Drag chain cable production data management method and system

Abstract

The invention relates to the technical field of data processing, in particular to a tow chain cable production data management method and system, and the method comprises the steps: obtaining multi-dimensional parameter data at each moment in the production process of a tow chain cable; and performing anomaly detection on the multi-dimensional parameter data by using an improved robust random cutting forest algorithm to obtain abnormal parameter data, and marking and managing the detected abnormal parameter data. According to the method, the problems that the detection precision is reduced and the abnormity is not recognized or misjudged in time due to the fact that the sensitivity of the existing robust random cutting forest algorithm to the multi-parameter coupling abnormity is insufficient are solved.

Description

Technical Field

[0001] This invention relates to the field of data processing technology. More specifically, this invention relates to a method and system for managing production data of drag chain cables. Background Technology

[0002] Drag chain cables, as a type of special flexible cable used in high-frequency reciprocating motion applications, are a crucial core component ensuring the stable operation of industrial automation equipment and intelligent equipment. Their reliability and durability directly affect the continuity and safety of the production line. In actual production, the manufacturing of drag chain cables involves multiple precision stages, including extrusion, stranding, cabling, and sheathing. Each stage is accompanied by dynamic changes in multi-dimensional process parameters such as temperature, speed, tension, and dimensions. These parameters exhibit complex interdependencies and coupling characteristics. For example, there are significant physical correlations between parameters such as the screw speed of the extruder and the melt temperature, the traction speed and the finished cable diameter, and the unwinding speed and tension level. Even a slight deviation or fluctuation in any single parameter can lead to insufficient material plasticization, geometrical deviations, surface defects, or even finished product failure.

[0003] With the development of modern industrial production towards high automation and intelligence, the parameter data generated during the production of drag chain cables exhibits continuous, high-frequency, and streaming characteristics. This type of data is large in volume, updates rapidly, and possesses strong multidimensional coupling properties. Against this backdrop, achieving real-time and accurate anomaly detection of streaming data during the production process has become a key technological link in ensuring product quality, reducing production risks, and improving the level of intelligent production. Traditional data analysis and anomaly detection methods often rely on static data or single-parameter analysis, which is insufficient to meet the real-time and multi-parameter coupling requirements of drag chain cable production data streams. Therefore, they suffer from problems such as detection lag, high false alarm rates, and inability to capture nonlinear relationships. Among existing technologies, the robust randomized cut forest algorithm is increasingly being applied to similar scenarios due to its ability to handle high-dimensional streaming data and calculate anomaly scores online. This algorithm constructs a forest composed of multiple randomly cut trees, evaluates each newly generated data point in real time, and calculates anomaly scores using a tree structure built from historical data, thereby achieving online anomaly detection.

[0004] While the robust randomized cut forest algorithm has certain advantages in handling continuous data streams, its original construction method assumes that each data dimension is independent when randomly selecting the cut dimensions, ignoring the strong coupling relationships between parameters in the drag chain cable production process. This independent cutting approach leads to insufficient sensitivity of the algorithm to anomalies caused by multi-parameter coupling in practical applications, easily resulting in decreased detection accuracy, failure to identify anomalies in a timely manner, or misjudgment. Consequently, it is difficult to meet the actual needs of high-precision, low-latency anomaly detection in drag chain cable production. Summary of the Invention

[0005] To address the problems of decreased detection accuracy, failure to identify anomalies in a timely manner, or misjudgment mentioned in the background art, the present invention provides solutions in the following aspects.

[0006] In a first aspect, the present invention provides a method for managing drag chain cable production data, comprising: acquiring multidimensional parameter data at various moments during the drag chain cable production process; performing anomaly detection on the multidimensional parameter data using an improved robust random cutting forest algorithm to obtain abnormal parameter data, and marking and managing the detected abnormal parameter data; the improved robust random cutting forest algorithm includes constructing a random cutting tree, specifically: calculating the correlation degree between any two dimensions of parameter data; obtaining the sensitivity between any two dimensions of parameter data based on the correlation degree; determining the node stability of each parameter data based on the sensitivity degree; and sequentially segmenting each parameter data based on the node stability to construct a random cutting tree.

