Production quality monitoring method and system for ultrahigh-strength hemp core steel wire

Abstract

The invention relates to the technical field of production quality monitoring, and discloses a production quality monitoring method and system for an ultrahigh-strength hemp core steel wire. The method comprises the steps of collecting production data; respectively extracting a first sub-region and a second sub-region, combining the first sub-region and the second sub-region into a combined region, and setting the combined region corresponding to the minimum segmentation error as an optimal region; inputting the optimal region into a quality prediction model, and obtaining a prediction error of a quality prediction value based on an actual quality value; judging whether the quality prediction model is converged or not, and if not, dividing the production variables and the product variables again until the quality prediction model is converged; and respectively calculating the difference values between the quality prediction values and the target quality indexes, if the difference values are smaller than or equal to the first type of threshold values, judging that the hemp core steel wire is qualified in production, otherwise, adjusting the production variables. The real-time performance and the accuracy of production quality monitoring of the ultra-high-strength hemp core steel wire are improved.

Description

Technical Field

[0001] The present application relates to the technical field of production quality monitoring, and in particular to a production quality monitoring method and system for ultra-high strength hemp core steel wire. Background Art

[0002] Ultra-high-strength hemp-core steel wire is a composite material with excellent properties such as high strength, high toughness, and corrosion resistance. It is widely used in aerospace, bridge construction, marine engineering, building structures, the automotive industry, and high-end machinery and equipment. Although the production process of ultra-high-strength hemp-core steel wire is relatively mature, some quality control challenges and limitations still exist in the actual production process. For example, multi-step quality control is difficult, real-time monitoring is insufficient, data integration and analysis are difficult, and human factors exist during manual inspection and operation. Therefore, there is a need to provide an efficient and reliable method for monitoring the production quality of ultra-high-strength hemp-core steel wire.

[0003] Similar prior art includes Chinese patent application CN117235648A, which discloses a data-processing-based integrated management system for the entire steel wire processing process. The system includes a data acquisition module, a feature data acquisition module, a partitioning threshold acquisition module, and an abnormal data screening module. The module calculates the uniformity of each data sequence in the processing equipment data, obtains a feature data sequence based on the uniformity, calculates the degree of outliers for each data in each feature data sequence, obtains the minimum and maximum clusters for each feature data sequence based on the degree of outliers, calculates the degree of fluctuation of the minimum and maximum clusters, obtains all partitioning thresholds for each feature data sequence based on the degree of fluctuation, obtains an isolation forest based on all partitioning thresholds, obtains abnormal data based on the isolation forest, and issues equipment abnormality warnings based on the number of abnormal data. This invention improves the accuracy of data anomaly score calculation, thereby improving the efficiency and accuracy of abnormal data detection. A Chinese patent application, publication number CN118735718A, discloses a method, apparatus, and equipment for producing silver-plated copper-clad steel wire. The method comprises: obtaining order data for the silver-plated copper-clad steel wire; analyzing and processing the implementation data to obtain ideal data; analyzing and processing the demand data to obtain actual data; performing data analysis based on the ideal and actual data to obtain production data; and controlling production equipment to produce the silver-plated copper-clad steel wire based on the production data. The silver-plated copper-clad steel wire production method provided in this application can address defects and deficiencies in the silver-plated copper-clad steel wire, effectively reducing the gradual decline in product quality produced by the production equipment, and improving the stability and reliability of the silver-plated copper-clad steel wire.

[0004] The shortcomings of existing technologies are mainly reflected in the limited data processing and equipment anomaly detection during the steel wire processing process, the lack of analysis of the multi-dimensional key quality indicators of the steel wire, and the lack of comprehensiveness and real-time quality monitoring of the silver-plated copper-clad steel wire. In actual situations, it is necessary to build a comprehensive quality monitoring system, conduct intelligent analysis of production data, and improve the real-time monitoring capabilities of production quality. Summary of the Invention

[0005] The present application provides a method and system for monitoring the production quality of ultra-high strength hemp core steel wire, which improves the real-time and accuracy of the production quality monitoring of ultra-high strength hemp core steel wire.

[0006] In a first aspect, the present application provides a method for monitoring the production quality of ultra-high-strength hemp core steel wire, the method comprising:

[0007] S1: Setting target quality indicators for hemp core steel wire production, and setting data acquisition equipment based on the production process to collect production data, wherein the production data includes production variables and product variables;

[0008] S2: Divide the production variable into a plurality of first sub-regions, and divide the product variable into a plurality of second sub-regions, extract one first sub-region and one second sub-region respectively and combine them into a combined region, calculate the segmentation error of any of the combined regions based on a normal distribution function, and set the combined region corresponding to the minimum segmentation error as the optimal region;

[0009] S3: constructing a quality prediction model, inputting the optimal region into the quality prediction model, outputting a quality prediction value, obtaining an actual quality value corresponding to the optimal region, and obtaining a prediction error of the quality prediction value based on the actual quality value;

[0010] S4: judging whether the quality prediction model has converged based on the prediction error; if not, proceeding to step S2 to re-divide the production variables and the product variables until the quality prediction model converges;

[0011] S5: Calculate the difference between the quality prediction value and the target quality index based on the preset target category. If the difference is less than or equal to the first category threshold, determine that the hemp core steel wire production is qualified. Otherwise, adjust the production variable based on the preset target category.

