A convexity self-learning size gear optimization method and system

Abstract

This invention discloses a convexity self-learning dimension level optimization method and system, belonging to the field of metal rolling technology. The method includes: collecting actual production data within a preset production cycle at the production site and integrating the data required for level optimization; determining the effective target size range of the strip and dividing the effective target size range into multiple intervals; constructing a loss function to evaluate the rationality of the dimension level division; taking the number of levels and the number of intervals contained in each level as the optimization object, and based on the loss function, using a swarm intelligence algorithm to solve for the optimal solution of the optimization object, thereby realizing convexity self-learning dimension level optimization. Using the convexity self-learning dimension level optimization method and system provided by this invention can improve the scientific reliability of optimization work, while also increasing optimization efficiency, enhancing the performance of the self-learning scheme in terms of accuracy and stability, and ultimately effectively improving the strip shape control effect.

Description

Technical Field

[0001] This invention relates to the field of metal rolling technology, and in particular to a method and system for optimizing crown self-learning dimension levels. Background Technology

[0002] Strip products are widely used in various industries. The quality of strip shape directly determines the overall competitiveness of the product. Due to factors such as fluctuations in operating conditions and changes in the environment, whose influencing mechanisms are often unknown, existing strip shape control models must employ data-based statistical prediction or data-driven methods to achieve error compensation and improve control effectiveness. Based on exponential smoothing, real-time compensation within the same roll period (referring to the work roll changing cycle) is achieved. Subsequently, a tiered memory is formed according to product type and specifications, enabling the recall of historical compensation at different roll periods. This self-learning scheme is the mainstream approach commonly used in current strip shape control models. Crown is an important indicator for evaluating strip shape, and crown self-learning is a crucial way to maintain effective strip shape control.

[0003] Mainstream self-learning solutions employ a combination of immediate compensation (short-term self-learning) and historical compensation (long-term self-learning) to identify and memorize patterns in deviations, ultimately reducing the frequency of requiring significant immediate compensation and improving the stability of sheet shape control. Products are categorized by material, width, and thickness, and deviations corresponding to different materials or sizes are memorized within the corresponding categorization level. However, in actual production, unreasonable categorization often affects sheet shape quality. For example, in a production line with convexity self-learning, the range [1400mm, 1600mm] is categorized into the same width categorization. This means that, with the material and thickness categorizations unchanged, all products within this width range will share the same long-term self-learning value for error compensation. If the width categorization is appropriate, all sheets and strips produced under similar conditions should exhibit comparable convexity performance across their corresponding width ranges, and the long-term self-learning value should remain relatively stable. Production data shows that within the same rolling period, when rolling products with widths of [1400mm, 1500mm] and [1500mm, 1600mm] respectively, the long-term self-learning value consistently exhibits significant changes, and the crown of the first product after the width transition shows considerable fluctuations. Similar issues exist within the thickness range [2mm, 3mm], with 2.8mm as the dividing point, and will not be elaborated further. Therefore, the rationality of the range division directly determines the stability and accuracy of the long-term self-learning value, thus affecting the final sheet shape control effect.

[0004] However, the problem of how to scientifically optimize the size range division remains unresolved. Currently, there are two main approaches to addressing improper size range division. One approach relies on experience to manually adjust long-term learning values ​​to an intermediate level and raise the threshold for triggering long-term self-learning. While this approach does reduce fluctuations in long-term self-learning values, it increases the model's dependence on short-term self-learning, reducing the stability of shape control and contradicting the original intention of long-term self-learning to uncover and memorize regular deviations. The other approach involves manually analyzing data to identify locations where improper size range division is evident. Then, the corresponding size range is roughly adjusted, and after rolling for a period, the data is analyzed and adjusted again, repeating this process until the problem is resolved. To a certain extent, this approach can improve the rationality of size range division, but due to its trial-and-error nature, achieving optimization requires significant time and manpower. From a problem-solving efficiency perspective, neither of these methods adequately meets the needs of practical applications.

[0005] Furthermore, in existing technologies, researchers generally focus on accurately obtaining the convexity of the strip itself. For example, some scholars have proposed a convexity prediction method based on a neural network model, a high-precision convexity calculation framework based on an ensemble learning model, and a convexity optimization control framework based on model-driven digital twins. All of these methods aim to improve the accuracy of convexity calculation and control performance, collecting production data and extracting features, and employing statistical regression and machine learning algorithms as implementation approaches. Although these methods improve the accuracy of convexity calculation to some extent, due to the "black box" nature of machine learning, they currently deal more with theoretical research. Their practicality, stability, and interpretability cannot compare with existing convexity acquisition solutions currently in industrial applications.

[0006] In summary, there is still an urgent need for a scientific and effective solution to improve the self-learning dimension division of convexity in the current strip shape control optimization. Summary of the Invention

[0007] This invention provides a method and system for optimizing convexity self-learning size ranges to solve the technical problem that existing schemes for optimizing size range division are difficult to meet the needs of practical applications.

[0008] To solve the above-mentioned technical problems, the present invention provides the following technical solution: On one hand, the present invention provides a method for optimizing convexity self-learning size levels, including: Collect actual production data from the production site within the preset production cycle, and integrate the data required for gear optimization. Determine the effective target size range of the strip, divide the effective target size range into multiple intervals, and calculate the feature data corresponding to each interval; Construct a loss function to evaluate the rationality of size segmentation; Taking the number of gears and the number of intervals contained in each gear as the optimization objects, and based on the loss function, the optimal solution of the optimization objects is solved by swarm intelligence algorithm to realize the self-learning size gear optimization of convexity.

[0009] Furthermore, the actual production data includes the model-calculated values ​​of the target thickness, target width, and convexity of each strip, the measured values ​​of the convexity, the long-term self-learning value of the convexity, and the short-term self-learning value of the convexity. The integrated data required for gear optimization includes: Calculate the total number of strips, M, produced within the preset production cycle at the production site; Calculate the target dimensions for each strip; where the target dimensions are the target thickness or target width. Calculate the convexity calculation residual for each strip; wherein the convexity calculation residual is calculated based on the model calculation value of convexity, the measured value of convexity, the long-term self-learning value of convexity, and the short-term self-learning value of convexity.

[0010] Furthermore, the formula for calculating the residual of the convexity calculation is as follows:

[0011] in, Calculate the residuals for convexity; This represents the measured value of the convexity. The calculated value for the convexity model; The sum of the long-term self-learning value and the short-term self-learning value of convexity.

