scenarios of the traceability unit set, constructs a granularity fit matrix, extracts the fit equilibrium values ​​of different dimensions, determines whether the matching degree of different dimensions is balanced, and if unbalanced, locates the granularity imbalance fit fault; Boundary Delineation Module: This module extracts missing node pairs covering the imbalanced units in the fit fault, establishes a fault causal vector by extracting the fault causal features of the missing node pairs, performs failure candidate analysis on the fault causal vector, and defines the applicable boundary of the traceability model; Granularity Calibration Module: This module extracts the core features of the applicable boundary of the traceability model, establishes a coupling mapping relationship by combining the core features of granularity and fault, and dynamically calibrates the granularity adjustment coefficient based on the coupling mapping relationship to achieve dynamic granularity calibration.

[0026] The foregoing has provided a detailed description of one embodiment of the present invention, but this description is merely a preferred embodiment and should not be construed as limiting the scope of the invention. All equivalent variations and modifications made within the scope of the present invention should still fall within the scope of the present invention.

Claims

1. A cable production traceability method based on big data, characterized by: The process includes the following steps: collecting cable traceability data from cable production; establishing a traceability model to process the cable traceability data with variable granularity and constructing a traceability unit set with variable granularity; based on the traceability unit set, performing granularity matching analysis on the traceability scenarios of the traceability unit set, constructing a granularity fit matrix and extracting the fit equilibrium value of different dimensions to determine whether the matching degree of different dimensions is balanced. If it is unbalanced, the fit fault of granularity imbalance is located; extracting missing node pairs covering the unbalanced units in the fit fault, establishing a fault causal vector by extracting the fault causal features of the missing node pairs, and performing failure candidate analysis on the fault causal vector to define the applicable boundary of the traceability model; extracting the core features of the applicable boundary of the traceability model, establishing a coupling mapping relationship by combining the core features of granularity and fault, and dynamically calibrating the granularity adjustment coefficient based on the coupling mapping relationship to achieve dynamic calibration of granularity.

2. The cable production traceability method based on big data according to claim 1, characterized in that: The variable granularity processing is performed as follows: a source tracing model is constructed, and the disturbance transmission chain of different processes in the source tracing model and the causal disturbance interval of the disturbance transmission chain are extracted; the granularity unit duration is dynamically divided based on the causal disturbance interval.

3. The cable production traceability method based on big data according to claim 1, characterized in that: The method to determine whether the matching degree of different dimensions is balanced is as follows: the traceability scenario is divided into quality traceability scenario, process optimization scenario and responsibility definition scenario; from the traceability unit set, the granularity matching degree of each granular unit in the three traceability scenarios is obtained, and a granularity fit degree matrix with the behavior scenario column as the dimension is constructed. The fitness balance analysis is performed based on the granularity fitness matrix to obtain the fitness balance value; the fitness balance value is used to determine whether the matching degree of each dimension is balanced, and if it is not balanced, the deviation analysis dimension is extracted. Extract node coverage from the deviation analysis dimension, and filter coverage imbalance units based on node coverage; based on the coverage imbalance units, determine whether the coverage imbalance units are at the sensitive points of causal chain break.

4. The cable production traceability method based on big data according to claim 3, characterized in that: The method for extracting the deviation analysis dimensions is as follows: if the matching degree of each dimension is unbalanced, calculate the deviation ratio between the granular matching degree of each dimension and the preset ideal matching degree to obtain the matching deviation rate; filter the negative matching deviation rates, sort the negative matching deviation rates in ascending order, and select the dimension with the smallest negative matching deviation rate as the deviation analysis dimension.

5. The cable production traceability method based on big data according to claim 1, characterized in that: The method for locating the adaptation fault of the granularity imbalance is as follows: if the coverage imbalance unit is located at the break sensitive point of the causal chain, the granularity unit duration of the coverage imbalance unit is extracted to determine the granularity imbalance type of the coverage imbalance unit; traceability blocking analysis is performed for each traceability scenario to obtain the fault blocking coefficient and associate it with the granularity imbalance type, and missing node pairs are extracted to determine the adaptation fault.

6. The cable production traceability method based on big data according to claim 1, characterized in that: The failure candidate analysis is performed as follows: based on the cable tracing data, the fault causal features of missing node pairs are extracted, and a fault causal vector is established; based on the fault causal vector, failure candidate analysis is performed to locate the failure candidate interval; the significance of the applicable boundary of the failure candidate interval is verified to determine the final applicable boundary.

7. The cable production traceability method based on big data according to claim 6, characterized in that: The method for verifying the significance of the applicable boundary is as follows: obtain the model failure confidence level through the model failure confidence analysis equation; calculate the statistical significance level coefficient of the failure confidence level; if the statistical significance level coefficient satisfies the statistical significance, then determine the failure candidate interval as the interval where the model has significant failure. Extract the boundary parameter values ​​of the process parameter intervals adjacent to the intervals of significant failure as the final applicable boundaries.

8. The cable production traceability method based on big data according to claim 1, characterized in that: The coupling mapping relationship is established by: obtaining the core features of granularity imbalance, adaptation fault, and applicable boundary, and establishing a core feature vector; Implicit correlation analysis is performed on process parameters to obtain the implicit coupling strength of process parameters, and a coupling mapping relationship is established based on the core feature vector and the implicit coupling strength; The granularity adjustment coefficient is dynamically calibrated based on the coupling mapping relationship.

9. The cable production traceability method based on big data according to claim 8, characterized in that: The method for dynamically calibrating the granularity adjustment coefficient is as follows: the implicit coupling strength and the granularity adjustment coefficient are combined to obtain the parameter group to be adjusted; the mean of the total matching degree of the granularity fit matrix of each parameter group to be adjusted is obtained from historical production data as the traceability accuracy; with the goal of maximizing the traceability accuracy, a continuous mapping function is constructed through a multiple linear regression algorithm. The calibration granularity adjustment coefficient is based on the continuous mapping function.

10. A cable production traceability system based on big data, used to implement the cable production traceability method based on big data as described in any one of claims 1-9, characterized in that: The system includes the following modules: Unit Construction Module: This module collects cable traceability data from cable production, establishes a traceability model, performs variable-granularity processing on the cable traceability data, and constructs a variable-granularity traceability unit set; Fault Extraction Module: Based on the traceability unit set, it performs granularity matching analysis on the traceability scenarios of the traceability unit set, constructs a granularity fit matrix, extracts the fit equilibrium value of different dimensions, determines whether the matching degree of different dimensions is balanced, and if unbalanced, locates the granularity imbalance fit fault; Boundary Delineation Module: This module extracts missing node pairs covering imbalanced units in the fit fault, establishes a fault causal vector by extracting the fault causal features of the missing node pairs, and performs failure candidate analysis on the fault causal vector to define the applicable boundary of the traceability model; Granularity Calibration Module: This module extracts the core features of the applicable boundary of the traceability model, establishes a coupling mapping relationship by combining the core features of granularity and fault, and dynamically calibrates the granularity adjustment coefficient based on the coupling mapping relationship to achieve dynamic granularity calibration.

Images