[0007] The above technical solution introduces an improved robust random cut forest algorithm into the drag chain cable production data management process, which realizes accurate anomaly detection and systematic management of multi-dimensional parameter data. It solves the problem that the existing algorithm is not sensitive enough to multi-parameter coupled anomalies, and is prone to decreased detection accuracy, failure to identify anomalies in time or misjudgment.

[0008] Furthermore, the degree of correlation specifically refers to: constructing a parameter sequence from the parameter data of the target time and its multiple historical times in the same dimension to obtain multiple parameter sequences, wherein the target time is any time in each time; Then the target time Down Parameters and The correlation of the parameters is as follows: taking the target time respectively Corresponding Parameter sequence and Parameter sequence; calculation The first parameter in the sequence Each element and The first parameter in the sequence The joint probability distribution of each element; simultaneously, calculate... The first parameter in the sequence The marginal distribution probability of each element, and The first parameter in the sequence The marginal distribution probabilities of each element; construct a logarithmic ratio from the joint distribution probability and the marginal distribution probabilities, and multiply it by the joint distribution probability; finally, calculate the marginal distribution probabilities of all elements. and The calculation results are double-accumulated to obtain the target time. Down Parameters and The degree of correlation between parameters.

[0009] The above technical solution constructs parameter sequences at the target time and multiple historical time points, and calculates information content measurement based on the joint distribution and marginal distribution between sequences, thereby achieving accurate quantification of the correlation between different parameters.

[0010] Furthermore, the sensitivity level is specifically defined as: constructing the target time. Down Parameters and Relationship matrix between parameters , , , , To be respectively the target time The corresponding parameter sequence of the first The parameter and the first The parameter, the first The parameter and the first The parameter, the first The parameter and the first The parameter, the first The parameter and the first The degree of correlation between the parameters Given the size of the parameter sequence, the target time... Down Parameters and Parameter sensitivity for, , For the target time Down Parameters and The maximum value of the eigenvalues ​​of the relation matrix between parameters. For the target time Down Parameters and The first parameter of the relation matrix 1 eigenvalue, The total number of eigenvalues. and These are the preset hyperparameters.

[0011] The above technical solution constructs a relationship matrix by the degree of correlation between two parameters at the target time, and uses the distribution of matrix eigenvalues ​​to quantify the sensitivity between parameters, thereby achieving a fine characterization of the strength of parameter interaction and the dominant relationship.

[0012] Furthermore, node stability specifically refers to: obtaining the target time and multiple historical time points. The sensitivity of the parameter to other parameters, then the target time The following parameters node stability for, , Parameters at multiple historical moments With the The mean of the sensitivity between the parameters, Parameters at multiple historical moments With the The standard deviation of the sensitivity between the parameters The total number of dimensions in the multidimensional parameter data.

[0013] The above technical solution performs statistical analysis on the sensitivity of each parameter to other parameters at the target time and at multiple historical times, and combines the mean and volatility to form node stability, thereby achieving a quantitative assessment of the stability and reliability of the parameters in the overall operation.

[0014] Furthermore, the multidimensional parameter data includes: temperature, screw speed, traction speed, wire feeding speed, and current.

[0015] Furthermore, the multidimensional parameter data is filled with missing values ​​and standardized.

[0016] Furthermore, the missing value imputation is mean imputation, and the standardization process is Z-score standardization.

[0017] Furthermore, the step of marking and managing the detected abnormal parameter data includes: adding abnormal tags to the abnormal parameter data and storing them in an abnormal data management library.

[0018] The above technical solution achieves systematic recording and tracking of abnormal information by marking and managing the detected abnormal parameter data. By adding tags to each abnormal data and storing it in a dedicated management database, it can not only quickly distinguish between normal and abnormal data, facilitating real-time querying and historical backtracking, but also save the context information when the abnormality occurred, providing a basis for subsequent analysis.