[0012] Furthermore, the effective target size range is a range consisting of the minimum and maximum values ​​of the target size of the strip as specified in the product outline; The calculation of the feature data corresponding to each interval includes: Calculate the total number of intervals divided by the effective target size range; For each interval, calculate the mean value of the convexity calculation residuals of all strips whose target size falls within its range, and record it as the mean residual value within the interval. For each interval, calculate the ratio of the number of strips with the target size falling within its range to M, and denot it as the production quantity percentage for that interval. For example, if the size interval is [lb, ub), and the number of strips with the target size size∈[lb, ub) is n, then the production quantity percentage for that interval is n / M.

[0013] Furthermore, the expression for the loss function is:

[0014] in, This represents the calculated loss value; This indicates the loss caused by differences within the file; This indicates the loss caused by differences between grades; , and As preset weight values, the sum of the two is 1. This represents the variance of the mean residuals within all intervals corresponding to the ranges whose loss value is to be calculated. This represents the variance of the production quantity percentage of all intervals included in the tier for which the loss value is to be calculated. , B This represents the baseline value, which takes the value of... The maximum value, e Represents the base of natural numbers. W g Indicates the gain coefficient. This indicates the gear at which the loss value is to be calculated and the gear between adjacent gears. The absolute gap This indicates the gear at which the loss value is to be calculated and the gear between adjacent gears. The absolute gap.

[0015] Furthermore, the optimization of gear positions, using the number of gears and the number of intervals contained in each gear position as the optimization object, and based on the loss function, using a swarm intelligence algorithm to iteratively solve for the optimal solution of the optimization object to achieve convexity self-learning size gear position optimization, includes: Determine the threshold for the number of gears and the threshold for the number of intervals contained in a single gear; Based on the threshold of the number of gears, determine all possible values ​​of the number of gears and form a set of gear numbers; Iterate through the set of gear numbers; For each gear number traversed, based on the loss function, under the constraint of the threshold number of intervals contained in a single gear number, the optimal solution for the number of intervals contained in each gear number under the current gear number is iteratively solved using a swarm intelligence algorithm, until the set of gear numbers is traversed and the optimal solution for the number of intervals contained in each gear number under different gear numbers is obtained. Among the optimal solutions for the number of intervals contained in each gear under different gear numbers, the optimal solution with the minimum loss value is selected, and the corresponding gear number and the number of intervals contained in each gear under this gear number are taken as the gear optimization result.

[0016] Furthermore, the step of using a swarm intelligence algorithm to iteratively solve for the optimal solution of the number of intervals contained in each gear under the current gear position includes: Represent the result of each iteration as a gear size vector: [ x 1 ,x 2 ,… , x n];in, x i Indicates the first i The number of intervals contained in each gear. i =1,2,…, n , n Indicates the current gear number; right x 1 ,x 2 ,… , x n The value is adjusted to satisfy: x 1 ,x 2 ,… , x n The sum equals the total number of intervals divided by the effective target size range, resulting in the adjusted... x 1 ,x 2 ,… , x n The value; The effective target size range is divided into n There are several gear levels; among them, the number of intervals contained in each gear level and the adjusted... x 1 ,x 2 ,… , x n The values ​​correspond one-to-one, and the divided gears are used as the current iteration result; Calculate the loss value for each level in the current iteration result and sum the calculation results to obtain the loss value of the current iteration result; wherein, the loss value for each level is calculated using the loss function. The iteration result with the minimum loss value is determined, and the number of intervals contained in each gear is taken as the optimal solution for the number of intervals contained in each gear under the current gear number.

[0017] Furthermore, the aforementioned x 1 ,x 2 ,… , x n The value is adjusted to satisfy: x 1 ,x 2 ,… , x n The sum equals the total number of intervals divided by the effective target size range, including: judge x 1 ,x 2 ,… , x n If the sum of the values ​​is not equal to the total number of intervals divided by the effective target size range, then...x 1 ,x 2 ,… , x n The value is scaled proportionally, and the scaling result is rounded to the nearest integer to obtain the initially adjusted value. x 1 ,x 2 ,… , x n The value; where the scaling ratio is the sum of the total number of intervals divided by the effective target size range and the value of the scaling ratio. x 1 , x 2 ,… , x n The ratio of the sum to the sum; If the initial adjustment x 1 ,x 2 ,… , x n The value still does not meet the requirements: x 1 ,x 2 ,… , x n The sum equals the total number of intervals divided by the effective target size range; then calculate the preliminary adjusted... x 1 ,x 2 ,… , x n The difference between the sum of the values ​​and the total number of intervals divided by the effective target size range; if the difference is greater than 0, then the initially adjusted... x 1 ,x 2 ,… , x n Subtract the difference from the maximum value in the range. If the difference is less than 0, then the initially adjusted value will be... x 1 ,x 2 ,… , x n The minimum value in the range is added to the difference to obtain the final adjusted value. x 1 ,x 2 ,… , x n The value of .

[0018] Furthermore, after using swarm intelligence algorithms to find the optimal solution for the optimization object, the method also includes: Suppose that the optimal solution of the optimization object corresponds to the number of gears, s; The effective target size range is divided into s levels; wherein, the number of intervals contained in each level corresponds one-to-one with the number of intervals contained in each level corresponding to the optimal solution of the optimization object. The level is used as the optimized level to obtain the value range of the target size corresponding to each level.

[0019] On the other hand, the present invention also provides a convexity self-learning size level optimization system, comprising: The data acquisition and integration module is used to collect actual production data within a preset production cycle at the production site and integrate the data required for gear optimization. The data processing module is used to determine the effective target size range of the strip, divide the effective target size range into multiple intervals, and calculate the feature data corresponding to each interval. The gear optimization module is used for: Construct a loss function to evaluate the rationality of size segmentation; Taking the number of gears and the number of intervals contained in each gear as the optimization objects, and based on the loss function, the optimal solution of the optimization objects is solved by swarm intelligence algorithm to realize the self-learning size gear optimization of convexity.

[0020] In another aspect, the present invention also provides an electronic device comprising a processor and a memory; wherein the memory stores at least one instruction, which is loaded and executed by the processor to implement the above-described method.

[0021] In another aspect, the present invention also provides a computer-readable storage medium storing at least one instruction, which is loaded and executed by a processor to implement the above method.

[0022] The beneficial effects of the technical solution provided by this invention include at least the following: 1. Based on the idea of ​​minimum interval, this invention transforms the abstract problem of convexity self-learning size gear optimization (which is essentially the optimization of the size range corresponding to the gear) into the problem of finding the optimal number of intervals that a specific gear can contain, thus achieving the effect of reasonably simplifying the problem.