[0012] In combination with the first aspect, obtaining the operation correlation between any two monitoring devices based on the breeding plan, dividing the production variables into a plurality of first sub-areas includes:

[0013] Obtaining correlations between the production variables, dividing all the production variables into first-category variables and second-category variables based on the correlations, and establishing a first spatial coordinate system, wherein the horizontal axis of the first spatial coordinate system represents the first-category variables and the vertical axis of the first spatial coordinate system represents the second-category variables;

[0014] A clustering algorithm is used to respectively obtain the boundary values of the first category variables and the second category variables, the boundary values are set as division points, and the first spatial coordinates are divided based on the division points in a direction parallel to the horizontal axis or the vertical axis to generate multiple first sub-areas.

[0015] In combination with the first aspect, dividing the product variable into a plurality of second sub-areas includes:

[0016] Obtain relevant variable values of any of the product variables, generate random numbers for all the relevant variable values, set multiple division thresholds, group all the random numbers based on the division thresholds, generate multiple random arrays, and set the product variables contained in the random arrays as the second sub-area.

[0017] In combination with the first aspect, the construction quality prediction model includes:

[0018] Setting the production data of the same production batch as a first data set, and setting all the production data as a second data set;

[0019] Extracting length data of the hemp core steel wire from any of the first data sets, setting the length data in a row direction, setting other data of the first data set in a column direction, and generating a two-dimensional matrix of the first data set based on the row direction and the column direction;

[0020] Building a first prediction model based on a convolutional neural network, inputting all the two-dimensional matrices into the first prediction model for training, and adjusting model parameters;

[0021] constructing a second prediction model based on an exponential smoothing method, inputting the second data set into the second prediction model, and calculating an exponential smoothing coefficient of the second prediction model;

[0022] The trained first prediction model and the calculated second prediction model are combined and set as the quality prediction model.

[0023] In conjunction with the first aspect, the output quality prediction value includes:

[0024] The quality prediction model obtains the production data contained in the optimal area and sets it as data to be analyzed, and inputs the data to be analyzed into the first prediction model and the second prediction model after data preprocessing;

[0025] The first prediction model outputs a first performance prediction result of the hemp-core steel wire, and the second prediction model outputs a smoothing result and a proportional coefficient of the hemp-core steel wire based on the production time of the data to be analyzed. If the deviation of the data to be analyzed within the production time is greater than a first preset value, the product of the smoothing result and the proportional coefficient is set as the second performance prediction result; otherwise, the sum of the smoothing result and the proportional coefficient is set as the second performance prediction result;

[0026] The first performance prediction result and the second performance prediction result are weighted and summarized based on the prediction data category to generate the quality prediction value.

[0027] In combination with the first aspect, after dividing the production variable into a plurality of first sub-regions, re-dividing the production variable includes:

[0028] Acquire data points included in the first sub-region, construct a first function based on the covariance matrix, and calculate a first function value of the data point based on the first function;

[0029] Calculating a regional prediction value of the data point based on the first function value, and calculating a regional error of the first sub-region according to the regional prediction value using a weighted average method;

[0030] Repeat this step to calculate the regional errors of all the first sub-regions, and set the first sub-region corresponding to the largest regional error as the region to be divided;

[0031] New first spatial coordinates are constructed for the area to be divided, and the new first spatial coordinates are divided.

[0032] In combination with the first aspect, determining whether the quality prediction model has converged includes:

[0033] Extracting a first subset from the quality prediction values, calculating an error mean and an error standard deviation of the first subset based on the prediction error, setting a fluctuation coefficient, multiplying the fluctuation coefficient by the error standard deviation and then adding the result to the error mean to generate a second preset value;

[0034] If the prediction error is less than or equal to the second preset value, it is determined that the quality prediction model has converged; otherwise, the quality prediction model has not converged.

[0035] In combination with the first aspect, adjusting the production variable based on the preset target category includes:

[0036] extracting the preset target category whose difference is greater than the first category threshold and setting it as the category to be adjusted, obtaining historical fluctuation data of the production variable in any of the categories to be adjusted, calculating the correlation between any two groups of the historical fluctuation data, and combining the historical fluctuation data with a correlation greater than a third preset value into a correlated adjustment group;

[0037] The fluctuation ratio mean of the associated adjustment group is calculated, and the numerical value of the production variable is adjusted based on the fluctuation ratio mean.

[0038] In conjunction with the first aspect, the collecting of production data includes:

[0039] Using spectral analysis to obtain the chemical composition of the hemp core material, performing physical testing on the hemp core material to obtain physical property data, the data acquisition device obtains drawing parameters, using a high-resolution camera to perform real-time imaging of the surface of the hemp core steel wire to generate surface quality data, and using an ultrasonic detector to test the hemp core steel wire to output tensile strength and elongation;

[0040] The chemical composition, the physical property data, the drawing parameters, the surface quality data, the tensile strength and the elongation are summarized based on the production process and are all set as the production data.

[0041] In a second aspect, the present application provides a production quality monitoring system for ultra-high strength hemp core steel wire, the production quality monitoring system for ultra-high strength hemp core steel wire comprising:

[0042] An acquisition module is used to set target quality indicators for hemp core steel wire production and to set data acquisition equipment based on the production process to collect production data, wherein the production data includes production variables and product variables;

[0043] a combining module, configured to divide the production variable into a plurality of first sub-regions, divide the product variable into a plurality of second sub-regions, extract one of the first sub-regions and one of the second sub-regions respectively and combine them into a combined region, calculate a segmentation error of any of the combined regions based on a normal distribution function, and set the combined region corresponding to the minimum segmentation error as the optimal region;

[0044] a prediction module, configured to construct a quality prediction model, input the optimal region into the quality prediction model, output a quality prediction value, obtain an actual quality value corresponding to the optimal region, and obtain a prediction error of the quality prediction value based on the actual quality value;

[0045] a verification module, configured to determine whether the quality prediction model has converged based on the prediction error, and if not, re-divide the production variables and the product variables until the quality prediction model has converged;

[0046] The control module is used to calculate the difference between the quality prediction value and the target quality index according to the preset target category. If the difference is less than or equal to the first category threshold, the hemp core steel wire is judged to be qualified; otherwise, the production variable is adjusted based on the preset target category.

