Safety early warning method and system for cable production

Abstract

The present application relates to the technical field of cable production, and discloses a safety early warning method and system for cable production, the method comprising: performing multivariate regression analysis and formula optimization on raw material ratio data of cable production to obtain formula data of each component; performing Fourier transform and blending parameter calculation to obtain optimal blending parameters; performing material mixing treatment to obtain heating element particles and performing particle size analysis to obtain particle size distribution data; performing influencing factor analysis and extrusion parameter optimization to obtain target extrusion parameters; performing multi-layer structure preparation and performing multi-scale feature extraction and multi-layer preparation speed control to obtain prepared heating wire; performing performance testing to obtain performance test data and performing anomaly detection and production process parameter optimization to generate target production process parameters, thereby realizing intelligent monitoring and early warning of the entire cable production process, and significantly improving the safety and controllability of the production process.

Description

Technical Field

[0001] The present application relates to the technical field of cable production, and in particular to a safety early warning method and system for cable production. Background Art

[0002] As an important carrier of power transmission and communication in modern society, the production quality of cables is directly related to the safe and stable operation of power systems and communication networks. The cable production process is constantly optimized and the production efficiency is significantly improved, but it also brings higher quality control requirements and safety risks.

[0003] Traditional cable production mainly relies on experience and manual monitoring, which makes it difficult to timely discover and warn of potential quality problems and safety hazards. Cable production involves multiple complex links such as raw material ratio, mixing process, extrusion molding, multi-layer structure preparation, etc. There are complex mutual influences and coupling relationships between each link. How to effectively integrate multi-source heterogeneous data and establish an intelligent early warning model for the entire process still faces many challenges. In addition, the abnormal state in the cable production process is hidden and diverse, and the traditional single threshold alarm method is difficult to accurately identify complex abnormal patterns. How to build a multi-dimensional and multi-scale anomaly detection mechanism to achieve accurate identification and timely warning of various potential risks in the production process is a key issue that needs to be urgently solved in the current research on cable production safety early warning. Summary of the invention

[0004] The present application provides a safety warning method and system for cable production, thereby realizing intelligent monitoring and early warning of the entire cable production process, significantly improving the safety and controllability of the production process.

[0005] In a first aspect, the present application provides a safety early warning method for cable production, the safety early warning method for cable production comprising:

[0006] Perform multivariate regression analysis and formula optimization on the raw material ratio data of cable production to obtain the formula data of each component;

[0007] Based on the formula data, Fourier transform is performed on the mixing process data of each component and blending parameters are calculated to obtain optimal blending parameters;

[0008] Performing material mixing processing on each component based on the optimal blending parameters to obtain heating element particles, and performing particle size analysis on the heating element particles to obtain particle size distribution data;

[0009] Based on the particle size distribution data, the temperature and pressure data of the extrusion molding process are analyzed for influencing factors and extrusion parameters are optimized to obtain target extrusion parameters;

[0010] A multi-layer structure is prepared by using the target extrusion parameters, and multi-scale feature extraction and multi-layer preparation speed control are performed on thickness data of the preparation process to obtain a prepared heating wire;

[0011] The prepared heating wire is subjected to a performance test to obtain performance test data, and the performance test data is subjected to abnormality detection and production process parameter optimization to generate target production process parameters.

[0012] A second aspect of the present application provides a safety early warning system for cable production, the safety early warning system for cable production comprising:

[0013] The analysis module is used to perform multivariate regression analysis and formula optimization on the raw material ratio data of cable production to obtain the formula data of each component;

[0014] A calculation module, used for performing Fourier transform and blending parameter calculation on the mixing process data of each component based on the formula data to obtain the optimal blending parameters;

[0015] A processing module, used for performing material mixing processing on each component based on the optimal blending parameters to obtain heating element particles, and performing particle size analysis on the heating element particles to obtain particle size distribution data;

[0016] An optimization module, for performing influencing factor analysis and extrusion parameter optimization on the temperature and pressure data of the extrusion molding process based on the particle size distribution data, to obtain target extrusion parameters;

[0017] A control module, used to prepare a multi-layer structure by using the target extrusion parameters, and to perform multi-scale feature extraction and multi-layer preparation speed control on thickness data of the preparation process, so as to obtain a prepared heating wire;

[0018] The detection module is used to perform a performance test on the prepared heating wire to obtain performance test data, and perform abnormality detection and production process parameter optimization on the performance test data to generate target production process parameters.

[0019] The third aspect of the present application provides an electronic device, comprising: a memory and at least one processor, wherein the memory stores instructions; the at least one processor calls the instructions in the memory so that the electronic device executes the above-mentioned safety warning method for cable production.

[0020] A fourth aspect of the present application provides a computer-readable storage medium, wherein the computer-readable storage medium stores instructions, which, when executed on a computer, enable the computer to execute the above-mentioned safety warning method for cable production.

[0021] Compared with the prior art, the present application has the following beneficial effects: by performing multivariate regression analysis and formula optimization on raw material ratio data, the optimal ratio of each component can be determined more accurately, the utilization efficiency of raw materials can be improved, and the production cost can be reduced. The Fourier transform and blending parameter calculation method can be used to quantitatively evaluate the uniformity of the mixing process, optimize the blending parameters, and improve the quality and stability of material mixing. Combined with particle size analysis and extrusion parameter optimization, the particle size distribution of the heating original particles can be better controlled, and the stability of the extrusion molding process and the consistency of product quality can be improved. Through multi-scale feature extraction and multi-layer preparation speed control, the precise control of the multi-layer structure preparation process is achieved, and the structural uniformity and performance stability of the product are improved. By using anomaly detection and production process parameter optimization methods, abnormal situations in the production process can be discovered in time, and targeted optimization strategies can be given to improve production efficiency and product qualification rate. Through the combination of data-driven and intelligent algorithms, intelligent monitoring and early warning of the entire process of cable production are realized, which significantly improves the safety and controllability of the production process. BRIEF DESCRIPTION OF THE DRAWINGS

[0022] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative labor.

[0023] The structures, proportions, sizes, etc. illustrated in the drawings of this specification are only used to match the contents disclosed in the specification so as to facilitate understanding and reading by persons familiar with this technology. They are not used to limit the conditions under which the present invention can be implemented, and therefore have no substantive technical significance. Any structural modification, change in proportion or adjustment of size, without affecting the effects and purposes that can be achieved by the present invention, should still fall within the scope of the technical contents disclosed by the present invention.

[0024] Figure 1 It is a schematic diagram of the process of the safety early warning method for cable production provided by an embodiment of the present invention;

[0025] Figure 2 It is a schematic block diagram of the structure of a safety warning system for cable production provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0026] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0027] The flowcharts shown in the accompanying drawings are only examples and do not necessarily include all the contents and operations / steps, nor must they be executed in the order described. For example, some operations / steps may also be decomposed, combined or partially merged, so the actual execution order may change according to actual conditions.

[0028] It should also be understood that the terms used in this application specification are only for the purpose of describing specific embodiments and are not intended to limit the application. As used in this application specification and the appended claims, unless the context clearly indicates otherwise, the singular forms "a", "an" and "the" are intended to include plural forms.

[0029] It should be further understood that the term "and / or" used in the specification and appended claims of this application refers to any combination and all possible combinations of one or more of the associated listed items, and includes these combinations. Figure 1 , an embodiment of the safety early warning method for cable production in the embodiment of the present application includes:

[0030] Step 100: Perform multivariate regression analysis and formula optimization on raw material ratio data for cable production to obtain formula data of each component;

[0031] It is understandable that the execution subject of the present application can be a safety warning system for cable production, or a terminal or a server, which is not limited here. The present application embodiment is described by taking a server as the execution subject as an example.

[0032] Specifically, the influencing factors of the raw material ratio data are extracted to obtain the target influencing factors, so as to better understand the influence of different raw material components on the performance of the final product. Based on the target influencing factors, the raw material ratio data are subjected to multivariate linear regression analysis to obtain the initial regression model. The residual analysis method is used to analyze the residuals of the initial regression model to obtain the residual distribution characteristics. The residual distribution characteristics can reveal the prediction errors of the regression model under different conditions. Through these characteristics, the deficiencies of the model can be found and improved. The initial regression model is optimized by weighted least squares method according to the residual distribution characteristics to reduce the influence of the residuals on the model prediction results, and the optimized regression model is obtained. Based on the optimized regression model, the first mapping relationship between the raw material ratio and the product performance is constructed to describe the change law of the product performance under different ratio conditions. The first mapping relationship is analyzed by the ratio scheme, and the optimal ratio scheme is determined by analyzing the influence of different ratio schemes on the product performance. The optimal ratio scheme represents the best combination of raw material ratios under the premise of meeting the product performance requirements. The theoretical formula data of each component is generated according to the optimal ratio scheme. These data are derived based on the optimized regression model and are intended to guide the raw material ratio in actual production. The error distribution of the theoretical formula data is calculated to obtain the formula error distribution. The formula error distribution can help identify and quantify possible deviations in the production process. The theoretical formula data is discretized according to the formula error distribution. By considering the fluctuations and errors in actual production, it is ensured that the final formula data of each component not only conforms to the theoretical optimization results, but also has certain production operability.