[0019] Furthermore, the step of sequentially segmenting the parameter data based on node stability specifically involves: obtaining the node stability of each parameter data and sequentially segmenting the parameter data according to the magnitude of the node stability to construct a random cutting tree.

[0020] In a second aspect, the present invention provides a drag chain cable production data management system, including a memory and a processor, wherein the memory stores computer program instructions, and when the computer program instructions are executed by the processor, a drag chain cable production data management method as described above is implemented.

[0021] The beneficial effects of this invention are as follows: This invention constructs a data management method and system for drag chain cable production based on an improved robust random cutting forest algorithm, achieving online anomaly detection and systematic management of multi-dimensional production parameters. By quantifying the correlation, sensitivity, and node stability among parameters, the method accurately reveals the nonlinear coupling relationships and dynamic interaction characteristics between parameters. When constructing the random cutting tree, the dominant role of key parameters is prioritized, significantly improving the accuracy and robustness of anomaly detection. Simultaneously, the detected anomaly data is tagged and centrally managed, enabling real-time tracking, historical review, and pattern analysis of anomaly information, providing reliable data support for production anomaly early warning, process optimization, and equipment maintenance. Attached Figure Description

[0022] Figure 1 This is a schematic flowchart illustrating a drag chain cable production data management method according to an embodiment of the present invention; Figure 2 This is a schematic diagram illustrating the structural block diagram of a drag chain cable production data management system according to an embodiment of the present invention. Detailed Implementation

[0023] An embodiment of a method for managing production data of drag chain cables.

[0024] like Figure 1 The flowchart shown is a method for managing production data of drag chain cables according to an embodiment of the present invention, which includes the following steps: S1: Obtain multi-dimensional parameter data at various moments during the production process of drag chain cables.

[0025] In a preferred embodiment, the multidimensional parameter data includes: temperature, screw speed, traction speed, wire feeding speed, and current. These parameters correspond to different stages in the production process. For example, temperature reflects the thermal state of the production environment and material processing, and has a direct impact on the melting uniformity and molding quality of the material. Screw speed determines the extrusion efficiency and melt flowability; excessively high or low speeds may lead to surface defects or unstable mechanical properties in the finished product. Traction speed directly affects the diameter and geometric stability of the finished cable. Wire feeding speed determines the matching degree between raw material supply and production progress; if uncoordinated, it may cause uneven stretching or surface damage. Current, as an important indicator of energy consumption and equipment operating conditions, can indirectly reflect whether the equipment has abnormal loads or operating resistance.

[0026] By synchronously acquiring the aforementioned multidimensional parameter data, a multi-faceted and comprehensive production process dataset can be constructed, laying the foundation for subsequent anomaly detection and process optimization.

[0027] After data acquisition, missing value imputation is required to address potential missing value issues. In this technical solution, mean imputation is preferred, which uses the mean of the same parameter over historical time series to replace missing values. On one hand, this method maximizes the overall consistency of parameter distribution, avoiding bias introduced by a small number of missing data points. On the other hand, mean imputation offers advantages such as ease of calculation and stable convergence when processing large-scale continuous production data, contributing to improved accuracy in subsequent modeling and analysis.

[0028] Subsequently, to eliminate the adverse effects of differences in the dimensions and value ranges of different parameters, it is necessary to standardize the filled multidimensional parameter data. This embodiment employs the Z-score standardization method, which involves subtracting the mean from each parameter data point and then dividing by the standard deviation, resulting in a standardized data with a mean of 0 and a variance of 1. This approach not only ensures that different parameters are on the same dimension, avoiding the dominance of large numerical parameters in the calculation, but also enhances the stability and convergence of the algorithm when processing high-dimensional data.

[0029] S2: The improved robust random cut forest algorithm is used to perform anomaly detection on the multidimensional parameter data to obtain abnormal parameter data.