[0047] The technical solution provided in this application first collects production data from multiple sources, covering key parameters of each production link from raw materials to finished products, which can fully reflect the production status. Based on the results of correlation analysis, the production variables are divided into first-category variables and second-category variables. The spatial coordinates are divided using a clustering algorithm to generate multiple first sub-regions. The product variables are divided into multiple second sub-regions using a random number generation and threshold partitioning method. The data-driven regional partitioning method avoids the subjectivity of human division, improves the scientific nature and accuracy of regional division, can adapt to changes in production conditions, and adjust the regional division strategy in a timely manner to ensure the accuracy and effectiveness of quality control and improve quality control efficiency. Then, a convolutional neural network is used to model the data of the same production batch to capture the complex relationship between production data and product quality within the batch. The exponential smoothing method is used to model the data of all production batches to capture the overall production trend and long-term laws, and generate a final quality prediction model. The model can predict the product quality of different production batches and different product types, has strong versatility, and improves the stability and accuracy of the prediction results. Finally, based on the difference between the prediction error and the target value, the category of production variables that need to be adjusted is identified, the correlation between historical fluctuation data is analyzed, and the data with high correlation are combined into associated adjustment groups. Based on the mean of the fluctuation ratio, the production variables are adjusted, avoiding the impact of a single variable adjustment on other variables, improving the controllability of the adjustment effect, and continuously adjusting the production variables to keep them in the optimal state, which is conducive to improving the stability of the production process and the consistency of product quality. BRIEF DESCRIPTION OF THE DRAWINGS

[0048] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0049] Figure 1 This is a schematic diagram of an embodiment of a method for monitoring the production quality of ultra-high-strength hemp-core steel wire in an embodiment of the present application;

[0050] Figure 2 This is a schematic diagram of an embodiment of outputting quality prediction values by the quality prediction model in an embodiment of the present application;

[0051] Figure 3 This is a schematic diagram of an embodiment of a production variable adjustment strategy in an embodiment of the present application;

[0052] Figure 4 This is a schematic diagram of an embodiment of a production quality monitoring system for ultra-high strength hemp core steel wire in an embodiment of the present application. DETAILED DESCRIPTION

[0053] The embodiments of the present application provide a method and system for monitoring the production quality of ultra-high strength hemp core steel wire. The terms "first", "second", "third", "fourth", etc. (if any) in the specification and claims of this application and the above-mentioned drawings are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that the data used in this way can be interchangeable where appropriate, so that the embodiments described herein can be implemented in an order other than that illustrated or described herein. In addition, the terms "including" or "having" and any variations thereof are intended to cover non-exclusive inclusions, for example, a process, method, system, product or device that includes a series of steps or units is not necessarily limited to those steps or units that are clearly listed, but may include other steps or units that are not clearly listed or that are inherent to these processes, methods, products or devices.

[0054] For ease of understanding, the specific process of the embodiment of the present application is described below. Figure 1 In one embodiment of the present application, a method for monitoring the production quality of ultra-high strength hemp core steel wire includes:

[0055] Step S1: setting target quality indicators for hemp core steel wire production, and setting data acquisition equipment based on the production process to collect production data, where the production data includes production variables and product variables.

[0056] It is understandable that the execution subject of this application can be a production quality monitoring device for ultra-high strength hemp core steel wire, or a terminal or a server, which is not limited here. The embodiment of this application is described by taking the server as the execution subject as an example.

[0057] Specifically, the target quality index refers to the performance index value of the hemp core steel wire, such as tensile strength index, elongation index, diameter deviation index, surface quality index, etc. The production process refers to the process steps for producing hemp core steel wire, and the data acquisition equipment refers to various sensor devices. Production data refers to data associated with monitoring the production quality of ultra-high strength hemp core steel wire. Production variables refer to operational data and quality data related to product quality, such as process parameters (such as drawing speed, drawing force, temperature, lubricant concentration, etc.), equipment status parameters, raw material data, environmental parameters, product quality data, etc. Product variables refer to variables that mark the classification information, status or category of the production of hemp core steel wire, which can be converted into numerical data through unique hot encoding, such as the production batch number of hemp core steel wire, the product specifications of hemp core materials, the number of production equipment, the number of operators, etc.

[0058] Step S2: Divide the production variables into multiple first sub-regions, and divide the product variables into multiple second sub-regions. Extract one first sub-region and one second sub-region respectively and combine them into a combined region. Calculate the segmentation error of any combined region based on the normal distribution function, and set the combined region corresponding to the minimum segmentation error as the optimal region.