[0033] Step 200: Perform Fourier transform and blending parameter calculation on the mixing process data of each component based on the formula data to obtain the optimal blending parameters;

[0034] Specifically, the mixing process data is segmented into time windows, and the mixing process data is segmented into several time series segments. The time series segments help capture the dynamic change characteristics of different time periods in the mixing process. Fast Fourier transform is performed on the time series segments to convert the mixing process data in the time domain into frequency domain feature data. The frequency domain feature data can reveal the oscillation and periodic changes of each component in the mixing process. The power spectral density is calculated based on the frequency domain feature data to obtain the frequency distribution characteristics of the mixing process, revealing the energy distribution of each frequency component in the mixing process, which helps to identify and analyze the main oscillation frequency in the mixing process. Peak detection is performed on the frequency distribution characteristics to identify the target oscillation frequency. The target oscillation frequency is a significant oscillation component in the mixing process, reflecting the periodic unevenness that may occur in the mixing process. The mixing process data is filtered using the target oscillation frequency to remove unnecessary noise and irrelevant components to obtain the filtered signal. The filtered signal is subjected to autocorrelation analysis to reveal the periodicity and repeatability in the signal and obtain a quantitative index of the mixing uniformity. The quantitative index reflects the uniformity in the mixing process and is an important parameter for measuring the mixing effect. Based on the quantitative index, the second mapping relationship between mixing uniformity and blending parameters is constructed, which links the mixing uniformity with different blending parameters to reflect how to adjust the blending parameters to optimize the mixing effect. By analyzing the blending parameters of the second mapping relationship, the optimal blending parameters are finally determined. The optimal blending parameters are the parameter combination that is most conducive to improving production efficiency and product quality while meeting the mixing uniformity requirements.

[0035] Step 300: performing material mixing processing on each component based on the optimal blending parameters to obtain heating element particles, and performing particle size analysis on the heating element particles to obtain particle size distribution data;

[0036] Specifically, each component is weighed according to the optimal blending parameters to obtain the raw materials to be mixed. The raw materials to be mixed are fully mixed to obtain the initial mixture. The uniformity of the initial mixture affects the stability of the material performance during the subsequent processing. The initial mixture is subjected to twin-screw extrusion. The twin-screw extrusion process can apply sufficient shear force and mixing force to the material to ensure that each component is evenly dispersed at the microscopic level to obtain a uniformly mixed composite material. The uniformly mixed composite material has excellent physical properties in the cable production process. During the extrusion process, the temperature and pressure of the material need to be strictly controlled to prevent thermal decomposition or performance degradation of the material. After the extrusion is completed, the uniformly mixed composite material is cooled and processed into initial original particles through a granulation device. The parameter control during the cooling granulation process affects the size and morphology of the particles. The initial original particles are subjected to laser particle size analysis to measure the size distribution of the particles and obtain the initial distribution data, which includes the size of the particles, the distribution range and other key information related to the particle size. By performing data analysis on the initial distribution data, the characteristic parameters of the particle size distribution are extracted. The characteristic parameters are the most important indicators in particle size analysis and can reflect the distribution morphology and concentration of the particles. Based on the characteristic parameters, a probability density function of particle size distribution is constructed. The probability density function can describe the distribution probability of particles in different size ranges. The probability density function is iteratively optimized using the expectation maximization algorithm, and the optimal solution is gradually approached through multiple iterations to obtain the optimized particle size distribution model. The initial original particles are screened according to the optimized particle size distribution model, and the particles that meet the target particle size range are screened out to obtain the heating original particles in the target particle size range. Image analysis is performed on the heating original particles in the target particle size range to confirm the morphology and distribution of the particles and obtain the particle size distribution data.

[0037] Step 400: Based on the particle size distribution data, the temperature and pressure data of the extrusion molding process are analyzed for influencing factors and extrusion parameters are optimized to obtain target extrusion parameters;

[0038] Specifically, the distribution characteristics of the particle size distribution data are extracted to obtain the key characteristics of the particle size distribution, such as the average particle size, distribution width and uniformity of particle morphology. Based on the extracted particle size distribution characteristics, the temperature and pressure data in the extrusion molding process are stratified sampled to obtain a highly representative target sample set. The target sample set retains the key data in the extrusion process and ensures the extensiveness and representativeness of the data through stratified sampling. Using the target sample set, a third mapping relationship between temperature, pressure and extrusion parameters is constructed. This mapping relationship can describe the influence of extrusion parameters on product quality under different temperature and pressure conditions. The weight coefficients of these factors in the third mapping relationship are determined by sensitivity analysis. The results of the sensitivity analysis reveal the relative importance of factors such as temperature and pressure to extrusion parameters. Based on the weight coefficient, the third mapping relationship is regularized to obtain a regularized regression model. Regularization can effectively prevent model overfitting and ensure that the model can stably predict appropriate extrusion parameters under different production conditions. Based on the regularized regression model, a fourth mapping relationship between extrusion parameters and product quality is constructed to reveal how extrusion parameters directly affect the quality characteristics of the final product, such as the surface finish, mechanical strength and dimensional accuracy of the product. The extrusion parameters of the fourth mapping relationship are analyzed to obtain the initial optimized extrusion parameters, which significantly improves product quality and production efficiency. The initial optimized extrusion parameters are optimized. Through the optimization solution, the optimal combination of extrusion parameters is found under various production conditions, thereby ensuring that in the actual production process, regardless of temperature fluctuations or pressure changes, the extrusion molding process can stably output high-quality products. The target extrusion parameters finally obtained can not only maximize product quality, but also improve production efficiency and reduce potential safety hazards in the production process.

[0039] Step 500, preparing a multi-layer structure by using target extrusion parameters, and performing multi-scale feature extraction and multi-layer preparation speed control on thickness data of the preparation process to obtain a prepared heating wire;

[0040] Specifically, according to the target extrusion parameters, the tinned soft copper stranded conductor is heated and coated to obtain the initial coating structure. In order to ensure the quality of the coating process, the initial coating structure is scanned in real time by X-ray to obtain the initial thickness distribution data. The thickness distribution of the coating layer is detected by X-ray scanning technology. The initial thickness distribution data is subjected to wavelet transform. Wavelet transform is a signal processing tool that can decompose the thickness distribution data into feature information of different scales and reveal the multi-level detail features in the data. The thickness distribution of the coating layer is reflected by analyzing the multi-scale feature information. Adaptive Kalman filter analysis is performed using the multi-scale feature information to generate a thickness prediction model. The thickness prediction model can predict the thickness change trend of the coating layer at a future moment based on the real-time monitoring data. Based on the thickness prediction model, the extrusion speed is adjusted in real time to ensure the thickness uniformity and quality stability of the coating layer. Through real-time adjustment, the optimized coating structure is obtained, so that the coating layer is more in line with the predetermined design requirements. Based on the optimized coating structure, the polyolefin insulation layer is extruded to form an insulation layer structure. The polyolefin insulation layer is an important component of the cable, and its thickness and uniformity have a direct impact on the insulation performance of the cable. The insulation layer structure is woven with tinned copper wire to form a shielding layer structure. The shielding layer structure can effectively prevent external electromagnetic interference and ensure the signal transmission quality of the cable. After the shielding layer is woven, the fluoropolymer sheath is extruded to obtain a multi-layer cable semi-finished product. The fluoropolymer sheath not only has excellent heat resistance and chemical resistance, but also can provide additional mechanical protection for the cable and enhance the overall durability of the cable. After the multi-layer semi-finished product is formed, the online thickness measurement technology is used to monitor the thickness in real time to obtain real-time thickness data. Based on the real-time thickness data, the fuzzy PID control algorithm is used to optimize and adjust the production process to control various process parameters in the preparation process. Through fuzzy PID control, the stability of the production process can be maintained in a dynamic environment, ensuring that the final heating wire has high quality and consistent thickness distribution.

[0041] Step 600: Perform a performance test on the prepared heating wire to obtain performance test data, perform anomaly detection and production process parameter optimization on the performance test data, and generate target production process parameters.

[0042] Specifically, the prepared heating wire is tested for tensile strength and elongation at break to obtain the mechanical property data of the heating wire, reflecting the strength and ductility of the heating wire under tensile conditions. The resistance temperature coefficient of the heating wire is tested to obtain its electrothermal performance data to evaluate the resistance change characteristics of the heating wire at different temperatures. A hot and cold cycle test is performed to obtain the power stability data of the heating wire. This test evaluates the power output stability of the heating wire in a repeated temperature change environment through multiple hot and cold alternating cycles. The collected mechanical property data, electrothermal performance data and power stability data constitute a multidimensional feature vector of the heating wire. The multidimensional feature vector is subjected to feature dimensionality reduction to reduce the complexity of the data and extract the most representative feature data. The feature data after dimensionality reduction is input into a pre-set isolation forest model for anomaly detection. The isolation forest model is a machine learning algorithm specifically used to detect abnormal samples, which can identify potential abnormal situations through the isolation of data. Through this model, abnormal samples in the performance test are identified and classified into an abnormal sample set. The abnormal sample set is analyzed, and association rules are mined according to the production parameters corresponding to these samples. By mining association rules, the relationship between abnormal samples and specific parameters in their production process is revealed, so as to find out the possible causes of abnormalities. Bayesian network analysis is performed based on the results of abnormal cause analysis. Bayesian network is a tool based on probabilistic reasoning. By modeling the causal relationship between different parameters, the process factors most likely to cause abnormalities are derived and corresponding optimization strategies are proposed. According to the process parameter optimization strategy, the target production process parameters are generated to minimize the probability of abnormalities in the production process and ensure the performance stability and reliability of the heating wire in production and use.