[0023] 2. When designing the loss function for evaluating the rationality of size gradation, this invention comprehensively considers two factors: the residual of convexity calculation and the production quantity ratio, thereby improving the scientific reliability of the gradation optimization results.

[0024] 3. This invention employs a swarm intelligence algorithm, which improves the efficiency of optimization. Furthermore, through self-checking of gear information, the total number of intervals can be kept constant, thus ensuring the practical usability of the optimization results. Attached Figure Description

[0025] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0026] Figure 1 This is a flowchart illustrating the convexity self-learning size level optimization method provided in an embodiment of the present invention; Figure 2 This is a schematic diagram of the convexity self-learning size level optimization method provided in the embodiment of the present invention; Figure 3 This is an embodiment of the invention providing the residual level of interval convexity calculation and the division of gears before and after optimization; Figure 4 This refers to the production quantity ratio of different zones and the division of production levels before and after optimization, as provided in the embodiments of the present invention. Figure 5 This is a system block diagram of the electronic device provided in the embodiments of the present invention. Detailed Implementation

[0027] To make the objectives, technical solutions, and advantages of the present invention clearer, the embodiments of the present invention will be described in further detail below with reference to the accompanying drawings.

[0028] First, it should be noted that in the embodiments of the present invention, the words "exemplarily," "for example," etc., are used to indicate that they are examples, illustrations, or descriptions. Any embodiment or design scheme described as "exemplary" in the present invention should not be construed as being more preferred or advantageous than other embodiments or design schemes. Specifically, the use of the term "exemplarily" is intended to present the concept in a specific manner. Furthermore, in the embodiments of the present invention, the meaning expressed by "and / or" can be both, or it can be either one or the other.

[0029] First Embodiment

[0030] This embodiment provides a convexity self-learning dimension level optimization method, which can be implemented by an electronic device. The execution flow of this method is as follows: Figure 1 As shown, it includes the following steps: S1 collects actual production data within a preset production cycle at the production site and integrates the data required for gear optimization. The actual production data includes the target thickness, target width, model-calculated convexity values, measured convexity values, long-term self-learning values, and short-term self-learning values ​​of convexity for each strip. Furthermore, it should be noted that to ensure the reliability of the results, the collected data should cover the main products on-site and have sufficiently rich operating conditions; it is recommended to use the production data from the most recent year on-site.

[0031] Furthermore, the data required for gear optimization was integrated, including: S11, Count the total number of strips M produced in the preset production cycle at the production site; S12, Calculate the target size for each strip; where the target size is determined based on the optimization target: if the width size is optimized, then the target width is corresponding to it; if the thickness size is optimized, then the target thickness is corresponding to it. S13, calculate the convexity calculation residual for each strip; where the formula for calculating the convexity calculation residual is:

[0032] in, Calculate the residuals (hereinafter referred to as residuals) for convexity. This represents the measured value of the convexity. The calculated value for the convexity model; The sum of the long-term self-learning value and the short-term self-learning value of convexity. The model-calculated value of convexity, the long-term self-learning value, and the short-term self-learning value are all calculated online by the plate shape control model and can be exported from the database of the online system.

[0033] S2, determine the effective target size range of the strip, divide the effective target size range into multiple intervals, and calculate the feature data corresponding to each interval; The effective target size range is defined by the minimum and maximum values ​​of the target dimensions (target width or target thickness) of the strip as specified in the field product outline. Each interval is obtained by fully subdividing the effective size range according to equal distances or custom rules. The interval lengths may vary, but should be sufficiently refined to improve optimization performance. Each grade may encompass multiple intervals.

[0034] Furthermore, the feature data corresponding to each interval is calculated, including: S21, Calculate the total number K of intervals divided by the effective target size range; S22, For each interval, calculate the mean value of the convexity calculation residuals of all strips whose target size falls within its range, and record it as the mean value of the residuals within the interval. S23, for each interval, calculate the ratio of the number of strips with the target size falling within its range to M, and record it as the interval production quantity percentage.

[0035] S3, Construct a loss function to evaluate the rationality of size segmentation; Specifically, in this embodiment, considering both the residual from convexity calculation and the production quantity ratio, a loss function is designed to evaluate the rationality of size grading. This mainly includes three steps: identifying influencing factors, determining characterization indicators, and designing the loss function. The influencing factors include two aspects: intra-grading differences, representing the differences between different intervals encompassed by a grading level; and inter-grading differences, representing the differences between a grading level and its adjacent grading levels. Characterization indicators include: ;in, It is the variance of the residual levels corresponding to multiple intervals (the residual level of an interval corresponds to the mean of the residuals in that interval). It is the variance of the proportion of quantities corresponding to multiple intervals (hereinafter referred to as proportion). It is the mean of the residual levels corresponding to multiple intervals. It is the average of the proportions of corresponding quantities across multiple intervals. The calculation formulas for each indicator are as follows:

[0036] in, k This represents the number of intervals contained in the gear. This represents the residual level (i.e., the residual mean) for each interval. P i This represents the percentage of quantities in each interval. Note that before calculating all performance indicators, it is necessary to first... and P i Normalization is performed using 0-1 standardization.

[0037] The loss function is constructed by considering influencing factors and combining characteristic indicators, and its expression is as follows:

[0038] in, Loss Represents the total loss. Loss in Losses due to discrepancies within the same document. Loss out This represents the loss caused by differences between grade levels.

[0039] Losses due to differences within the file Loss in The formula is as follows:

[0040] in, W e The weights considered represent the residual variance.W p The weights representing the variance of the proportions are considered, and their sum equals 1. Within the same range, the larger the residual variance and the proportion variance of each interval, the greater the dispersion of the two, indicating that the intervals within the current range are less similar. Therefore, the corresponding loss function value should be increased to avoid or reduce the possibility of the corresponding partitioning method.