[0059] Specifically, dividing the production variables can capture the quality change patterns under different operating conditions. Each first sub-region corresponds to a different combination of production conditions. For example, a region with higher temperature and lower pressure may correspond to another quality change trend. Dividing the product variables can obtain different categories of product variables (for example, different batches of raw materials, different production lines) that will have different effects on product quality, where each second sub-region corresponds to a different category or state. For example, different batches of hemp core materials may lead to systematic differences in product quality. The data contained in the combined area can be used to achieve local modeling of production data, thereby improving the real-time and flexibility of subsequent quality monitoring. The combined area contains multiple data points. With the center of gravity of the combined area as the center, the normal distribution function is used to describe the correlation between each data point and the combined area. The calculation formula for the segmentation error is: M1 = ∑[(x1-v) 2 ] / σ 2 , where M1 is the segmentation error of any combined region, x1 is any data point in the combined region, v is the average of all data points in the combined region, and σ is the standard deviation of all data points in the combined region. Since there are multiple first and second subregions, there are also multiple combined regions. The combined region with the minimum segmentation error is selected as the optimal region, which represents the optimal data point distribution for hemp-core steel wire production.

[0060] Step S3: construct a quality prediction model, input the optimal area into the quality prediction model, output the quality prediction value, obtain the actual quality value corresponding to the optimal area, and obtain the prediction error of the quality prediction value based on the actual quality value.

[0061] Specifically, the quality prediction model is used to analyze the quality of data points in the optimal region. Because the actual quality values of hemp-core steel wire vary, the quality prediction values also vary. Prediction error = |Prediction value - Actual quality value| / Actual quality value. The prediction error can be used to assess the accuracy of the quality prediction model, and the actual quality value refers to the quality value of the hemp-core steel wire after real-time production.

[0062] Step S4: determine whether the quality prediction model has converged based on the prediction error. If not, proceed to step S2 to re-divide the production variables and product variables until the quality prediction model converges.

[0063] Specifically, the convergence of the quality prediction model is assessed to determine whether the model has fully learned the data patterns and whether the prediction error is stable within an acceptable range. If the quality prediction model does not converge, it indicates that the prediction error is large and the corresponding optimal region does not correspond to the optimal data point distribution, which means it cannot represent the optimal quality of hemp-core steel wire production. Therefore, it is necessary to re-divide the production variables and product variables and repeat the steps until the quality prediction model converges.

[0064] Step S5: Calculate the difference between the quality prediction value and the target quality index based on the preset target category. If the difference is less than or equal to the first category threshold, the hemp core steel wire is determined to be qualified. Otherwise, adjust the production variables based on the preset target category.

[0065] Specifically, the preset target category refers to the quality requirements for ultra-high-strength hemp-core steel wire after production, such as tensile strength, elongation, and other categories. The difference between the quality prediction value and the target quality index for the same preset target category is calculated and then compared with the sub-thresholds included in the first-category threshold to determine whether the preset target category meets the requirements, and further comprehensively determine whether the hemp-core steel wire has been produced to quality. Since production variables directly affect the actual quality value of the hemp-core steel wire during the production process, if the hemp-core steel wire fails to meet any of the preset target categories, the value of the production variable is further adjusted.

[0066] In a specific embodiment, the production variables are divided into a plurality of first sub-areas, including:

[0067] (1) Obtain the correlation between production variables, divide all production variables into first-category variables and second-category variables based on the correlation, and establish a first spatial coordinate, where the horizontal axis of the first spatial coordinate represents the first-category variables and the vertical axis of the first spatial coordinate represents the second-category variables.

[0068] (2) Using a clustering algorithm to obtain the boundary values of the first category variables and the second category variables respectively, setting the boundary values as division points, dividing the first spatial coordinates in a direction parallel to the horizontal axis or the vertical axis based on the division points, and generating multiple first sub-regions.

[0069] Specifically, since production variables contain multiple data categories and there may be influencing factors between different production variables, for example, when producing hemp core steel wire, the pulling force will affect the pulling speed. Therefore, it is necessary to use the Pearson Correlation Coefficient to measure the linear correlation between production variables (such as pulling speed, pulling force, temperature, lubricant concentration, mold wear, diameter deviation, surface defects, etc.). The production variables corresponding to strong correlation are set as first-class variables, and the production variables corresponding to weak correlation are set as second-class variables, where both the first-class variables and the second-class variables contain multiple production variables. The first spatial coordinate is a coordinate system drawn by the horizontal axis and the vertical axis.

[0070] Cluster the data points for the first and second category variables using the K-means algorithm. Calculate the mean of all data points in each cluster and set this mean as the boundary value. The number of clusters must be 2 or greater. For example, for the first category variable, if the boundary values are 90 m / min pulling speed and 30 kN pulling force, Cluster 1 contains data points with lower pulling speed and higher pulling force, while Cluster 2 contains data points with higher pulling speed and lower pulling force. Draw parallel lines along the horizontal and vertical axes from the dividing points to form multiple first sub-regions.

[0071] In a specific embodiment, the product variable is divided into a plurality of second sub-areas, including:

[0072] Obtain the relevant variable values of any product variable, generate random numbers for all relevant variable values, set multiple partitioning thresholds, group all random numbers based on the partitioning thresholds, generate multiple random arrays, and set the product variables included in the random arrays as the second sub-area.

[0073] Specifically, the relevant variable value refers to the numerical type of any product variable. The range of the random number is 0 to 1. The division threshold can be set to 0.3 or 0.7. If the random number is <0.3, it is allocated to the second sub-area V1. If 0.3≤random number≤0.7, it is allocated to the second sub-area V2. If the random number is >0.7, it is allocated to the second sub-area V3.

[0074] In a specific embodiment, constructing a quality prediction model includes:

[0075] (1) The production data of the same production batch is set as the first data set, and all production data is set as the second data set.