[0043] Perform principal component analysis on the multidimensional feature vector and calculate the contribution rate of each principal component. The principal component contribution rate reflects the ability of each principal component to explain the variance of the original data. According to the principal component contribution rate, select the principal component with a cumulative contribution rate greater than the preset threshold to ensure that the selected principal component can effectively retain the important information in the original data and obtain the feature data after dimensionality reduction. Standardize the feature data after dimensionality reduction to eliminate the differences between different feature dimensions so that each feature has equal importance in subsequent analysis. Perform cluster analysis on the standardized feature data to find samples with high similarity in the data set and determine the initial cluster center. Through cluster analysis, the data is divided into multiple clusters according to its internal structure, and each cluster corresponds to an initial cluster center. Based on the initial cluster center, calculate the local outlier factor of the standardized feature data to obtain the abnormality of the sample. The local outlier factor measures the density difference between each sample and its neighboring samples to identify samples with abnormal behavior relative to the surrounding samples. Construct a candidate set of abnormal samples based on the abnormality, and include those samples with higher abnormality in further analysis. Perform support vector data description analysis on the candidate set of abnormal samples to generate abnormal boundaries. Support vector data description constructs a minimized spherical boundary to surround normal samples, thereby identifying abnormal samples outside the boundary. According to the obtained abnormal boundary, the candidate set of abnormal samples is screened, and those samples inside the boundary are eliminated to obtain preliminary abnormal samples. The preliminary abnormal samples are analyzed using the isolation forest model. The isolation forest model detects abnormal samples by constructing isolation trees. Isolation trees are constructed based on preliminary abnormal samples, and multiple subtrees are generated by randomly splitting these isolation trees. In each subtree, the number of times the sample is split is reflected in the path length. The shorter the path, the more likely it is an abnormal sample. The path length of the subtree is calculated to obtain the abnormal score of each sample. According to the abnormal score, the preliminary abnormal samples are sorted and the samples with the highest score are identified. In order to ensure the flexibility and accuracy of anomaly detection, a dynamic threshold is set based on the sorting results of abnormal samples, and the abnormal sample set is finally determined. The dynamic threshold can be adjusted according to actual needs, making anomaly detection more adaptable to different production environments and data distributions.

[0044] The abnormal sample set is feature encoded, and the data of the abnormal sample is converted into a coding format suitable for data mining. Frequent item set mining is performed on the encoded abnormal sample data to identify the feature combinations that frequently appear in the abnormal sample set, reflecting the potential association relationship and pattern. According to the frequent item set, a candidate set of association rules is constructed, which contains possible association rules and describes the potential relationship between different features. By calculating the support and confidence of the candidate set of association rules, those association rules with high statistical significance are preliminarily screened out. Support measures the frequency of occurrence of the rule in the entire data set, while confidence measures the probability of the result occurring when the premise occurs. Through these two indicators, which rules have stronger reliability and representativeness are identified. The preliminary association rules are analyzed for lift. Lift is an important indicator for measuring the strength of association rules, reflecting the degree of increase in the probability of the result occurring relative to its natural occurrence probability when considering the association relationship. Through lift analysis, more accurate association rule evaluation results are obtained, and the preliminary association rules are screened according to the evaluation results to retain those effective association rules with high lift and practical significance. Based on the effective association rules after screening, a causal graph is constructed to intuitively display the causal relationship between each feature in the abnormal sample, helping to identify the potential cause of the abnormality. The causal graph is topologically sorted to obtain the hierarchical structure of the abnormal causes and reveal the hierarchical relationship between each abnormal cause. Markov blanket identification is performed based on the abnormal cause hierarchy. Markov blanket is a tool for identifying local dependencies, which can effectively find the direct dependency between a variable and other variables. Based on the identified local dependencies, a Bayesian network structure is constructed. Bayesian network is a model based on probabilistic reasoning. By describing the conditional independence and dependency between variables, it can effectively infer the possibility and potential causes of abnormality. Based on the constructed Bayesian network structure, a process parameter optimization strategy is formulated. Through the reasoning of the Bayesian network, it is identified which process parameters have a key impact on the occurrence of abnormalities, so as to propose targeted optimization strategies. These strategies can help adjust the key parameters in the production process, reduce the probability of abnormalities, and improve production stability and product quality.

[0045] In the embodiment of the present application, by performing multivariate regression analysis and formula optimization on the raw material ratio data, the optimal ratio of each component can be determined more accurately, the utilization efficiency of raw materials can be improved, and the production cost can be reduced. Using Fourier transform and blending parameter calculation method, the uniformity of the mixing process can be quantitatively evaluated, the blending parameters can be optimized, and the quality and stability of material mixing can be improved. Combined with particle size analysis and extrusion parameter optimization, the particle size distribution of the heating original particles can be better controlled, and the stability of the extrusion molding process and the consistency of product quality can be improved. Through multi-scale feature extraction and multi-layer preparation speed control, accurate control of the multilayer structure preparation process is achieved, and the structural uniformity and performance stability of the product are improved. Using anomaly detection and production process parameter optimization methods, abnormal situations in the production process can be discovered in time, and targeted optimization strategies are given to improve production efficiency and product qualification rate. Through the combination of data-driven and intelligent algorithms, intelligent monitoring and early warning of the entire process of cable production are realized, which significantly improves the safety and controllability of the production process.

[0046] In a specific embodiment, the process of executing step 100 may specifically include the following steps:

[0047] The influencing factors of the raw material ratio data of cable production are extracted to obtain the target influencing factors, and the raw material ratio data are subjected to multivariate linear regression analysis based on the target influencing factors to obtain the initial regression model;

[0048] Perform residual analysis on the initial regression model to obtain residual distribution characteristics, and optimize the initial regression model using weighted least squares method based on the residual distribution characteristics to obtain an optimized regression model;

[0049] A first mapping relationship between raw material ratio and product performance is constructed based on the optimized regression model, and a ratio scheme analysis is performed on the first mapping relationship to obtain an optimal ratio scheme;

[0050] Generate theoretical formula data of each component according to the optimal ratio scheme, and calculate the error distribution of the theoretical formula data to obtain the formula error distribution;

[0051] The theoretical formula data is discretized according to the formula error distribution to obtain the formula data of each component.

[0052] Specifically, the influencing factors of the raw material ratio data for cable production are extracted, and the key factors that have a significant impact on product performance are identified from the raw material ratio data to obtain the target influencing factors. The process of extracting influencing factors usually includes calculating the contribution of each raw material component to the performance of the final product, which can be done through statistical analysis or machine learning algorithms. For example, the correlation coefficient or regression coefficient is used to measure the degree of influence of each component on product performance, and those raw material components that have a significant impact on performance indicators (such as tensile strength, heat resistance, etc.) are selected as target influencing factors. Based on the target influencing factors, a multivariate linear regression analysis is performed on the raw material ratio data to establish an initial regression model. The form of the multivariate linear regression model is:

[0053] ;

[0054] in, Indicates the expected value of product performance, is the intercept of the model, For each raw material component The regression coefficient of is the error term. By minimizing the error term , regression analysis can find the regression coefficient that best fits the existing data and establish an initial regression model. Perform residual analysis on the initial regression model to check the difference between the model prediction value and the actual observation value (i.e., residual), and identify potential deficiencies of the model by analyzing the residual distribution characteristics. Defined as:

[0055] ;

[0056] in, is the actual observed value, is the predicted value of the model. By plotting the residual graph and observing whether the residuals are randomly distributed or whether there are systematic deviations, the quality of the model fit can be judged. If the residual distribution shows certain specific patterns or trends, it indicates that the initial regression model is imperfect and the model needs to be optimized. The weighted least squares method is used to correct the deviations in the model. The weighted least squares method gives different weights to different residuals to reduce the impact of large residuals on the model and obtain an optimized regression model. The optimized regression model is in the form of:

[0057] ;

[0058] in, It is with The weight corresponding to the sample residual, the larger the weight, the smaller the impact of the sample on the model. The optimized regression model is used to predict the impact of raw material ratio on product performance. Based on the optimized regression model, the first mapping relationship between raw material ratio and product performance is constructed. This mapping relationship is used to simulate the impact of different ratio schemes on product performance, so as to perform ratio scheme analysis. The goal of ratio scheme analysis is to find the optimal ratio scheme, that is, the ratio combination that can maximize or minimize certain key performance indicators. For example, in cable production, it may be desirable to maximize tensile strength while keeping resistivity within a certain range. In this process, the model can help evaluate product performance under different ratio schemes and select the optimal ratio scheme. Generate theoretical formula data for each component based on the optimal ratio scheme. The theoretical formula data is based on the optimized regression model and the optimal ratio scheme, and represents the optimal ratio of each component under ideal conditions. The error distribution of the theoretical formula data is calculated to evaluate the impact of these errors on the final formula. The error distribution is calculated by statistical methods, such as estimating the standard error or variance of each component by analyzing multiple experimental data. Assume that the theoretical ratio of a component is , its error distribution is normal distribution, then the actual ratio may follow the following distribution:

[0059] ;

[0060] in, Indicates the standard error of the ratio. Through error distribution calculation, the possible deviation range of the ratio of each component in actual production is determined. According to the calculated formula error distribution, the theoretical formula data is discretized to make it easier to operate and implement in actual production. Discretization is to convert the continuous theoretical ratio into a set of operable discrete values ​​to obtain the final formula data of each component.