[0030] In a preferred embodiment, the improved robust random cut forest algorithm includes constructing a random cut tree, specifically by calculating the degree of correlation between any two-dimensional parameter data. The degree of correlation specifically refers to: constructing a parameter sequence from the parameter data of the target time and its multiple historical times in the same dimension to obtain multiple parameter sequences, where the target time is any time within each sequence; then the target time... Down Parameters and The degree of correlation of parameters for, , For the target time Corresponding The first in the parameter sequence Each element and The first in the parameter sequence The joint probability distribution of the elements. For the target time Corresponding The first in the parameter sequence Marginal distribution probabilities of each element For the target time Corresponding The first in the parameter sequence Marginal distribution probabilities of each element For the target time The total number of elements in the corresponding parameter sequence.

[0031] By constructing a multidimensional parameter sequence at the target time and its historical times, and utilizing the logarithmic ratio between the joint distribution probability and the marginal distribution probability to calculate the correlation between parameters, this is essentially an information theory-based measurement method. Its core logic is that when two parameters are statistically independent, the joint distribution will approximately equal the product of the marginal distributions, and the correlation approaches zero; however, when there is a strong dependency between them, the joint distribution will significantly deviate from the product of the marginal distributions, and the correlation value will increase accordingly. This construction not only reveals linear relationships but also captures complex nonlinear dependency patterns, thus avoiding the limitation of traditional correlation coefficients, which can only measure linear correlations. Furthermore, the formula uses a global traversal to perform a weighted summation of all pairs of elements in the parameter sequence, ensuring that the calculation results reflect the overall patterns of long-term historical data, rather than relying solely on instantaneous values ​​at a single moment, thereby improving the robustness and noise resistance of the correlation measurement.

[0032] Based on the aforementioned correlation, the sensitivity between any two dimension parameter data is obtained, specifically by constructing the target time. Down Parameters and Relationship matrix between parameters , , , , To be respectively the target time The corresponding parameter sequence of the first The parameter and the first The parameter, the first The parameter and the first The parameter, the first The parameter and the first The parameter, the first The parameter and the first The degree of correlation between the parameters Given the size of the parameter sequence, the target time... Down Parameters and Parameter sensitivity for, , For the target time Down Parameters and The maximum value of the eigenvalues ​​of the relation matrix between parameters. For the target time Down Parameters and The first parameter of the relation matrix 1 eigenvalue, The total number of eigenvalues. and To preset hyperparameters, Used to avoid a denominator of zero; This is used to improve the sensitivity of feature values, enabling them to more sensitively identify changes in feature values. In this application, the value is set to 2, but it can also be set according to the actual situation. By constructing a parameter relationship matrix at the target time, both local correlations and overall coupling structures can be preserved in a two-dimensional space, allowing the multidimensional interaction characteristics between parameters to be uniformly characterized in matrix form. Subsequently, by performing eigenvalue decomposition on the relationship matrix, the main change patterns and energy distribution contained within the matrix can be extracted from an algebraic perspective. The maximum eigenvalue corresponds to the strongest coupling pattern in the matrix, i.e., the most dominant part among all potential interaction relationships, while other smaller eigenvalues ​​represent secondary action patterns. The formula further calculates the ratio of the maximum eigenvalue to each eigenvalue and sums them up, thereby achieving a hierarchical amplification of dominant and non-dominant relationships. This mechanism ensures that once a strong dependency pattern exists, the sensitivity index is significantly increased, accurately highlighting key parameter pairs.

[0033] Based on the aforementioned sensitivity, the node stability of each parameter data is determined. Specifically, node stability involves obtaining the target time and multiple historical time values. The sensitivity of the parameter to other parameters, then the target time The following parameters node stability for, , Parameters at multiple historical moments With the The mean of the sensitivity between the parameters, Parameters at multiple historical moments With the The standard deviation of the sensitivity between the parameters The total number of dimensions in the multidimensional parameter data.