[0076] (2) Extracting the length data of the hemp core steel wire from any first data set, setting the length data in the row direction, setting the other data of the first data set in the column direction, and generating a two-dimensional matrix of the first data set based on the row direction and the column direction.

[0077] (3) Constructing a first prediction model based on a convolutional neural network, inputting all two-dimensional matrices into the first prediction model for training, and adjusting the model parameters.

[0078] (4) Constructing a second prediction model based on the exponential smoothing method, inputting the second data set into the second prediction model, and calculating the exponential smoothing coefficient of the second prediction model.

[0079] (5) The trained first prediction model and the calculated second prediction model are combined and set as a quality prediction model.

[0080] Specifically, during the production process of hemp-core steel wire, there are multiple production batches. To accurately analyze the production quality of hemp-core steel wire, the production data can be divided into multiple categories. The first data set is the production data of the same production batch. Since there are multiple production batches, there are multiple first data sets. In addition, all production data is not divided and is directly set as the second data set.

[0081] Length data refers to the length record of the hemp core steel wire produced. In order to refine the production quality of the hemp core steel wire, the length data can be divided. For example, the row direction represents a specific position of the hemp core steel wire in the length direction (one data position per 1 meter, forming a 1-meter-long interval). The column direction represents the operating conditions contained in the different production data in the first data set, as well as the quality data of the product. Each element in the two-dimensional matrix represents a data point of corresponding position and corresponding condition. When the hemp core steel wire goes through multiple production processes, for each specific length data, the operating condition data of each process is associated with the quality data.

[0082] Convolutional neural networks are capable of feature extraction and excel at processing two-dimensional data. The first prediction model learns to identify the relationship between operating conditions such as raw material composition, drawing process, and heat treatment and the mechanical properties of the steel wire. It also identifies the relationship between hemp core characteristics and the bond strength of the steel wire, and the impact of the interactions between operating conditions in each process on the quality of the final product. The accuracy of the first prediction model for predicting the mechanical properties and surface quality of the hemp core steel wire is evaluated, and the model parameters are further adjusted.

[0083] Exponential smoothing is a statistical technique used for time series analysis and forecasting. It can be used to perform a weighted average based on historical production data, further analyzing the changing trends of real-time production data. The second dataset contains production data from a longer period of time, so a second forecasting model is used for analysis. The formula for exponential smoothing prediction in the second forecasting model is: y′(i) = λ × y(i-1) + (1-λ) × y′(i-1), where λ is the exponential smoothing coefficient, i is the data point of hemp-core steel wire production, y′(i) is the predicted quality value of the current hemp-core steel wire, y(i-1) is the actual quality value of the previous hemp-core steel wire production, and y′(i-1) is the predicted quality value of the previous hemp-core steel wire production. The exponential smoothing coefficient can be calculated using the second dataset.

[0084] Therefore, the two prediction models are combined to analyze the quality changes in the short term and long term during the production process of hemp core steel wire, and then a quality prediction model is constructed. Among them, the first prediction model belongs to short-term quality analysis, and the second prediction model belongs to long-term quality analysis.

[0085] In a specific embodiment, outputting the quality prediction value includes:

[0086] (1) The quality prediction model obtains the production data contained in the optimal area and sets it as the data to be analyzed. The data to be analyzed is preprocessed and then input into the first prediction model and the second prediction model respectively.

[0087] (2) The first prediction model outputs a first performance prediction result of the hemp core steel wire, and the second prediction model outputs a smoothing result and a proportional coefficient of the hemp core steel wire based on the production time of the data to be analyzed. If the deviation of the data to be analyzed within the production time is greater than a first preset value, the product of the smoothing result and the proportional coefficient is set as the second performance prediction result; otherwise, the sum of the smoothing result and the proportional coefficient is set as the second performance prediction result.

[0088] (3) Based on the prediction data category, the first performance prediction result and the second performance prediction result are weighted and summarized to generate a quality prediction value.

[0089] Specifically, Figure 2 A flowchart for outputting quality prediction values for the quality prediction model. The production data contained in the optimal region represents the expected optimal production quality for hemp-core steel wire production. Therefore, the data to be analyzed is preprocessed and then input into the quality prediction model. The first and second prediction models are used to analyze and predict the data to be analyzed. Data preprocessing includes steps such as data cleaning and data conversion.

[0090] The trained first prediction model extracts features of the data to be analyzed, produces a feature map, and outputs a prediction result after integrating feature information, namely the first performance prediction result, which refers to the production quality prediction corresponding to the data to be analyzed in the first prediction model.

[0091] Production time refers to the time record contained in the data to be analyzed in the production data, and the smoothing result refers to the quality prediction value of the hemp core steel wire calculated and output by the second prediction model using the exponential smoothing method for the data to be analyzed. The proportional coefficient refers to the factor used to adjust the second prediction model for long-term learning prediction. The ratio of the actual quality value corresponding to the previous production time to the quality prediction value output by the second prediction model is set as the proportional coefficient. Deviation refers to the fluctuation deviation of the data to be analyzed. When the deviation is greater than the first preset value, it means that the fluctuation of the data to be analyzed is large. The product of the smoothing result and the proportional coefficient can maintain the proportional relationship between the predicted value and the actual value. Therefore, it is set as the second performance prediction result. Otherwise, it means that the fluctuation of the data to be analyzed is small. The addition of the smoothing result and the proportional coefficient can maintain the absolute deviation between the predicted value and the actual value time, more accurately reflecting the changing trend of the data. Therefore, it is set as the second performance prediction result.