[0061] In a specific embodiment, the process of executing step 200 may specifically include the following steps:

[0062] Based on the formula data, the mixing process data of each component is segmented into time windows to obtain time series segments, and the time series segments are fast Fourier transformed to obtain frequency domain feature data;

[0063] The power spectrum density is calculated according to the frequency domain characteristic data to obtain the frequency distribution characteristics of the mixing process, and the peak value of the frequency distribution characteristics is detected to obtain the target oscillation frequency;

[0064] According to the target oscillation frequency, the mixing process data is filtered to obtain a filtered signal, and the filtered signal is subjected to autocorrelation analysis to obtain a quantitative index of mixing uniformity;

[0065] A second mapping relationship between mixing uniformity and blending parameters is constructed based on quantitative indicators, and blending parameters are analyzed on the second mapping relationship to obtain optimal blending parameters.

[0066] Specifically, the mixing process data of each component is segmented into time windows based on the recipe data, and the continuous mixing process data is divided into multiple time series segments. The choice of time window determines the resolution of data analysis. The appropriate time window length can capture the key dynamic changes in the mixing process while avoiding excessive noise interference. Assume that the mixing process data is , whose time series fragment can be expressed as ,in Represents different time windows. Fast Fourier transform is performed on the time series segments to convert the time domain signal into the frequency domain signal to obtain the frequency domain feature data. Fast Fourier transform can map the time variation characteristics of the mixing process into the frequency space, so that the system can identify the main frequency components of the mixing process. For each time series segment , its Fourier transform can be expressed as:

[0067] ;

[0068] in, is the frequency domain feature data, is the frequency, is an imaginary unit. By performing a fast Fourier transform on each time series segment, multiple frequency domain feature data are obtained to reflect the frequency components and their intensities in the mixing process. Based on the frequency domain feature data, the power spectrum density is calculated to obtain the frequency distribution characteristics of the mixing process. The power spectrum density is a tool to measure the power distribution of a signal at different frequencies, which can reveal the main oscillation frequencies and their intensities in the mixing process. Power spectrum density The calculation formula is:

[0069] ;

[0070] in, Indicates frequency The power spectral density at is the length of the time window. By calculating the power spectral density, the frequency distribution diagram of the mixing process is obtained. After obtaining the frequency distribution characteristics, peak detection is performed to identify the target oscillation frequency in the mixing process. The target oscillation frequency refers to the frequency with a significant peak in the power spectral density. These frequencies may correspond to uneven oscillation or fluctuation phenomena in the mixing process. By identifying the target oscillation frequency, we can better understand the potential problems or unevenness in the mixing process. According to the identified target oscillation frequency, the mixing process data is filtered to remove unnecessary noise or interference components in the mixing process and retain the frequency components that are most useful for uniformity analysis. Usually a bandpass filter is used to select the frequency band near the target oscillation frequency for retention. The filtered signal is smoother and better reflects the key features of the mixing process. Autocorrelation analysis is performed on the filtered signal to quantify the uniformity of the mixing process. Autocorrelation analysis can reveal the periodicity and repeatability in the signal and is an important tool for measuring signal uniformity. Autocorrelation function The calculation formula is:

[0071] ;

[0072] in, The lag time is The autocorrelation value at is the length of the signal, is the filtered signal. If the autocorrelation value is at a small lag time The signal decays rapidly under the influence of the autocorrelation analysis, indicating that the uniformity of the signal is good; otherwise, it indicates that there are significant periodic fluctuations in the signal and the uniformity is poor. Based on the uniformity quantitative index obtained by autocorrelation analysis, the second mapping relationship between mixing uniformity and blending parameters is constructed to reflect the influence of different blending parameters (such as mixing speed, mixing time, etc.) on uniformity. The blending parameters are analyzed for the second mapping relationship. By adjusting and optimizing the blending parameters, the uniformity in the mixing process is optimized.

[0073] In a specific embodiment, the process of executing step 300 may specifically include the following steps:

[0074] Weighing each component according to the optimal blending parameters to obtain raw materials to be mixed, and mixing the raw materials to be mixed to obtain an initial mixture;

[0075] Extruding the initial mixture through a twin-screw extruder to obtain a uniformly mixed composite material, and cooling and granulating the uniformly mixed composite material to obtain initial raw material particles;

[0076] Performing laser particle size analysis on the initial original particles to obtain initial distribution data, and performing data analysis based on the initial distribution data to obtain characteristic parameters of the particle size distribution;

[0077] A probability density function of particle size distribution is constructed based on characteristic parameters, and the probability density function is iterated by an expectation maximization algorithm to obtain an optimized particle size distribution model;

[0078] The initial original particles are screened according to the optimized particle size distribution model to obtain heating original particles within a target particle size range, and image analysis is performed on the heating original particles within the target particle size range to obtain particle size distribution data.

[0079] Specifically, each component is weighed according to the optimal blending parameters to obtain the raw materials to be mixed. After weighing, the raw materials to be mixed are placed in a mixing device for mixing to obtain an initial mixture. The mixing process is carried out according to preset mixing time and speed parameters to ensure that all components are evenly distributed in the mixture. After mixing, the initial mixture usually presents a uniform appearance, and the initial mixture is subjected to twin-screw extrusion. Twin-screw extrusion is a commonly used mixing and extrusion process, in which the mixture is pushed and further mixed by two intermeshing screws to form a uniform composite material. During the twin-screw extrusion process, parameters such as extrusion speed, temperature and screw speed need to be precisely controlled to prevent material degradation or performance loss. The uniformly mixed composite material obtained by this process is usually used as the base material for the outer sheath or insulation layer in cable production. After extrusion, the composite material is subjected to cooling granulation treatment to obtain initial original particles. The purpose of cooling granulation is to quickly cool the composite material in a hot melt state and cut it into particles of a certain size. The obtained initial original particles usually have a certain particle size distribution. The initial original particles are analyzed for particle size using a laser particle size analyzer. Laser particle size analysis is a technique for measuring particle size distribution. It uses a laser beam to illuminate particles, measure the angle and intensity of scattered light, and infer the size of the particles. Through this process, the initial distribution data is obtained, which is usually expressed as a histogram or probability density function of the particle size distribution, showing the distribution of particles in different size ranges. Data analysis is performed based on the initial distribution data to extract key particle size distribution characteristic parameters. Characteristic parameters include average particle size, median diameter, standard deviation, etc., which can describe the shape, concentration and uniformity of the particle size distribution. Through these characteristic parameters, the quality and consistency of the particles are preliminarily evaluated. Based on the extracted characteristic parameters, a probability density function of the particle size distribution is constructed. The probability density function can more comprehensively describe the distribution of particle size, especially for multi-peak distribution or asymmetric distribution, and can provide more detailed information than simple mean and standard deviation. The constructed probability density function is iteratively optimized by the expectation maximization algorithm. The expectation maximization algorithm is a statistical method commonly used for parameter estimation. By repeatedly adjusting the model parameters and maximizing the likelihood function of the observed data, the optimal particle size distribution model is obtained. The optimization process usually includes two steps: E step to calculate the expected value and M step to maximize the likelihood function. After multiple iterations, the optimized particle size distribution model is obtained. The initial original particles are screened according to the optimized particle size distribution model to obtain the heated original particles that meet the target particle size range. The screening process removes particles that do not meet the particle size requirements by setting a screen or using an airflow classification device to ensure that the particle size of the final product is within the specified range. Image analysis is performed on the heated original particles in the target particle size range. Image analysis technology can provide morphological characteristics of particles, such as particle shape, surface roughness, etc.These morphological characteristics are also important factors affecting particle performance. By combining them with particle size distribution data, the quality of particles can be more comprehensively evaluated.

[0080] In a specific embodiment, the process of executing step 400 may specifically include the following steps:

[0081] Extracting distribution characteristics of particle size distribution data to obtain particle size distribution characteristics, and performing stratified sampling on temperature and pressure data of the extrusion molding process according to the particle size distribution characteristics to obtain a target sample set;

[0082] A third mapping relationship among temperature, pressure and extrusion parameters is constructed according to the target sample set, and a sensitivity analysis is performed on the third mapping relationship to obtain the weight coefficient of each influencing factor;

[0083] Regularizing the third mapping relationship according to the weight coefficient to obtain a regularized regression model, and constructing a fourth mapping relationship between the extrusion parameters and the product quality based on the regularized regression model;

[0084] An extrusion parameter analysis is performed on the fourth mapping relationship to obtain initial optimized extrusion parameters, and an optimization solution is performed based on the initial optimized extrusion parameters to obtain target extrusion parameters.