[0041] Losses due to differences between grades Loss out The formula is as follows:

[0042] in, B Representing a benchmark value, it should be guaranteed that... Loss out and Loss in Using values ​​of the same order of magnitude as the specific standard for selection, and avoiding significant differences in the impact of the two types of losses, it is recommended to use... Loss in The maximum value. Mathematically, after a dataset is standardized to 0-1, the maximum variance of any subset is 0.25. This maximum value is reached if and only if the subset contains both 0 and 1 in equal proportions. Therefore, it is not difficult to deduce that: and P var The extreme values ​​are all 0.25, and because W e and W p The sum of all these is always equal to 1, which leads to the conclusion that... Loss in The maximum value, i.e. B Recommended value: 0.25; W g This represents the gain coefficient, which can be any positive value depending on the actual situation. Represents the relationship between the current gear and adjacent gears. The absolute difference; P diff Represents the relationship between the current gear and adjacent gears. P avg The absolute difference. The larger the absolute difference between the mean residual and the mean proportion between different grades, the greater the degree of difference between grades, indicating a low degree of approximation between different grades. Therefore, the loss function value should be reduced to increase the likelihood of the corresponding division method. and P diff The calculation is based on the actual number of adjacent gears available to the current gear: if the current gear does not include the boundary of the effective target size range, then... and P diffIt is the sum of the absolute values ​​of the differences between its two adjacent gears; if the current gear includes the boundary of the effective target size range, then... and P diff It is the absolute value of the difference between adjacent gears on the side with the same size.

[0043] S4, taking the number of gears and the number of intervals contained in each gear as the optimization object, and using the loss function, the optimal solution of the optimization object is solved by swarm intelligence algorithm to realize the convexity self-learning size gear optimization; Among them, swarm intelligence algorithms refer to group-based optimization algorithms that can find the optimal solution within a specified range. They use the loss function constructed in step S3 as the evaluation function.

[0044] The process of optimizing size ranges using swarm intelligence algorithms includes: determining the optimization target, determining the iteration range, and ensuring that the total number of intervals included in the size range remains constant after each iteration.

[0045] The optimization targets refer to the number of intervals and the number of gears contained in each gear.

[0046] The iteration range refers to the upper and lower limits of the number of intervals included in each gear (i.e., gear size threshold) and the upper and lower limits of the total number of gears (i.e., gear number threshold). The gear size threshold is determined based on expert experience; the gear number threshold is determined by the physical upper limit of the number and the original number of gears. The determination of both thresholds should be based on ensuring the differentiation effect of the self-learning value and minimizing gear transitions to maintain the stability of the plate shape quality. This is because when the number of gears is small, the differentiation effect is poor, and the stability of the self-learning value cannot be guaranteed; while increasing the number of gears may lead to more gear transitions in actual production. Gear transitions mean changes in long-term self-learning, and due to the current characteristics of the long-term and short-term coordination of self-learning, an increase in gear transitions will be detrimental to the stability of plate shape control. It should be noted that the gear size threshold will be used for swarm intelligence algorithms, and the gear number threshold will be used for traversal loops. That is, the method will find the optimal solution under different numbers of gears, and finally determine the best gear division method by comparing the evaluation values ​​(loss function calculation values).

[0047] Ensuring that the total number of gear positions within a given range remains constant after each iteration is achieved through a self-check of the gear position information. This self-check includes proportional scaling and fine-tuning.

[0048] Based on the above, in this embodiment, S4 includes the following steps: S41, Determine the threshold for the number of gears and the threshold for the number of intervals contained in a single gear; S42, Based on the threshold of the number of gears, determine all possible values ​​of the number of gears and form a set of the number of gears; S43, traverse the set of gear numbers; S44. For each gear number traversed, based on the loss function, under the constraint of the threshold number of intervals contained in a single gear number, the optimal solution of the number of intervals contained in each gear number under the current gear number is solved iteratively using a swarm intelligence algorithm until the set of gear numbers is traversed and the optimal solution of the number of intervals contained in each gear number under different gear numbers is obtained. Furthermore, the step of using a swarm intelligence algorithm to iteratively solve for the optimal solution of the number of intervals contained in each gear under the current gear position includes: S441, represent the result of each iteration as a gear size vector: [ x 1 ,x 2 ,… , x n ];in, x i Indicates the first i The number of intervals contained in each gear. i =1,2,…, n , n Indicates the current gear number; S442, for x 1 ,x 2 ,… , x n The value is adjusted to satisfy: x 1 ,x 2 ,… , x n The sum equals the total number of intervals divided by the effective target size range, resulting in the adjusted... x 1 ,x 2 ,… , x n The value; Furthermore, the aforementioned x 1 ,x 2 ,… , x n The value is adjusted to satisfy: x 1 ,x 2 ,… , x n The sum equals the total number of intervals divided by the effective target size range, including: S4421, judgment x 1 ,x 2 ,… , x n If the sum of the values ​​is not equal to the total number of intervals divided by the effective target size range, then... x 1 ,x 2,… , x n The value is scaled proportionally, and the scaling result is rounded to the nearest integer to obtain the initially adjusted value. x 1 ,x 2 ,… , x n The value; where the scaling ratio is the sum of the total number of intervals divided by the effective target size range and the value of the scaling ratio. x 1 ,x 2 ,… , x n The ratio of the sum to the sum; S4422, if the initial adjustment x 1 ,x 2 ,… , x n The value still does not meet the requirements: x 1 ,x 2 ,… , x n The sum equals the total number of intervals divided by the effective target size range; then calculate the preliminary adjusted... x 1 ,x 2 ,… , x n The difference between the sum of the values ​​and the total number of intervals divided by the effective target size range; if the difference is greater than 0, then the initially adjusted... x 1 ,x 2 ,… , x n Subtract the difference from the maximum value in the range. If the difference is less than 0, then the initially adjusted value will be... x 1 ,x 2 ,… , x n The minimum value in the range is added to the difference to obtain the final adjusted value. x 1 ,x 2 ,… , x n The value of .

[0049] S443, the effective target size range is divided into n There are several gear levels; among them, the number of intervals contained in each gear level and the adjusted... x 1 ,x 2 ,… , x n The values ​​correspond one-to-one, and the divided gears are used as the current iteration result; S444, calculate the loss value of each level in the current iteration result and sum the calculation results to obtain the loss value of the current iteration result; wherein, the loss value of each level is calculated by the loss function. S445, determine the iteration result with the minimum loss value, and take the number of intervals contained in each gear as the optimal solution for the number of intervals contained in each gear under the current gear number.

[0050] S45: Select the optimal solution with the smallest loss value from the optimal solutions for the number of intervals contained in each gear under different gear numbers, and take the corresponding number of gears and the number of intervals contained in each gear under this number of gears as the gear optimization result.