[0092] The prediction data categories refer to different quality categories, and the weighted aggregation refers to adding the first performance prediction result and the second performance prediction result by using a weighting coefficient to generate a quality prediction value.

[0093] In a specific embodiment, after dividing the production variables into a plurality of first sub-areas, re-dividing the production variables includes:

[0094] (1) Obtain data points included in the first sub-region, construct a first function based on the covariance matrix, and calculate the first function value of the data point based on the first function.

[0095] (2) Calculate the regional prediction value of the data point based on the first function value, and use the weighted average method to calculate the regional error of the first sub-region according to the regional prediction value.

[0096] (3) Repeat this step to calculate the regional errors of all first sub-regions, and set the first sub-region corresponding to the maximum regional error as the region to be divided.

[0097] (4) Construct new first spatial coordinates for the area to be divided, and divide the new first spatial coordinates.

[0098] Specifically, the formula of the first function is: Where G(j) is the value of the first function, j is the data point in the first subregion, r is the centroid of the first subregion, and J is the total number of data points in the first subregion. The first function can be used to evaluate the distribution trend of the data points in the first subregion.

[0099] The calculation formula for regional prediction value is: Where T is the actual quality value of the data point, and T′(j) is the regional prediction value of the jth data point. The weighted average method refers to the calculation process of the weighted average absolute error. The calculation formula of the regional error is: Wherein, H is the area error of the first sub-area.

[0100] Since there are multiple first sub-regions, it is necessary to repeatedly calculate the regional errors of all first sub-regions, and continue to divide the first sub-region with the largest regional error, that is, the region to be divided.

[0101] The specific steps of dividing the area to be divided are as described above.

[0102] In a specific embodiment, determining whether the quality prediction model has converged includes:

[0103] (1) Extract a first subset from the quality prediction values, calculate the error mean and error standard deviation of the first subset based on the prediction error, set a fluctuation coefficient, multiply the fluctuation coefficient by the error standard deviation, and then add the result to the error mean to generate a second preset value.

[0104] (2) If the prediction error is less than or equal to the second preset value, it is determined that the quality prediction model has converged; otherwise, the quality prediction model has not converged.

[0105] Specifically, the first subset can be used to evaluate the convergence of the quality prediction model. The first subset can be extracted according to time window sampling. The mean error refers to the average of the prediction errors of all first subsets. A smaller mean error indicates higher model prediction accuracy. The standard deviation of the error refers to the standard deviation of the prediction errors of all first subsets. A smaller standard deviation of the error indicates more stable model prediction results. The second preset value = fluctuation coefficient × standard deviation of the error + mean error. By comparing the prediction error with the second preset value, the learning error of the quality prediction model can be evaluated to ensure that the model achieves the expected prediction accuracy.

[0106] In one embodiment, adjusting production variables based on a preset target category includes:

[0107] (1) Extract the preset target categories whose difference is greater than the first category threshold and set them as the categories to be adjusted. Obtain the historical fluctuation data of the production variables in any category to be adjusted, calculate the correlation between any two groups of historical fluctuation data, and combine the historical fluctuation data with a correlation greater than a third preset value into an associated adjustment group.

[0108] (2) Calculate the mean fluctuation ratio of the associated adjustment group and adjust the numerical value of the production variable based on the mean fluctuation ratio.

[0109] Specifically, Figure 3This is a diagram of the production variable adjustment strategy. For each preset target category (such as tensile strength, elongation, and diameter deviation), the difference between the quality prediction value and the target quality indicator is calculated. All preset target categories with a difference greater than the first threshold are screened out. These categories are identified as requiring adjustment, i.e., the categories to be adjusted.

[0110] Historical fluctuation data refers to the historical data changes over a period of time for each production variable (e.g., drawing speed, drawing force, temperature, etc.) involved in each category to be adjusted. The correlation between any two sets of historical fluctuation data is calculated using the Pearson correlation coefficient. Correlated adjustment groups are groups of categories with strong correlations.

[0111] The mean fluctuation ratio refers to the average of the change in the production variable between the current time period and the previous time period. It measures the overall fluctuation of that group of production variables. A large mean fluctuation ratio indicates significant fluctuations in that group of production variables, requiring appropriate adjustments. The direction of adjustment is determined by the difference (positive or negative) between the predicted quality value and the target value, and the adjustment range is determined by the mean fluctuation ratio. For example, if the predicted tensile strength value is lower than the target value, the pull-out force needs to be increased or the pull-out speed needs to be reduced. The larger the mean fluctuation ratio, the larger the adjustment range. The adjustment range can be set using either a linear or nonlinear relationship.

[0112] In a specific embodiment, collecting production data includes:

[0113] (1) Use spectral analysis to obtain the chemical composition of the hemp core material, perform physical testing on the hemp core material to obtain physical performance data, use data acquisition equipment to obtain drawing parameters, use a high-resolution camera to perform real-time imaging of the surface of the hemp core steel wire to generate surface quality data, and use an ultrasonic detector to test the hemp core steel wire to output tensile strength and elongation.

[0114] (2) Based on the production process, the chemical composition, physical property data, drawing parameters, surface quality data, tensile strength and elongation are summarized and set as production data.

[0115] Specifically, hemp core material refers to the material used to produce hemp core steel wire. This material must be tested before production, including its chemical composition and physical properties. Drawing parameters include, but are not limited to, drawing speed and drawing force. Surface quality data includes, but is not limited to, defect type and number. An ultrasonic detector is used to assess the internal defects and mechanical properties of the hemp core steel wire. All test data is aggregated and set as production data.