[0085] Specifically, the particle size distribution data is subjected to distribution feature extraction, and key characteristic parameters such as average particle size (D50), standard deviation, skewness and kurtosis are extracted from the particle size distribution. These characteristic parameters can fully reflect the distribution of particles and provide an important basis for subsequent analysis. For example, the average particle size reflects the overall size level of the particles, the standard deviation describes the dispersion of the particle size, and the skewness and kurtosis reveal the symmetry and peak degree of the distribution. The temperature and pressure data of the extrusion molding process are stratified and sampled according to the particle size distribution characteristics. Stratified sampling is a statistical method that aims to ensure that representative samples are drawn from different characteristic levels to form a target sample set that can fully reflect the overall data characteristics. For example, if the product extruded under different temperature and pressure conditions exhibits different particle size distribution characteristics, samples should be drawn from these conditions separately to ensure that these diverse characteristics are included in the target sample set. Through stratified sampling, a target sample set is formed, which contains the corresponding extrusion parameters under various temperature and pressure conditions and their effects on particle size distribution. According to the target sample set, a third mapping relationship between temperature, pressure and extrusion parameters is constructed. The third mapping relationship is a multivariate regression model that describes how temperature, pressure, and extrusion parameters jointly affect the particle size distribution characteristics. For example, assuming that the temperature, pressure, and extrusion speed during the extrusion process are , and , then the third mapping relationship can be expressed as:

[0086] ;

[0087] in, is the average particle size, is the intercept, , , are the regression coefficients of temperature, pressure and extrusion speed, respectively. is the error term. By fitting the model, the quantitative effects of temperature, pressure and extrusion parameters on the particle size distribution characteristics are obtained. A sensitivity analysis is performed on the third mapping relationship, and the weight coefficients of each influencing factor, that is, their relative contribution to the quality of the final product, are determined by calculating the standardized regression coefficient. The standardized regression coefficient standardizes the effects of different variables, allowing the system to directly compare the relative importance of different influencing factors. For example, if the standardized regression coefficient of temperature is significantly higher than that of pressure and extrusion speed, it indicates that temperature has a greater impact on particle size distribution. After obtaining the weight coefficients of each influencing factor, the third mapping relationship is regularized to prevent the model from overfitting. Regularization is a commonly used regression technique that introduces a penalty term to constrain the size of the regression coefficient, thereby improving the generalization ability of the model. Common regularization methods include Lasso regression and ridge regression. The form of the regularized regression model is:

[0088] ;

[0089] in, is the actual observed value, It is The sample eigenvalues, is the regression coefficient, is the regularization parameter. By adjusting , control the size of the regression coefficient to obtain a more stable model. Based on the regularized regression model, the fourth mapping relationship between extrusion parameters and product quality is constructed. The fourth mapping relationship aims to quantify the impact of extrusion parameters on the quality of the final product (such as tensile strength, surface finish, etc.). For example, in cable production, changes in extrusion speed and temperature will significantly affect the thickness and uniformity of the insulation layer, which are directly related to the insulation performance and durability of the cable. The fourth mapping relationship is analyzed for extrusion parameters, and the extrusion parameters are preliminarily optimized to improve product quality. By simulating different extrusion conditions and analyzing their impact on product quality, a set of initial optimized extrusion parameters are identified. The optimization solution is performed based on the initial optimized extrusion parameters, which is implemented through a variety of algorithms, such as gradient descent or genetic algorithms, which iteratively search for parameter combinations that can maximize or minimize the objective function.

[0090] In a specific embodiment, the process of executing step 500 may specifically include the following steps:

[0091] The tinned soft copper stranded conductor is coated with a heating element according to target extrusion parameters to obtain an initial coating structure, and the initial coating structure is scanned in real time by X-ray to obtain initial thickness distribution data;

[0092] Perform wavelet transform on the initial thickness distribution data to obtain multi-scale feature information, and perform adaptive Kalman filter analysis based on the multi-scale feature information to generate a thickness prediction model;

[0093] The extrusion speed is adjusted in real time based on the thickness prediction model to obtain an optimized coating structure, and a polyolefin insulation layer is extruded on the optimized coating structure to obtain an insulation layer structure;

[0094] The insulating layer structure is braided with tinned copper wire to obtain a shielding layer structure, and a fluorine-containing polymer sheath is extruded on the shielding layer structure to obtain a multi-layer structure semi-finished product;

[0095] The thickness of the multi-layer semi-finished product is measured online to obtain real-time thickness data, and fuzzy PID control optimization is performed based on the real-time thickness data to obtain the prepared heating wire.

[0096] Specifically, according to the target extrusion parameters, the tinned soft copper stranded conductor is coated with a heating element. This step involves passing the conductor through an extruder and uniformly coating the heating material on the surface of the conductor at a high temperature to form an initial coating structure. The initial coating structure is scanned in real time by X-rays. X-ray scanning technology can non-destructively detect the thickness distribution of the coating layer, thereby providing instant feedback during the production process. The initial thickness distribution data obtained by X-ray scanning is usually represented in the form of a two-dimensional image or digital signal, showing the thickness of the coating layer at different positions. The initial thickness distribution data is subjected to a wavelet transform, and the thickness data is decomposed into feature information of different scales, revealing local details and global trends in the data. The result of the wavelet transform can usually be expressed as:

[0097] ;

[0098] in, are the wavelet transform coefficients, is the initial thickness distribution signal, is the mother wavelet function, and They are scale and translation parameters respectively. By adjusting and , wavelet transform can capture the changing characteristics of thickness data at different times and frequencies. Based on the multi-scale feature information obtained by wavelet transform, adaptive Kalman filter analysis is performed to generate a thickness prediction model. Kalman filter is a recursive estimation algorithm widely used in dynamic systems, which can make the best estimate of the system state in the presence of noise. Through adaptive Kalman filtering, the parameters of the prediction model are continuously adjusted according to the real-time thickness data to achieve accurate prediction of the coating thickness. The basic form of Kalman filter is:

[0099] ;

[0100] in, It is at the moment An estimate of the thickness, is the Kalman gain, is the measured value, is the measurement matrix. Kalman gain Dynamic adjustment is made according to the ratio of measurement noise and system noise to achieve adaptive filtering. The extrusion speed is adjusted in real time based on the thickness prediction model to obtain the optimized coating structure. By adjusting the extrusion speed, the thickness deviation caused by temperature change or material fluctuation is compensated to ensure the uniform thickness of the coating layer over the entire conductor. This process is a dynamic control process, and the thickness prediction model needs to be closely integrated with the extrusion speed control system to achieve real-time adjustment. The optimized coating structure is extruded with a polyolefin insulation layer. Polyolefin materials are widely used in cable production due to their excellent electrical insulation properties and mechanical durability. The extrusion process is similar to the heating element coating in the previous step, but due to different materials and functions, extrusion parameters such as temperature and pressure may need to be readjusted. By controlling the extrusion process, the thickness and uniformity of the polyolefin insulation layer meet the design requirements to form a high-quality insulation layer structure. The insulation layer structure is woven with tinned copper wire to form a shielding layer structure. The main function of the shielding layer is to prevent electromagnetic interference, thereby ensuring the stability of the cable during signal transmission. The weaving process requires that the copper wires are arranged closely and evenly, which must ensure the shielding effect without affecting the flexibility of the cable. After the shielding layer structure is completed, the fluoropolymer sheath is extruded. Fluoropolymer has excellent high temperature resistance, corrosion resistance and wear resistance, providing additional mechanical protection for the cable to form the final multi-layer semi-finished product. The multi-layer semi-finished product is measured online. The online thickness measurement technology can detect the thickness of each layer of material in real time and transmit the data to the control system for dynamic adjustment. The real-time thickness data obtained by online thickness measurement is optimized by the fuzzy PID control algorithm. Fuzzy PID control combines the advantages of traditional PID control and the flexibility of fuzzy logic, and can achieve stable control in systems with strong nonlinearity and uncertainty. The basic form of PID controller is:

[0101] ;

[0102] in, is the control input, is the error, , and They are proportional, integral and differential coefficients respectively. These coefficients are dynamically adjusted through fuzzy logic to achieve precise control of thickness and obtain the prepared high-quality heating wire.

[0103] In a specific embodiment, the process of executing step 600 may specifically include the following steps:

[0104] The prepared heating wire is subjected to tensile strength and elongation at break tests to obtain mechanical property data, and the heating wire is subjected to resistance temperature coefficient test to obtain electric heating performance data;

[0105] Conduct hot and cold cycle tests on the heating wire to obtain power stability data, and construct a multidimensional feature vector based on the mechanical performance data, electrothermal performance data, and power stability data;

[0106] Performing feature dimensionality reduction on the multi-dimensional feature vector to obtain feature data after dimensionality reduction, and inputting the feature data after dimensionality reduction into a preset isolation forest model for anomaly detection to obtain an abnormal sample set;

[0107] Association rule mining is performed based on the production parameters corresponding to the abnormal sample set to obtain the abnormal cause analysis results, and Bayesian network analysis is performed based on the abnormal cause analysis results to obtain the process parameter optimization strategy;

[0108] Generate target production process parameters based on process parameter optimization strategy.