[0051] Furthermore, after using swarm intelligence algorithms to find the optimal solution for the optimization object, the method also includes: Suppose that the optimal solution of the optimization object corresponds to the number of gears, s; The effective target size range is divided into s levels; wherein, the number of intervals contained in each level corresponds one-to-one with the number of intervals contained in each level corresponding to the optimal solution of the optimization object. The level is used as the optimized level to obtain the value range of the target size corresponding to each level.

[0052] In summary, this embodiment provides a scientifically feasible solution for optimizing convexity self-learning dimension levels, improving the scientific reliability of the optimization work, enhancing optimization efficiency, and strengthening the performance of the self-learning scheme in terms of accuracy and stability. Ultimately, it can effectively improve the shape control effect and greatly improve the efficiency of shape control optimization.

[0053] Second Embodiment

[0054] This embodiment uses width dimension level optimization as an example to illustrate the implementation process and final effect of the method of the present invention. Specifically, the process of achieving width dimension level optimization through the method of the present invention is as follows: S1. Collect and integrate the data required for gear optimization, including product target size specifications, crown calculation residuals, etc. This embodiment collects a total of 100,286 data points from a production line, some of which are shown below: Table 1. Partial Data from a Production Line

[0055] Based on the data in Table 1 above, the convexity calculation residual for each strip can be easily calculated using the formula for calculating the convexity calculation residual. For example, the convexity calculation residual corresponding to the first data in Table 1 is: .

[0056] After integration and processing, the dataset required for gear optimization is obtained, a portion of which is shown below: Table 2. Data required for gear optimization (partial)

[0057] S2: Determine the effective size range and the minimum interval for gear optimization.

[0058] Based on the on-site product outline, this embodiment obtains an effective width range of [1000mm, 2000mm]. This effective width range is then divided into 100 equal parts, and the minimum interval length for gear optimization is determined to be 10mm. The mean residual value and the production quantity ratio within each interval are then calculated for use in subsequent optimization processes.

[0059] Due to space limitations, some statistical information for each interval is shown in the table below.

[0060] Table 3. Interval Statistics (Partial)

[0061] S3: Design a loss function to evaluate the rationality of size gradation.

[0062] Considering the influencing factors of both intra-class and inter-class differences, and combining four basic characterization indicators—variance of multi-interval residuals, mean of multi-interval residuals, variance of multi-interval proportion, and mean of multi-interval proportion—the loss function is designed as follows:

[0063] in, Loss Represents the total loss; Loss in The residuals represent the loss caused by in-file differences, with weights considered. W e And the proportion takes into account weight W p Each is taken as 0.5, and the expression is as follows:

[0064] in, Loss out This represents the loss caused by differences between grade levels, and no gain processing is applied; it corresponds to the coefficient. W g Take 1; B Pick Loss in Maximum value 0.25; residuals take weight. W e And the proportion takes into account weight W p and Loss in The values ​​are the same, and the expressions are as follows:

[0065] S4: Optimize size ranges using swarm intelligence algorithms.

[0066] Swarm intelligence algorithm: Particle swarm optimization algorithm is selected as the solution method.

[0067] Determine the optimization targets: the number of intervals and the number of gears contained in each gear; Determine the iteration range: Based on expert experience, and taking into account the gear differentiation effect and production stability requirements, the gear size was finally determined to be [5,25]. The original number of width gears before optimization was 5, and the self-learning system limited the upper limit of the number of width gears to 10, so the number of gears was finally determined to be [5,10].

[0068] Gear Information Self-Check: During iterative optimization, before each evaluation of the rationality of the division, the total number of intervals included in the gear is checked and corresponding preprocessing is performed. Taking a gear of 5 as an example, after a certain iteration of the particle swarm optimization algorithm, the vector composed of the number of intervals included in each gear (hereinafter referred to as: gear size vector) is: [20,22,24,25,20]. The sum of its components is 111, which is greater than the actual total number of intervals of 100, resulting in an excess of 11. First, each component is scaled proportionally, with the ratio being the ratio of the actual total number of intervals to the sum of the current components, i.e., 100 / 111. The scaling result is rounded to the nearest integer, so the scaled gear size vector is: [18,20,22,23,18]. At this time, the sum of the components is 101, which is still greater than the actual total number of intervals of 100, resulting in an excess of 1. Then, a self-check and fine-tuning is performed, selecting the current maximum component 23. The margin relative to the lower limit of gear size 5 is 18, which is greater than the excess, so it is adjusted downward to 22. It should be noted that this part involves fine-tuning after proportional scaling, aiming to ensure that the total number of intervals after iteration is constant with the actual number of intervals, thus satisfying the actual situation. The self-checking design of this invention, which first scales proportionally and then fine-tunes, theoretically avoids the so-called "less than excessive" situation, which can be disregarded. If it occurs, then the largest component at that time is selected again for similar adjustment until the condition of a constant number of intervals is met. After the gear information self-check, the actual gear size vector used for loss function calculation in this iteration is: [18, 20, 22, 22, 18]. The process can be referred to in the table below: Table 4 Comparison of gear information before and after self-check

[0069] Similarly, after a certain iteration, the gear size vector is [18,17,19,23,16]. Its gear information self-check process, specifically the scaling step, is similar to the previous example. The difference is that before the self-check fine-tuning loop, the sum of the components is 99, which is less than 100, indicating a missing component of 1. Therefore, during the self-check fine-tuning, the current minimum component 17 is selected. The margin relative to the gear size upper limit of 25 is 8, which is greater than the missing component, so it is adjusted upwards to 18. Therefore, the actual gear size vector used for loss function calculation is [19,18,20,25,18]. The process can be seen in the table below. Other cases are similar and will not be elaborated further.

[0070] Table 5 Comparison of gear information before and after self-check

[0071] Rationality Evaluation: Using the loss function designed in step S3 as the evaluation function, the gear information after self-checking is input into the calculation of the corresponding loss value to complete the rationality evaluation of this (iteration step). For example: After a certain iteration, the gear information input into the evaluation function is [25,6,22,7,25,15]. Taking gears 3 and 6 as examples, the loss value calculation process is explained. First, the width range and corresponding statistical information of each gear are shown in the following table: Table 6. Width range and corresponding statistics for each gear level

[0072] The loss caused by the intra-gear difference corresponding to Gear 3 is:

[0073] The loss caused by the difference between gear levels corresponding to level 3 is:

[0074] Therefore, the total loss corresponding to level 3 is:

[0075] The loss caused by the difference within the 6th gear is:

[0076] The loss caused by the gear difference corresponding to gear 6 is:

[0077] Therefore, the total loss corresponding to level 6 is:

[0078] By calculating and organizing using the above methods, the loss information corresponding to each gear in this iteration can be obtained as shown in the following table: Table 7 Loss Information for Each Gear

[0079] Therefore, the total loss value for each gear in this iteration is 1.2858.