[0116] The above describes the production quality monitoring method for ultra-high strength hemp core steel wire in the embodiment of the present application. The following describes the production quality monitoring system for ultra-high strength hemp core steel wire in the embodiment of the present application. Figure 4In one embodiment of the present application, a production quality monitoring system for ultra-high strength hemp core steel wire includes:

[0117] The acquisition module 201 is used to set target quality indicators for the production of hemp core steel wire, set data acquisition equipment based on the production process, and collect production data, which includes production variables and product variables.

[0118] The combination module 202 is used to divide the production variables into multiple first sub-regions and the product variables into multiple second sub-regions, extract one first sub-region and one second sub-region respectively and combine them into a combination region, calculate the segmentation error of any combination region based on the normal distribution function, and set the combination region corresponding to the minimum segmentation error as the optimal region.

[0119] The prediction module 203 is used to build a quality prediction model, input the optimal area into the quality prediction model, output a quality prediction value, obtain the actual quality value corresponding to the optimal area, and obtain a prediction error of the quality prediction value based on the actual quality value.

[0120] The verification module 204 is used to determine whether the quality prediction model has converged based on the prediction error. If not, the production variables and product variables are re-divided until the quality prediction model converges.

[0121] The control module 205 is used to calculate the difference between the quality prediction value and the target quality index according to the preset target category. If the difference is less than or equal to the first category threshold, the hemp core steel wire is determined to be qualified. Otherwise, the production variables are adjusted based on the preset target category.

[0122] Through the collaborative efforts of the above components, production data is first collected from multiple sources, covering key parameters of each production link from raw materials to finished products, which can fully reflect the production status. Based on the results of correlation analysis, production variables are divided into first-category variables and second-category variables. A clustering algorithm is used to divide the spatial coordinates to generate multiple first sub-regions. Random number generation and threshold partitioning are used to divide the product variables into multiple second sub-regions. This data-driven regional division method avoids the subjectivity of human division, improves the scientific nature and accuracy of regional division, and can adapt to changes in production conditions and adjust the regional division strategy in a timely manner, ensuring the accuracy and effectiveness of quality control and improving quality control efficiency. Then, a convolutional neural network is used to model data from the same production batch to capture the complex relationship between production data and product quality within the batch. An exponential smoothing method is used to model data from all production batches to capture overall production trends and long-term patterns. The final quality prediction model is generated. It can predict product quality across different production batches and product types, has strong versatility, and improves the stability and accuracy of the prediction results. Finally, based on the difference between the forecast error and the target value, the categories of production variables that need to be adjusted are identified, the correlation between historical fluctuation data is analyzed, and the data with high correlation are combined into associated adjustment groups. Based on the mean of the fluctuation ratio, the production variables are adjusted, avoiding the impact of a single variable adjustment on other variables, improving the controllability of the adjustment effect, and continuously adjusting the production variables to keep them in the optimal state, which is conducive to improving the stability of the production process and the consistency of product quality.

[0123] Those skilled in the art will 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.

[0124] 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, and the computer software product is stored in a storage medium, including a number of instructions for enabling a computer device (which can be a personal computer, a server, or a network device, etc.) to execute 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, and other media that can store program code.

[0125] 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 above embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the above 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 method for monitoring the production quality of ultra-high strength hemp core steel wire, characterized in that: The production quality monitoring method for ultra-high strength hemp core steel wire comprises: S1: Setting target quality indicators for hemp core steel wire production, and setting data acquisition equipment based on the production process to collect production data, wherein the production data includes production variables and product variables; S2: Divide the production variable into a plurality of first sub-regions, and divide the product variable into a plurality of second sub-regions, extract one first sub-region and one second sub-region respectively and combine them into a combined region, calculate the segmentation error of any of the combined regions based on a normal distribution function, and set the combined region corresponding to the minimum segmentation error as the optimal region; S3: constructing a quality prediction model, inputting the optimal region into the quality prediction model, outputting a quality prediction value, obtaining an actual quality value corresponding to the optimal region, and obtaining a prediction error of the quality prediction value based on the actual quality value; S4: judging whether the quality prediction model has converged based on the prediction error; if not, proceeding to step S2 to re-divide the production variables and the product variables until the quality prediction model converges; S5: Calculate the difference between the quality prediction value and the target quality index based on the preset target category. If the difference is less than or equal to the first category threshold, determine that the hemp core steel wire production is qualified. Otherwise, adjust the production variable based on the preset target category.

2. The method for monitoring the production quality of ultra-high strength hemp core steel wire according to claim 1, characterized in that: The dividing the production variable into a plurality of first sub-areas comprises: Obtaining correlations between the production variables, dividing all the production variables into first-category variables and second-category variables based on the correlations, and establishing a first spatial coordinate system, wherein the horizontal axis of the first spatial coordinate system represents the first-category variables and the vertical axis of the first spatial coordinate system represents the second-category variables; A clustering algorithm is used to respectively obtain the boundary values of the first category variables and the second category variables, the boundary values are set as division points, and the first spatial coordinates are divided based on the division points in a direction parallel to the horizontal axis or the vertical axis to generate multiple first sub-areas.

3. The method for monitoring the production quality of ultra-high strength hemp core steel wire according to claim 1, characterized in that: The dividing the product variable into a plurality of second sub-areas includes: Obtain relevant variable values of any of the product variables, generate random numbers for all the relevant variable values, set multiple division thresholds, group all the random numbers based on the division thresholds, generate multiple random arrays, and set the product variables contained in the random arrays as the second sub-area.