[0109] Specifically, the prepared heating wire is subjected to tensile strength and elongation at break tests to obtain mechanical performance data. The tensile strength test measures the maximum stress that the material can withstand before breaking by gradually increasing the pulling force until the wire breaks. This data can reflect the strength and durability of the wire. The elongation at break test measures the ductility of the wire before breaking, that is, the ratio of the length change when the wire is stretched to break to the original length, which can indicate the toughness of the wire. The heating wire is subjected to a resistance temperature coefficient test to obtain its electrothermal performance data. The resistance temperature coefficient refers to the sensitivity of the material's resistance to temperature changes. The resistance temperature coefficient is usually determined by heating the wire to different temperatures, measuring its resistance values ​​at different temperatures, and calculating the resistance change rate. This data can help determine whether the wire can work stably in the actual working environment, avoid resistance instability caused by temperature changes, and thus affect the conductive performance of the cable. The heating wire is subjected to a hot and cold cycle test to obtain its power stability data. The hot and cold cycle test simulates the working conditions of the wire under extreme temperature changes, and observes the stability of its power output through repeated heating and cooling. The power stability data can reveal the reliability of the wire in long-term use, that is, its ability to maintain stable power output under different temperature conditions. Through the mechanical performance data, electrothermal performance data and power stability data, a multidimensional feature vector is constructed to fully characterize the performance of the heating wire. The multidimensional feature vector can be expressed as:

[0110] ;

[0111] in, Represents tensile strength, represents the elongation at break, represents the temperature coefficient of resistance, Indicates power stability. Perform feature dimensionality reduction on the multidimensional feature vector, and retain the features that best reflect the changes in the original data by projecting the high-dimensional data into a low-dimensional space. Input the reduced feature data into the preset isolation forest model for anomaly detection. The isolation forest model is an unsupervised learning method that gradually isolates data points by randomly selecting features and segmentation points, thereby identifying abnormal samples. The anomaly detection process of the isolation forest can be described as:

[0112] ;

[0113] in, It is a sample The anomaly score, It is a sample The expected value of the path length in the isolation tree, is a constant used to normalize path lengths, is the number of samples. The higher the score, the more likely the sample is abnormal. In this way, an abnormal sample set is obtained, that is, those heated wire samples that perform abnormally in the performance test. Association rule mining is performed based on the production parameters corresponding to the abnormal sample set to identify the cause of the abnormality. Association rule mining discovers process conditions that may cause abnormalities by analyzing the frequent co-occurrence relationship between production parameters and abnormal samples. For example, if the same high extrusion temperature is found in multiple abnormal samples, then this temperature may be the key factor causing the abnormality. The results of association rule mining are usually expressed in the form of "if-then" rules, for example: "If the extrusion temperature exceeds 200°C, then the elongation at break is less than 10%." Based on the results of association rule mining, Bayesian network analysis is performed. Bayesian network is a probabilistic model used to represent conditional dependencies between variables. It can infer the probability path of abnormal occurrence based on observed anomalies and process parameters, thereby revealing potential process problems. The reasoning process of Bayesian network usually involves calculating conditional probabilities:

[0114] ;

[0115] in, Indicates that under known conditions Next event The probability of occurrence. Through reasoning, determine which production parameters are most likely to cause performance anomalies, and propose optimization strategies accordingly. Generate target production process parameters based on the process parameter optimization strategy obtained by Bayesian network analysis. The optimization parameters aim to adjust key links in the production process, such as extrusion temperature, cooling rate or material ratio, to eliminate factors that may cause anomalies and improve the overall quality and consistency of the heating wire.

[0116] In a specific embodiment, the execution step performs feature dimension reduction on the multidimensional feature vector to obtain feature data after dimension reduction, and inputs the feature data after dimension reduction into a preset isolation forest model for anomaly detection. The process of obtaining the abnormal sample set can specifically include the following steps:

[0117] Calculate the principal components of the multidimensional feature vector to obtain the principal component contribution rate, and select the principal component whose cumulative contribution rate is greater than a preset threshold according to the principal component contribution rate to obtain the feature data after dimensionality reduction;

[0118] The feature data after dimension reduction is standardized to obtain standardized feature data, and cluster analysis is performed on the standardized feature data to obtain initial cluster centers;

[0119] The local outlier factor of the standardized feature data is calculated according to the initial cluster center to obtain the sample abnormality, and the abnormal sample candidate set is constructed according to the sample abnormality;

[0120] Perform support vector data description analysis on the abnormal sample candidate set to obtain the abnormal boundary, and screen the abnormal sample candidate set according to the abnormal boundary to obtain preliminary abnormal samples;

[0121] An isolation tree is constructed based on the preliminary abnormal samples through the isolation forest model, and the isolation tree is randomly split to obtain multiple subtrees. The path lengths of the multiple subtrees are calculated to obtain the abnormality score;

[0122] The preliminary abnormal samples are sorted according to the abnormal scores to obtain the abnormal sample sorting results, and the dynamic threshold is set based on the abnormal sample sorting results to obtain the abnormal sample set.

[0123] Specifically, principal component analysis is performed on the multidimensional feature vector to extract the most important features. Principal component analysis calculates the covariance matrix of the data and decomposes its eigenvalues ​​to find the principal components. These principal components are linear combinations of the original features and can explain most of the variance in the data. For each principal component, its contribution rate is calculated, that is, the proportion of the total variance that the principal component can explain. The principal component contribution rate can be expressed as:

[0124] ;

[0125] in, It is The eigenvalues ​​of the principal components, is the total number of features. By accumulating the principal component contribution rate, it is determined how many principal components are sufficient to retain most of the information in the data. According to a preset threshold, such as 90%, the principal components with a cumulative contribution rate greater than the threshold are selected to form the feature data after dimensionality reduction. This step can effectively reduce the dimension of the data and reduce the computational complexity while retaining the main information of the data. The feature data after dimensionality reduction is standardized and converted into a standard normal distribution with a mean of 0 and a variance of 1 to eliminate the dimensional differences between different features so that each feature has equal influence in subsequent analysis. The standardized feature data will be used for cluster analysis to identify natural groups in the data. Cluster analysis uses algorithms such as K-means to divide data points into multiple clusters based on the similarity of features and determine the initial cluster center of each cluster. The cluster center represents the center position of each cluster. The local outlier factor of the standardized feature data is calculated based on the initial cluster center, and the abnormality of the data point is evaluated by comparing the local density difference between each data point and its neighbors. The higher the local outlier factor value, the more likely the data point is to be abnormal in its neighborhood. The calculation formula of the local outlier factor is:

[0126] ;

[0127] in, represents a data point, yes of Neighbors, is the local reachable density of the point. According to the calculated LOF value, the samples are sorted and a candidate set of abnormal samples is constructed, including those samples with higher LOF values. Support vector data description analysis is performed on the candidate set of abnormal samples. Support vector data description is a technology used for anomaly detection. It constructs a minimized hypersphere to surround normal data points and exclude abnormal samples. The goal of support vector data description is to find an optimal sphere that can surround as many normal samples as possible while excluding those abnormal samples outside the boundary. The optimization goal of support vector data description can be expressed as:

[0128] ;

[0129] in, is the radius of the sphere, is the penalty parameter, is a slack variable used to tolerate some samples being excluded from the sphere. By solving this optimization problem, the boundary of the abnormal samples is determined. According to the boundary, the candidate set of abnormal samples is screened to obtain preliminary abnormal samples. The preliminary abnormal samples are analyzed using the isolation forest model. The isolation forest model gradually splits the samples by constructing multiple randomly split trees (i.e., isolation trees) until the samples are isolated. The shorter the path length in the isolation tree, the easier it is for the sample to be isolated, and therefore the more likely it is to be an abnormal sample. The anomaly score is calculated based on the average path length of the sample in the isolation tree:

[0130] ;

[0131] in, is a constant used to normalize path lengths, is the number of samples. By calculating the anomaly scores of all samples, the preliminary anomaly samples are sorted to identify those samples that are most likely to be anomalies. The preliminary anomaly samples are sorted according to the anomaly scores to obtain the sorting results of the anomaly samples. A dynamic threshold is set based on these scores. The setting of the dynamic threshold can be adjusted according to actual business needs and the distribution of anomaly scores to ensure the accuracy and recall rate of anomaly detection. For example, if the anomaly scores show a clear bimodal distribution, the threshold can be set at the valley between the two peaks to effectively distinguish normal samples from abnormal samples.