[0080] Optimization Logic Overview: The Particle Swarm Optimization (PSO) algorithm primarily achieves information sharing and collective optimization by having different individuals in the population adjust their movement direction by tracking their individual and group historical best positions. After determining the population parameters (as shown in Table 8, where individual dimensions 5-10 represent all cases with traversal widths from 5 to 10; for detailed procedures, please refer to...), the algorithm... Figure 2 The process initializes the positions of all individuals, then determines whether the termination condition has been met, and decides whether to proceed to the next iteration. In this embodiment, the termination condition for a single optimization is reaching a specified number of iterations (150 steps). To ensure optimal results, this embodiment selects multiple optimizations in a single round (each round corresponds to one individual dimension) to obtain the best result (10 times). Each iteration step will complete five steps in sequence: updating individual positions and speeds, self-checking gear information, calculating the evaluation values ​​of all individuals, updating the historical best evaluation values ​​and positions of individuals, and updating the historical best evaluation values ​​and positions of the group, before re-entering the determination process. This cycle repeats until the iteration ends.

[0081] It should be noted that since the particle swarm optimization algorithm is relatively mature and this method does not involve the development or improvement of the algorithm itself, it will not be elaborated here.

[0082] Table 8 Population Parameters

[0083] The optimized results are compared with those before optimization in the table below. For details, please refer to the table. Figure 3 , Figure 4 : Table 9 Comparison of gear shifting before and after optimization

[0084] Based on the above results, the long-term self-learning width settings for convexity were optimized. The optimized width settings (1350mm-1500mm) corresponded to the original width settings (1400mm-1600mm), with the mean square of the residuals decreasing by approximately 38.6%, indicating that the dispersion level of the residuals within the optimized settings was more consistent. After a period of observation and statistics, the convexity control effect improved for all width ranges. For products with widths above 1650mm, the convexity hit rate of the corresponding strip increased by more than 2.7% after optimization compared to before optimization, thus proving the scientific effectiveness of the convexity size setting optimization method provided in this invention.

[0085] Third Embodiment

[0086] This embodiment provides a convexity self-learning size level optimization system, including the following modules: The data acquisition and integration module is used to collect actual production data within a preset production cycle at the production site and integrate the data required for gear optimization. The data processing module is used to determine the effective target size range of the strip, divide the effective target size range into multiple intervals, and calculate the feature data corresponding to each interval. The gear optimization module is used for: Construct a loss function to evaluate the rationality of size segmentation; Taking the number of gears and the number of intervals contained in each gear as the optimization objects, and based on the loss function, the optimal solution of the optimization objects is solved by swarm intelligence algorithm to realize the self-learning size gear optimization of convexity.

[0087] It should be noted that the convexity self-learning size level optimization system of this embodiment corresponds to the convexity self-learning size level optimization method of the first embodiment described above; the functions implemented by each functional module in the convexity self-learning size level optimization system of this embodiment correspond one-to-one with the process steps in the convexity self-learning size level optimization method of the first embodiment described above; therefore, they will not be described again here.

[0088] Fourth embodiment

[0089] This embodiment provides an electronic device, such as... Figure 5 As shown, the electronic device includes a processor and a memory; wherein the processor and the memory can be connected via a communication bus; the memory stores at least one instruction, which is loaded and executed by the processor to implement the method of the first embodiment described above. Furthermore, the electronic device may also include a transceiver, the processor and the transceiver can be connected via a communication bus, and the transceiver is used to communicate with other devices.

[0090] Below, in conjunction with Figure 5A detailed introduction to each component of this electronic device is provided below: The processor is the control center of the electronic device. The electronic device may include multiple processors, each of which can be a single-core processor (single-CPU) or a multi-core processor (multi-CPU). The term "processor" can refer to a single processor or a collective term for multiple processing elements. For example, a processor can be one or more central processing units (CPUs), other general-purpose processors, application-specific integrated circuits (ASICs), or one or more integrated circuits configured to implement embodiments of the present invention, such as one or more digital signal processors (DSPs), one or more field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. A general-purpose processor can be a microprocessor or any conventional processor. The processor can perform various functions of the electronic device by running or executing software programs stored in memory and by calling data stored in memory.

[0091] In a specific implementation, as one example, the processor may include one or more CPUs, for example... Figure 5 CPU0 and CPU1 shown are, of course, merely illustrative examples.

[0092] The memory is used to store the software program that executes the solution of the present invention, and the processor controls its execution. For specific implementation methods, please refer to the above method embodiments, which will not be repeated here.

[0093] Optionally, the memory may be a read-only memory (ROM) or other type of static storage device capable of storing static information and instructions, random access memory (RAM) or other type of dynamic storage device capable of storing information and instructions, or electrically erasable programmable read-only memory (EEPROM), compact disc read-only memory (CD-ROM) or other optical disc storage, optical disc storage (including compressed optical discs, laser discs, optical discs, digital universal optical discs, Blu-ray discs, etc.), magnetic disk storage media or other magnetic storage devices, or any other medium capable of carrying or storing desired program code in the form of instructions or data structures and accessible by a computer, but not limited thereto. The memory may be integrated with the processor or exist independently, and may be accessed through the interface circuit of the electronic device ( Figure 5 (Not shown in the image) is coupled to the processor; however, this embodiment of the invention does not impose specific limitations on this.

[0094] The transceiver may include a receiver and a transmitter. Figure 5 (Not shown separately). The receiver is used to implement the receiving function, and the transmitter is used to implement the transmitting function. The transceiver can be integrated with the processor or exist independently, and can be connected through the interface circuit of the electronic device (…). Figure 5 (Not shown in the image) is coupled to the processor, and this embodiment of the invention does not specifically limit this.

[0095] In addition, it should be noted that, Figure 5 The structure of the electronic device shown is not intended to limit the device. Actual devices may include more or fewer components than shown, or combine certain components, or have different component arrangements. Furthermore, the technical effects achieved by this electronic device when performing the method of the first embodiment described above can be referenced to the technical effects described in the first embodiment; therefore, they will not be repeated here.

[0096] Fifth Embodiment

[0097] This embodiment provides a computer-readable storage medium storing at least one instruction, which is loaded and executed by a processor to implement the method of the first embodiment described above. The computer-readable storage medium may be a ROM, random access memory, CD-ROM, magnetic tape, floppy disk, or optical data storage device, etc. The instruction stored therein can be loaded and executed by a processor in a terminal.