4. The method for monitoring the production quality of ultra-high strength hemp core steel wire according to claim 1, characterized in that: The construction quality prediction model includes: Setting the production data of the same production batch as a first data set, and setting all the production data as a second data set; Extracting length data of the hemp core steel wire from any of the first data sets, setting the length data in a row direction, setting other data of the first data set in a column direction, and generating a two-dimensional matrix of the first data set based on the row direction and the column direction; Building a first prediction model based on a convolutional neural network, inputting all the two-dimensional matrices into the first prediction model for training, and adjusting model parameters; constructing a second prediction model based on an exponential smoothing method, inputting the second data set into the second prediction model, and calculating an exponential smoothing coefficient of the second prediction model; The trained first prediction model and the calculated second prediction model are combined and set as the quality prediction model.

5. The method for monitoring the production quality of ultra-high strength hemp core steel wire according to claim 4, characterized in that: The output quality prediction value includes: The quality prediction model obtains the production data contained in the optimal area and sets it as data to be analyzed, and inputs the data to be analyzed into the first prediction model and the second prediction model after data preprocessing; The first prediction model outputs a first performance prediction result of the hemp-core steel wire, and the second prediction model outputs a smoothing result and a proportional coefficient of the hemp-core steel wire based on the production time of the data to be analyzed. If the deviation of the data to be analyzed within the production time is greater than a first preset value, the product of the smoothing result and the proportional coefficient is set as the second performance prediction result; otherwise, the sum of the smoothing result and the proportional coefficient is set as the second performance prediction result; The first performance prediction result and the second performance prediction result are weighted and summarized based on the prediction data category to generate the quality prediction value.

6. The method for monitoring the production quality of ultra-high strength hemp core steel wire according to claim 2, characterized in that: After dividing the production variable into a plurality of first sub-areas, re-dividing the production variable includes: Acquire data points included in the first sub-region, construct a first function based on the covariance matrix, and calculate a first function value of the data point based on the first function; Calculating a regional prediction value of the data point based on the first function value, and calculating a regional error of the first sub-region according to the regional prediction value using a weighted average method; Repeat this step to calculate the regional errors of all the first sub-regions, and set the first sub-region corresponding to the largest regional error as the region to be divided; New first spatial coordinates are constructed for the area to be divided, and the new first spatial coordinates are divided.

7. The method for monitoring the production quality of ultra-high strength hemp core steel wire according to claim 1, characterized in that: Determining whether the quality prediction model has converged includes: Extracting a first subset from the quality prediction values, calculating an error mean and an error standard deviation of the first subset based on the prediction error, setting a fluctuation coefficient, multiplying the fluctuation coefficient by the error standard deviation and then adding the result to the error mean to generate a second preset value; If the prediction error is less than or equal to the second preset value, it is determined that the quality prediction model has converged; otherwise, the quality prediction model has not converged.

8. The method for monitoring the production quality of ultra-high strength hemp core steel wire according to claim 1, characterized in that: The adjusting the production variable based on the preset target category includes: extracting the preset target category whose difference is greater than the first category threshold and setting it as the category to be adjusted, obtaining historical fluctuation data of the production variable in any of the categories to be adjusted, calculating the correlation between any two groups of the historical fluctuation data, and combining the historical fluctuation data with a correlation greater than a third preset value into a correlated adjustment group; The fluctuation ratio mean of the associated adjustment group is calculated, and the numerical value of the production variable is adjusted based on the fluctuation ratio mean.

9. The method for monitoring the production quality of ultra-high strength hemp core steel wire according to claim 1, characterized in that: The collection of production data includes: Using spectral analysis to obtain the chemical composition of the hemp core material, performing physical testing on the hemp core material to obtain physical property data, the data acquisition device obtains drawing parameters, using a high-resolution camera to perform real-time imaging of the surface of the hemp core steel wire to generate surface quality data, and using an ultrasonic detector to test the hemp core steel wire to output tensile strength and elongation; The chemical composition, the physical property data, the drawing parameters, the surface quality data, the tensile strength and the elongation are summarized based on the production process and are all set as the production data.

10. A production quality monitoring system for ultra-high strength hemp core steel wire, characterized in that: The production quality monitoring system for ultra-high strength hemp core steel wire comprises: An acquisition module is used to set target quality indicators for hemp core steel wire production and to set data acquisition equipment based on the production process to collect production data, wherein the production data includes production variables and product variables; a combining module, configured to divide the production variable into a plurality of first sub-regions, divide the product variable into a plurality of second sub-regions, extract one of the first sub-regions and one of the second sub-regions respectively and combine them into a combined region, calculate a segmentation error of any of the combined regions based on a normal distribution function, and set the combined region corresponding to the minimum segmentation error as the optimal region; a prediction module, configured to construct a quality prediction model, input the optimal region into the quality prediction model, output a quality prediction value, obtain an actual quality value corresponding to the optimal region, and obtain a prediction error of the quality prediction value based on the actual quality value; a verification module, configured to determine whether the quality prediction model has converged based on the prediction error, and if not, re-divide the production variables and the product variables until the quality prediction model has converged; The control module is used to calculate the difference between the quality prediction value and the target quality index according to the preset target category. If the difference is less than or equal to the first category threshold, the hemp core steel wire is judged to be qualified; otherwise, the production variable is adjusted based on the preset target category.

Images