[0034] By introducing the calculation of node stability, this method analyzes the sensitivity of a single parameter to all other parameters over historical time series using mean and volatility analysis, and then aggregates them using weighted ratios to quantitatively characterize the stability of that parameter within the overall system. When the sensitivity relationship between a parameter and other parameters remains relatively stable over different time periods, its mean is high and its volatility is low, indicating improved node stability and strong predictability and reliability of the parameter during system operation. Conversely, if the sensitivity relationship between a parameter and other parameters fluctuates significantly, node stability decreases, indicating that this parameter contributes more to the overall uncertainty of the system and is more likely to become a trigger for anomalies or instability. Therefore, this method can quickly identify core parameters that play a crucial role in system stability from multidimensional monitoring data and provide a scientific basis for anomaly early warning, parameter tuning, and process optimization, thereby significantly enhancing the robustness and quality assurance capabilities of the drag chain cable production process.

[0035] The parameter data is sequentially segmented based on node stability to construct a random cutting tree. Specifically, this segmentation involves obtaining the node stability of each parameter data point and then segmenting the data sequentially according to the magnitude of the node stability to construct a random cutting tree. The process of constructing the random cutting tree is a well-known technique, and its specific details will not be elaborated upon in this solution.

[0036] S3: And mark and manage the detected abnormal parameter data.

[0037] In a preferred embodiment, the marking and management of detected abnormal parameter data is not limited to simple recording, but is achieved by establishing a complete abnormal data management mechanism. Specifically, after detecting abnormal parameter data, the data is first marked as abnormal by attaching label information related to the abnormal characteristics to the original parameter data, such as the time of the abnormality and the parameter dimensions involved. In this way, not only can abnormal data be identified and tracked in real time, but normal data can also be quickly distinguished from abnormal data in subsequent analysis, thereby improving the efficiency of data retrieval and comparison.

[0038] Furthermore, after labeling, the abnormal parameter data is categorized and managed according to different classification rules. For example, it can be classified in multiple dimensions based on the source of the abnormality (such as temperature control or current fluctuation) and the severity of the abnormality. This hierarchical classification method helps to grasp the distribution of abnormalities in the production process from a global perspective, identify high-frequency links or potential bottlenecks where abnormalities occur, and thus provide more targeted guidance for optimizing production processes and maintaining equipment.

[0039] The present invention provides a data management method and system for drag chain cable production based on an improved robust random cutting forest algorithm, achieving high-precision real-time anomaly detection and systematic management of multi-dimensional production parameters. By quantifying the correlation, sensitivity, and node stability among parameters, and constructing a random cutting tree based on these parameters, this method can fully reveal the nonlinear coupling relationships and complex interaction characteristics between parameters in the production process, thereby effectively identifying isolated anomalies and potential anomalies caused by combinations of key parameters.

[0040] Simultaneously, the detected abnormal parameters are marked and centrally managed, enabling real-time tracking, historical review, and pattern analysis of abnormal data. This provides reliable data support for production anomaly early warning, process optimization, and equipment maintenance. Overall, this invention not only improves the accuracy and real-time performance of anomaly detection but also enhances the controllability and stability of the production process, significantly ensuring the product quality and production efficiency of drag chain cables, and realizing intelligent management and efficient application of streaming production data.

[0041] An embodiment of a drag chain cable production data management system: like Figure 2 As shown in the figure, a structural block diagram of a drag chain cable production data management system according to an embodiment of the present invention includes a processor and a memory.

[0042] This invention also provides a data management system for drag chain cable production. For example... Figure 2 As shown, the system includes a processor and a memory, the memory storing computer program instructions, which, when executed by the processor, implement a drag chain cable production data management method according to the present invention.

[0043] The drag chain cable production data management system also includes other components well known to those skilled in the art, such as communication interfaces. Their settings and functions are known in the art and will not be described in detail here.

[0044] In this invention, the aforementioned memory can be any tangible medium containing or storing a program that can be used or combined with an instruction execution system, apparatus, or device. For example, a computer-readable storage medium can be any suitable magnetic or magneto-optical storage medium, such as Resistive Random Access Memory (RRAM), Dynamic Random Access Memory (DRAM), Static Random Access Memory (SRAM), Enhanced Dynamic Random Access Memory (EDRAM), High-Bandwidth Memory (HBM), Hybrid Memory Cube (HMC), etc., or any other medium that can be used to store desired information and can be accessed by an application, module, or both. Any such computer storage medium can be part of a device or accessible to or connected to a device. Any application or module described in this invention can be implemented by computer-readable / executable instructions stored or otherwise maintained on such a computer-readable medium.