[0098] Furthermore, it should be noted that the present invention can be provided as a method, apparatus, or computer program product. Therefore, embodiments of the present invention can take the form of a completely or partially hardware embodiment, a completely or partially software embodiment, or an embodiment combining software and hardware aspects. Moreover, when implemented in software, embodiments of the present invention can take the form of a computer program product implemented on one or more computer-usable storage media containing computer-usable program code. The computer program product includes one or more computer instructions or computer programs. When the computer instructions or computer program are loaded or executed on a computer, all or part of the processes or functions described in the embodiments of the present invention are generated. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions can be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another. For example, the computer instructions can be transmitted from one website, computer, server, or data center to another website, computer, server, or data center via wired (e.g., infrared, wireless, microwave, etc.) means. The computer-readable storage medium can be any usable medium accessible to a computer or a data storage device such as a server or data center containing one or more sets of usable media. The available medium can be a magnetic medium (e.g., floppy disk, hard disk, magnetic tape), an optical medium (e.g., DVD), or a semiconductor medium. A semiconductor medium can be a solid-state drive (SSD).

[0099] Embodiments of the present invention are described with reference to flowchart illustrations and / or block diagrams of methods, terminal devices (systems), and computer program products according to embodiments of the invention. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, embedded processor, or other programmable data processing terminal device to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing terminal device, generate instructions for implementing the flowchart illustrations. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0100] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing terminal device to operate in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1The functions specified in one or more boxes. These computer program instructions may also be loaded onto a computer or other programmable data processing terminal equipment to cause a series of operational steps to be performed on the computer or other programmable terminal equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable terminal equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0101] It should also be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. The terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or terminal device that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or terminal device. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or terminal device that includes said element. Furthermore, the term "and / or" is merely a description of the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can represent: A alone, A and B simultaneously, and B alone, where A and B can be singular or plural. Additionally, the character " / " in this text generally indicates an "or" relationship between the preceding and following objects, but it can also indicate an "AND / OR" relationship. Please refer to the context for specific interpretations. "At least one" refers to one or more items, while "more than" refers to two or more items. "At least one of the following" or similar expressions refer to any combination of these items, including any combination of single or multiple items. For example, at least one of a, b, or c can be represented as: a, b, c, ab, ac, bc, or abc, where a, b, and c can be single or multiple.

[0102] Furthermore, it is understood that in various embodiments of the present invention, the order of the above-mentioned process numbers does not imply the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of the present invention.

[0103] Those skilled in the art will recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementations should not be considered beyond the scope of this invention.

[0104] In the several embodiments provided by this invention, it should be understood that the disclosed devices, apparatuses, and methods can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative. For instance, the division of functional modules / units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another device, or some features may be ignored or not executed. Furthermore, the shown or discussed mutual couplings or direct couplings or communication connections may be through some interfaces; indirect couplings or communication connections between devices or units may be electrical, mechanical, or other forms. Units described as separate components may or may not be physically separate, and components shown as units may or may not be physical units, i.e., they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs. Additionally, the functional units in the various embodiments of this invention may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit.

[0105] If the method is implemented as 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 invention, or the part that contributes to the prior art, or a part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0106] Finally, it should be noted that the above description is merely a preferred embodiment of the present invention. It should be pointed out that although preferred embodiments of the present invention have been described, those skilled in the art, once they understand the basic inventive concept of the present invention, can make several improvements and modifications without departing from the principles described herein. These improvements and modifications should also be considered within the scope of protection of the present invention. Therefore, the appended claims are intended to be interpreted as including the preferred embodiments as well as all changes and modifications falling within the scope of the embodiments of the present invention.

Claims

1. A method for optimizing convexity self-learning dimension levels, characterized in that, include: Collect actual production data from the production site within the preset production cycle, and integrate the data required for gear optimization. Determine the effective target size range of the strip, divide the effective target size range into multiple intervals, and calculate the feature data corresponding to each interval; Construct a loss function to evaluate the rationality of size segmentation; Taking the number of gears and the number of intervals contained in each gear as the optimization objects, and based on the loss function, the optimal solution of the optimization objects is solved by swarm intelligence algorithm to realize the self-learning size gear optimization of convexity; The optimization targets are the number of gear positions and the number of intervals contained in each gear position. Based on the loss function, a swarm intelligence algorithm is used to iteratively solve for the optimal solution of the optimization targets, thereby achieving convexity self-learning size gear position optimization, including: Determine the threshold for the number of gears and the threshold for the number of intervals contained in a single gear; Based on the threshold of the number of gears, determine all possible values ​​of the number of gears and form a set of gear numbers; Iterate through the set of gear numbers; For each gear number traversed, based on the loss function, under the constraint of the threshold number of intervals contained in a single gear number, the optimal solution for the number of intervals contained in each gear number under the current gear number is iteratively solved using a swarm intelligence algorithm, until the set of gear numbers is traversed and the optimal solution for the number of intervals contained in each gear number under different gear numbers is obtained. Among the optimal solutions for the number of intervals contained in each gear under different gear numbers, select the optimal solution with the minimum loss value, and take the corresponding gear number and the number of intervals contained in each gear under this gear number as the gear optimization result. The method of using swarm intelligence algorithms to iteratively solve for the optimal solution of the number of intervals contained in each gear under the current gear position includes: Represent the result of each iteration as a gear size vector: [x1, x2, ..., x n ]; where x i This represents the number of intervals contained in the i-th gear, where i = 1, 2, ..., n, and n represents the number of gears in the current gear. For x1,x2,…,x n The values ​​are adjusted to satisfy: x1, x2, ..., x n The sum of these values ​​equals the total number of intervals divided by the effective target size range, resulting in the adjusted x1, x2, ..., x n The value; The effective target size range is divided into n levels; wherein, the number of intervals contained in each level is equal to the adjusted x1, x2, ..., x n The values ​​correspond one-to-one, and the divided gears are used as the current iteration result; Calculate the loss value for each level in the current iteration result and sum the calculation results to obtain the loss value of the current iteration result; wherein, the loss value for each level is calculated using the loss function. The iteration result with the minimum loss value is determined, and the number of intervals contained in each gear is taken as the optimal solution for the number of intervals contained in each gear under the current gear number.