[0132] In a specific embodiment, the execution step performs association rule mining according to the production parameters corresponding to the abnormal sample set to obtain the abnormal cause analysis result, and performs Bayesian network analysis based on the abnormal cause analysis result to obtain the process parameter optimization strategy. The process may specifically include the following steps:

[0133] Perform feature encoding on the abnormal sample set to obtain encoded abnormal sample data, and perform frequent item set mining on the encoded abnormal sample data to obtain frequent item sets;

[0134] Construct association rule candidate sets based on frequent item sets, and calculate support and confidence for the association rule candidate sets to obtain preliminary association rules;

[0135] Perform lift analysis on the preliminary association rules to obtain the association rule evaluation results, and screen the rules based on the association rule evaluation results to obtain effective association rules;

[0136] Construct a causal graph based on effective association rules, and perform topological sorting on the causal graph to obtain the abnormal cause hierarchy;

[0137] Markov blanket identification is performed according to the abnormal cause hierarchy to obtain local dependencies, and a Bayesian network structure is constructed based on the local dependencies. A process parameter optimization strategy is constructed based on the Bayesian network structure.

[0138] Specifically, feature encoding is performed on the abnormal sample set, and the original data is converted into a coding form suitable for mining. For example, if an abnormal sample has multiple attributes, such as temperature, pressure, time, etc., the different values ​​of these attributes are mapped to unique coding values ​​to form the encoded abnormal sample data. Frequent item set mining is performed on the encoded abnormal sample data to identify attribute combinations that often appear at the same time in the abnormal sample set. These combinations may represent potential association patterns. Frequent item sets can be mined using the Apriori algorithm, which recursively detects the frequency of each item set and screens out item sets whose support (i.e., the frequency of the item set appearing in the sample set) is greater than the set threshold. The support formula for frequent item sets is:

[0139] ;

[0140] in, is an item set. The higher the support, the more common the item set is in the abnormal sample set. The candidate set of association rules is constructed based on the frequent item set. The association rule refers to an association relationship between the attributes in a frequent item set. The candidate set of association rules is generated by combining the various parts of the frequent item set. The candidate rules are screened by calculating the support and confidence. The confidence refers to the probability that the result will occur when the known premise occurs. The calculation formula is:

[0141] ;

[0142] in, represents an association rule, and support represents the item set and The frequency of co-occurrence. By calculating the confidence, the association rules with higher certainty are selected, namely the preliminary association rules. The preliminary association rules are subjected to lift analysis to evaluate the strength of the association rules. The lift is an important indicator to measure the strength of the association rules, which indicates the degree of increase in the probability of the result occurring relative to its natural probability when considering the association relationship. The lift formula is:

[0143] ;

[0144] A lift greater than 1 indicates that the rule has a positive association, less than 1 indicates a negative association, and equal to 1 indicates no association. Through lift analysis, effective association rules are screened out, and those rules with high confidence but no practical significance are eliminated. A causal graph is constructed based on the screened effective association rules. A causal graph is a graph structure used to represent the causal relationship between events. Nodes represent events or attributes, and edges represent causal relationships. When constructing a causal graph, the premise and result in the effective association rule are mapped to nodes in the graph respectively, and directed edges are constructed in the direction of the association rule to form a preliminary causal structure. The causal graph is topologically sorted to obtain the hierarchical structure of abnormal causes. Topological sorting is an algorithm that sorts the nodes of a directed acyclic graph according to dependencies. The result can clearly show which causes are upstream of other causes, that is, more fundamental causes, and which are downstream results. Through topological sorting, the hierarchy of abnormal phenomena is intuitively presented. Markov blanket recognition is performed on the hierarchy of abnormal causes. Markov blanket refers to the set of direct parent nodes, child nodes and common parent nodes of a variable in a Bayesian network, which constitute the local dependency of the variable. By identifying the Markov blanket, we can determine which variables have a direct impact on a certain abnormal cause, simplify the causal graph, and retain the most critical dependencies. Based on the identified local dependencies, a Bayesian network structure is constructed. A Bayesian network is a directed acyclic graph model that represents conditional independence and dependencies between variables. It can analyze the mutual influence of various production parameters through probabilistic reasoning and identify the most likely causes of abnormalities. The basic reasoning formula of the Bayesian network is:

[0145] ;

[0146] in, Indicates that when the parent node is known, the node The conditional probability of each production parameter is obtained through probabilistic reasoning, and finally the weak links in the process parameters are identified. According to the reasoning results constructed by the Bayesian network structure, a process parameter optimization strategy is formulated. The optimization strategy usually includes adjusting key parameters in the production process, such as temperature, pressure, speed, etc., to reduce the probability of abnormality. Through optimization, the production process can more stably output products that meet quality standards and reduce the production of defective products.

[0147] The above describes the safety warning method for cable production in the embodiment of the present application. The following describes the safety warning system 10 for cable production in the embodiment of the present application. Figure 2 In the embodiment of the present application, an embodiment of the safety warning system 10 for cable production includes:

[0148] The analysis module 11 is used to perform multivariate regression analysis and formula optimization on the raw material ratio data of cable production to obtain the formula data of each component;

[0149] A calculation module 12, for performing Fourier transformation and blending parameter calculation on the mixing process data of each component based on the formula data to obtain the optimal blending parameters;

[0150] The processing module 13 is used to perform material mixing processing on each component based on the optimal blending parameters to obtain heating element particles, and to perform particle size analysis on the heating element particles to obtain particle size distribution data;

[0151] An optimization module 14 is used to analyze the influencing factors and optimize the extrusion parameters of the temperature and pressure data of the extrusion molding process based on the particle size distribution data to obtain the target extrusion parameters;

[0152] A control module 15 is used to prepare a multi-layer structure by using target extrusion parameters, and to perform multi-scale feature extraction and multi-layer preparation speed control on thickness data of the preparation process to obtain a prepared heating wire;

[0153] The detection module 16 is used to perform a performance test on the prepared heating wire to obtain performance test data, perform anomaly detection on the performance test data and optimize production process parameters to generate target production process parameters.

[0154] Through the synergy of the above components, the optimal ratio of each component can be determined more accurately by multivariate regression analysis and formula optimization of raw material ratio data, the utilization efficiency of raw materials can be improved, and production costs can be reduced. The Fourier transform and blending parameter calculation methods can quantitatively evaluate the uniformity of the mixing process, optimize the blending parameters, and improve the quality and stability of material mixing. Combined with particle size analysis and extrusion parameter optimization, the particle size distribution of the heating element particles can be better controlled, and the stability of the extrusion molding process and the consistency of product quality can be improved. Through multi-scale feature extraction and multi-layer preparation speed control, the precise control of the multi-layer structure preparation process is achieved, and the structural uniformity and performance stability of the product are improved. The use of anomaly detection and production process parameter optimization methods can timely detect abnormal situations in the production process, and give targeted optimization strategies to improve production efficiency and product qualification rate. Through the combination of data-driven and intelligent algorithms, intelligent monitoring and early warning of the entire cable production process are realized, which significantly improves the safety and controllability of the production process.

[0155] The present application also provides an electronic device, which includes a memory and a processor, wherein the memory stores computer-readable instructions, and when the computer-readable instructions are executed by the processor, the processor executes the steps of the safety warning method for cable production in the above-mentioned embodiments.

[0156] The present application also provides a computer-readable storage medium, which may be a non-volatile computer-readable storage medium or a volatile computer-readable storage medium. Instructions are stored in the computer-readable storage medium. When the instructions are executed on a computer, the computer executes the steps of the safety warning method for cable production.

[0157] Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working processes of the above-described systems, systems and units can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.

[0158] If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present application is essentially or the part that contributes to the prior art or all or part of the technical solution can be embodied in the form of a software product. The computer software product is stored in a storage medium, including several instructions to enable an electronic device (which can be a personal computer, server, or network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present application. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM), random access memory (RAM), disk or optical disk, etc., various media that can store program codes.

[0159] As described above, the above embodiments are only used to illustrate the technical solutions of the present application, rather than to limit them. Although the present application has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present application.