[0045] In the description of this specification, "multiple" or "several" means at least two, such as two, three or more, unless otherwise explicitly specified.

[0046] While this specification has shown and described numerous embodiments of the invention, it will be apparent to those skilled in the art that such embodiments are provided by way of example only. Many modifications, alterations, and alternatives will occur to those skilled in the art without departing from the spirit and essence of the invention. It should be understood that various alternatives to the embodiments of the invention described herein may be employed in the practice of this invention.

Claims

1. A method of managing production data of a drag chain cable, characterized by, The method comprises: acquiring multi-dimensional parameter data at each moment in the production process of the drag chain cable; performing anomaly detection on the multi-dimensional parameter data by using an improved robust random cut forest algorithm to obtain abnormal parameter data, and marking and managing the detected abnormal parameter data; the improved robust random cut forest algorithm comprises constructing a random cut tree, specifically comprising: calculating the correlation degree between any two-dimensional parameter data; and obtaining the sensitivity degree between any two-dimensional parameter data based on the correlation degree; determining the node stability of each parameter data based on the sensitivity degree, and sequentially segmenting each parameter data based on the node stability to construct a random cut tree.

2. The method of claim 1, wherein, The correlation degree is specifically: constructing parameter data at the target moment and at a plurality of historical moments in the same dimension into a parameter sequence to obtain a plurality of parameter sequences, the target moment being any moment at each moment. The target time point The target time point The target time point The correlation degree of the parameters is that the target time point The target time point The target time point The target time point The target time point The target time point The target time point The target time point The target time point The target time point The target time point The target time point The target time point The target time point The target time point The target time point The target time point 3. The method of claim 2, wherein, The sensitivity degree is specifically: constructing a target moment The parameters and the relationship matrix between the parameters , , , , is the correlation degree of the first parameter in the parameter sequence corresponding to the target moment and the first parameter, the first parameter and the first parameter, the first parameter and the first parameter, the first parameter and the first parameter, is the size of the parameter sequence, and the target moment The sensitivity degree of the parameters and the parameters is , is the maximum value of the eigenvalue of the relationship matrix between the parameters at the target moment The sensitivity degree of the parameters and the relationship matrix between the parameters is the eigenvalue of the relationship matrix between the parameters at the target moment is the eigenvalue, is the total number of eigenvalues, and is a preset hyperparameter.

4. The method of claim 1, wherein, The node stability is specifically: obtaining the target moment and a plurality of historical moments The sensitivity between the parameters and other parameters, and the target moment The parameter The node stability of the parameter is, , The average of the sensitivity between the parameter and the first parameter at the plurality of historical moments, The standard deviation of the sensitivity between the parameter and the first parameter at the plurality of historical moments, The total number of dimensions of the multi-dimensional parameter data.

5. The method of claim 1, wherein, The multi-dimensional parameter data comprises temperature, screw speed, traction speed, pay-off speed and current.

6. The method of claim 1, wherein, The multi-dimensional parameter data is subjected to missing value filling and standardization processing.

7. The method of claim 6, wherein, The missing value filling is mean filling, and the standardization processing is Z-score standardization.

8. The method of claim 1, wherein, The marking and management of the detected abnormal parameter data comprises adding an abnormal label to the abnormal parameter data and storing it in an abnormal data management library.

9. The method of claim 1, wherein, The sequentially segmenting each parameter data based on the node stability is specifically: obtaining the node stability of each parameter data, and sequentially segmenting each parameter data according to the size of the node stability to construct a random cut tree.

10. A data management system for production of a drag chain cable, characterized in that, The method comprises a memory and a processor, and the memory stores computer program instructions, which, when executed by the processor, realize the method of any one of claims 1-9.

Images