Claims

1. A safety early warning method for cable production, characterized in that: The safety early warning method for cable production includes: Perform multivariate regression analysis and formula optimization on the raw material ratio data of cable production to obtain the formula data of each component; Based on the formula data, the mixing process data of each component is subjected to Fourier transformation and blending parameter calculation to obtain the optimal blending parameters; specifically comprising: based on the formula data, the mixing process data of each component is subjected to time window segmentation to obtain time series segments, and the time series segments are subjected to fast Fourier transformation to obtain frequency domain feature data; according to the frequency domain feature data, the power spectrum density is calculated to obtain the frequency distribution characteristics of the mixing process, and the frequency distribution characteristics are subjected to peak detection to obtain the target oscillation frequency; according to the target oscillation frequency, the mixing process data is filtered to obtain the filtered signal, and the filtered signal is subjected to autocorrelation analysis to obtain the quantitative index of mixing uniformity; based on the quantitative index, a second mapping relationship between mixing uniformity and blending parameters is constructed, and the second mapping relationship is subjected to blending parameter analysis to obtain the optimal blending parameters; Performing material mixing processing on each component based on the optimal blending parameters to obtain heating element particles, and performing particle size analysis on the heating element particles to obtain particle size distribution data; Based on the particle size distribution data, the temperature and pressure data of the extrusion molding process are subjected to influencing factor analysis and extrusion parameter optimization to obtain target extrusion parameters; specifically comprising: extracting distribution characteristics of the particle size distribution data to obtain particle size distribution characteristics, and performing stratified sampling on the temperature and pressure data of the extrusion molding process according to the particle size distribution characteristics to obtain a target sample set; constructing a third mapping relationship between temperature, pressure and extrusion parameters according to the target sample set, and performing sensitivity analysis on the third mapping relationship to obtain weight coefficients of various influencing factors; regularizing the third mapping relationship according to the weight coefficient to obtain a regularized regression model, and constructing a fourth mapping relationship between extrusion parameters and product quality based on the regularized regression model; performing extrusion parameter analysis on the fourth mapping relationship to obtain initially optimized extrusion parameters, and performing optimization solution based on the initially optimized extrusion parameters to obtain target extrusion parameters; A multi-layer structure is prepared by using the target extrusion parameters, and multi-scale feature extraction and multi-layer preparation speed control are performed on thickness data of the preparation process to obtain a prepared heating wire; The prepared heating wire is subjected to a performance test to obtain performance test data, and the performance test data is subjected to abnormality detection and production process parameter optimization to generate target production process parameters.

2. The safety early warning method for cable production according to claim 1 is characterized in that: The raw material ratio data for cable production is subjected to multivariate regression analysis and formula optimization to obtain the formula data of each component, including: Extract influencing factors from raw material ratio data of cable production to obtain target influencing factors, and perform multivariate linear regression analysis on the raw material ratio data according to the target influencing factors to obtain an initial regression model; Performing residual analysis on the initial regression model to obtain residual distribution characteristics, and optimizing the initial regression model using weighted least squares method according to the residual distribution characteristics to obtain an optimized regression model; Constructing a first mapping relationship between raw material ratio and product performance based on the optimized regression model, and performing a ratio scheme analysis on the first mapping relationship to obtain an optimal ratio scheme; Generating theoretical formula data of each component according to the optimal ratio scheme, and performing error distribution calculation on the theoretical formula data to obtain formula error distribution; The theoretical formula data is discretized according to the formula error distribution to obtain formula data of each component.

3. The safety early warning method for cable production according to claim 1 is characterized in that: The material mixing process is performed on each component based on the optimal blending parameters to obtain heating element particles, and the particle size analysis is performed on the heating element particles to obtain particle size distribution data, including: Weighing each component according to the optimal blending parameters to obtain a raw material to be mixed, and mixing the raw materials to be mixed to obtain an initial mixture; Extruding the initial mixture through a twin-screw extruder to obtain a uniformly mixed composite material, and cooling and granulating the uniformly mixed composite material to obtain initial raw material particles; Performing laser particle size analysis on the initial original particles to obtain initial distribution data, and performing data analysis based on the initial distribution data to obtain characteristic parameters of particle size distribution; Constructing a probability density function of particle size distribution based on the characteristic parameters, and iterating the probability density function using an expectation maximization algorithm to obtain an optimized particle size distribution model; The initial original particles are screened according to the optimized particle size distribution model to obtain heating original particles within a target particle size range, and image analysis is performed on the heating original particles within the target particle size range to obtain particle size distribution data.

4. The safety early warning method for cable production according to claim 1 is characterized in that: The method comprises: preparing a multilayer structure by using the target extrusion parameters, extracting multi-scale features of thickness data of the preparation process and controlling the multilayer preparation speed to obtain a prepared heating wire, comprising: Heating and coating the tinned soft copper stranded conductor according to the target extrusion parameters to obtain an initial coating structure, and performing X-ray real-time scanning on the initial coating structure to obtain initial thickness distribution data; Performing wavelet transform on the initial thickness distribution data to obtain multi-scale feature information, and performing adaptive Kalman filter analysis based on the multi-scale feature information to generate a thickness prediction model; Based on the thickness prediction model, the extrusion speed is adjusted in real time to obtain an optimized coating structure, and a polyolefin insulation layer is extruded on the optimized coating structure to obtain an insulation layer structure; The insulating layer structure is braided with tinned copper wires to obtain a shielding layer structure, and a fluorine-containing polymer sheath is extruded on the shielding layer structure to obtain a multi-layer structure semi-finished product; The thickness of the multi-layer semi-finished product is measured online to obtain real-time thickness data, and fuzzy PID control optimization is performed according to the real-time thickness data to obtain the prepared heating wire.

5. The safety early warning method for cable production according to claim 1, characterized in that: The method of performing a performance test on the prepared heating wire to obtain performance test data, and performing abnormality detection and production process parameter optimization on the performance test data to generate target production process parameters includes: The prepared heating wire is subjected to a tensile strength and elongation at break test to obtain mechanical property data, and the heating wire is subjected to a resistance temperature coefficient test to obtain electrothermal property data; Performing a hot and cold cycle test on the heating wire to obtain power stability data, and constructing a multidimensional feature vector according to the mechanical property data, the electrothermal property data and the power stability data; Performing feature dimensionality reduction on the multidimensional feature vector to obtain feature data after dimensionality reduction, and inputting the feature data after dimensionality reduction into a preset isolation forest model for anomaly detection to obtain an abnormal sample set; Performing association rule mining according to the production parameters corresponding to the abnormal sample set to obtain abnormal cause analysis results, and performing Bayesian network analysis based on the abnormal cause analysis results to obtain a process parameter optimization strategy; Generate target production process parameters according to the process parameter optimization strategy.

6. The safety early warning method for cable production according to claim 5 is characterized in that: The multi-dimensional feature vector is subjected to feature dimensionality reduction to obtain feature data after dimensionality reduction, and the feature data after dimensionality reduction is input into a preset isolation forest model for anomaly detection to obtain an abnormal sample set, including: Performing principal component calculation on the multidimensional feature vector to obtain principal component contribution rate, and selecting principal components whose cumulative contribution rate is greater than a preset threshold value according to the principal component contribution rate to obtain feature data after dimensionality reduction; Performing standardization processing on the feature data after dimension reduction to obtain standardized feature data, and performing cluster analysis on the standardized feature data to obtain initial cluster centers; Calculating the local outlier factor of the standardized feature data according to the initial cluster center to obtain the sample abnormality, and constructing an abnormal sample candidate set according to the sample abnormality; Performing support vector data description analysis on the abnormal sample candidate set to obtain an abnormal boundary, and screening the abnormal sample candidate set according to the abnormal boundary to obtain preliminary abnormal samples; An isolation tree is constructed based on the preliminary abnormal sample by using an isolation forest model, and the isolation tree is randomly split to obtain multiple subtrees, and path lengths of the multiple subtrees are calculated to obtain an abnormality score; The preliminary abnormal samples are sorted according to the abnormal scores to obtain an abnormal sample sorting result, and a dynamic threshold is set based on the abnormal sample sorting result to obtain an abnormal sample set.

7. The safety early warning method for cable production according to claim 5, characterized in that: The association rule mining is performed according to the production parameters corresponding to the abnormal sample set to obtain the abnormal cause analysis result, and the Bayesian network analysis is performed based on the abnormal cause analysis result to obtain the process parameter optimization strategy, including: Performing feature encoding on the abnormal sample set to obtain encoded abnormal sample data, and performing frequent item set mining on the encoded abnormal sample data to obtain frequent item sets; Constructing a candidate set of association rules according to the frequent item sets, and calculating support and confidence of the candidate set of association rules to obtain preliminary association rules; Performing a lift analysis on the preliminary association rules to obtain an association rule evaluation result, and performing rule screening according to the association rule evaluation result to obtain an effective association rule; Constructing a causal graph based on the effective association rules, and topologically sorting the causal graph to obtain an abnormal cause hierarchy; Markov blanket identification is performed according to the abnormal cause hierarchy to obtain local dependencies, and a Bayesian network structure is constructed based on the local dependencies, and a process parameter optimization strategy is constructed based on the Bayesian network structure.

8. A safety early warning system for cable production, characterized in that: Used to execute the safety early warning method for cable production according to any one of claims 1 to 7, the safety early warning system for cable production comprises: The analysis module is used to perform multivariate regression analysis and formula optimization on the raw material ratio data of cable production to obtain the formula data of each component; A calculation module, used for performing Fourier transform and blending parameter calculation on the mixing process data of each component based on the formula data to obtain the optimal blending parameters; A processing module, used for performing material mixing processing on each component based on the optimal blending parameters to obtain heating element particles, and performing particle size analysis on the heating element particles to obtain particle size distribution data; An optimization module, for performing influencing factor analysis and extrusion parameter optimization on the temperature and pressure data of the extrusion molding process based on the particle size distribution data, to obtain target extrusion parameters; A control module, used to prepare a multi-layer structure by using the target extrusion parameters, and to perform multi-scale feature extraction and multi-layer preparation speed control on thickness data of the preparation process, so as to obtain a prepared heating wire; The detection module is used to perform a performance test on the prepared heating wire to obtain performance test data, and perform abnormality detection and production process parameter optimization on the performance test data to generate target production process parameters.

Images