2. The convexity self-learning size level optimization method as described in claim 1, characterized in that, The actual production data includes the model-calculated values ​​of the target thickness, target width, and convexity of each strip, the measured values ​​of the convexity, the long-term self-learning value of the convexity, and the short-term self-learning value of the convexity. The integrated data required for gear optimization includes: Calculate the total number of strips, M, produced within the preset production cycle at the production site; Calculate the target dimensions for each strip; where the target dimensions are the target thickness or target width. Calculate the convexity calculation residual for each strip; wherein the convexity calculation residual is calculated based on the model calculation value of convexity, the measured value of convexity, the long-term self-learning value of convexity, and the short-term self-learning value of convexity.

3. The convexity self-learning size level optimization method as described in claim 2, characterized in that, The formula for calculating the residual of the convexity calculation is as follows: ； in, Calculate the residual for convexity; This represents the measured value of the convexity. The calculated value for the convexity model; The sum of the long-term self-learning value and the short-term self-learning value of convexity.

4. The convexity self-learning size range optimization method as described in claim 2, characterized in that, The effective target size range is a range consisting of the minimum and maximum values ​​of the target size of the strip as specified in the product outline; The calculation of the feature data corresponding to each interval includes: Calculate the total number of intervals divided by the effective target size range; For each interval, calculate the mean value of the convexity calculation residuals of all strips whose target size falls within its range, and record it as the mean residual value within the interval. For each interval, the ratio of the number of strips with the target size falling within its range to M is calculated and recorded as the interval production quantity percentage.

5. The convexity self-learning size range optimization method as described in claim 4, characterized in that, The expression for the loss function is: ； in, This represents the calculated loss value; This indicates the loss caused by differences within the file; This indicates the loss caused by differences between grades; , and As preset weight values, the sum of the two is 1. This represents the variance of the mean residuals within all intervals corresponding to the ranges whose loss value is to be calculated. This represents the variance of the production quantity percentage of all intervals included in the tier for which the loss value is to be calculated. B represents the baseline value, which takes the value of The maximum value of W, where e represents the base of the natural number. g Indicates the gain coefficient. This indicates the gear at which the loss value is to be calculated and the gear between adjacent gears. The absolute gap This indicates the gear at which the loss value is to be calculated and the gear adjacent to it. The absolute gap.

6. The convexity self-learning size level optimization method as described in claim 1, characterized in that, The terms x1, x2, ..., x n The values ​​are adjusted to satisfy: x1, x2, ..., x n The sum equals the total number of intervals divided by the effective target size range, including: Determine the relationship between x1, x2, ..., x n If the sum of x1, x2, ..., x2 is not equal to the total number of intervals divided by the effective target size range, then for x1, x2, ... ... n The values ​​are scaled proportionally, and the scaling result is rounded to the nearest integer to obtain the initially adjusted x1, x2, ..., x n The value of ; where the scaling ratio is the sum of the total number of intervals divided by the effective target size range and x1, x2, ..., x n The ratio of the sum to the sum; If the initial adjusted x1, x2, ..., x n The values ​​still do not satisfy: x1, x2, ..., x n The sum of these values ​​equals the total number of intervals divided by the effective target size range; then calculate the initially adjusted x1, x2, ..., x... n The difference between the sum of x1, x2, ..., x2 and the total number of intervals divided by the effective target size range. If the difference is greater than 0, then the initially adjusted x1, x2, ..., x2 will be... n Subtract the difference from the maximum value in the range. If the difference is less than 0, then the initially adjusted x1, x2, ..., x... n The minimum value in the range is added to the difference to obtain the final adjusted x1, x2, ..., x. n The value of .

7. The convexity self-learning size range optimization method as described in claim 1, characterized in that, After using swarm intelligence algorithms to find the optimal solution for the optimization object, the method further includes: Suppose that the optimal solution of the optimization object corresponds to the number of gears, s; The effective target size range is divided into s levels; wherein, the number of intervals contained in each level corresponds one-to-one with the number of intervals contained in each level corresponding to the optimal solution of the optimization object. The level is used as the optimized level to obtain the value range of the target size corresponding to each level.

8. A convexity self-learning size range optimization system, characterized in that, include: The data acquisition and integration module is used to collect actual production data within a preset production cycle at the production site and integrate the data required for gear optimization. The data processing module is used to determine the effective target size range of the strip, divide the effective target size range into multiple intervals, and calculate the feature data corresponding to each interval. The gear optimization module is used for: Construct a loss function to evaluate the rationality of size segmentation; Taking the number of gears and the number of intervals contained in each gear as the optimization objects, and based on the loss function, the optimal solution of the optimization objects is solved by swarm intelligence algorithm to realize the self-learning size gear optimization of convexity; Taking the number of gear positions and the number of intervals contained in each gear position as the optimization object, and based on the aforementioned loss function, a swarm intelligence algorithm is used to iteratively solve for the optimal solution of the optimization object, thereby achieving convexity self-learning size gear position optimization, including: Determine the threshold for the number of gears and the threshold for the number of intervals contained in a single gear; Based on the threshold of the number of gears, determine all possible values ​​of the number of gears and form a set of gear numbers; Iterate through the set of gear numbers; For each gear number traversed, based on the loss function, under the constraint of the threshold number of intervals contained in a single gear number, the optimal solution for the number of intervals contained in each gear number under the current gear number is iteratively solved using a swarm intelligence algorithm, until the set of gear numbers is traversed and the optimal solution for the number of intervals contained in each gear number under different gear numbers is obtained. Among the optimal solutions for the number of intervals contained in each gear under different gear numbers, select the optimal solution with the minimum loss value, and take the corresponding gear number and the number of intervals contained in each gear under this gear number as the gear optimization result. The method of using swarm intelligence algorithms to iteratively solve for the optimal solution of the number of intervals contained in each gear under the current gear position includes: Represent the result of each iteration as a gear size vector: [x1, x2, ..., x n ]; where x i This represents the number of intervals contained in the i-th gear, where i = 1, 2, ..., n, and n represents the number of gears in the current gear. For x1,x2,…,x n The values ​​are adjusted to satisfy: x1, x2, ..., x n The sum of these values ​​equals the total number of intervals divided by the effective target size range, resulting in the adjusted x1, x2, ..., x n The value; The effective target size range is divided into n levels; wherein, the number of intervals contained in each level is equal to the adjusted x1, x2, ..., x n The values ​​correspond one-to-one, and the divided gears are used as the current iteration result; Calculate the loss value for each level in the current iteration result and sum the calculation results to obtain the loss value of the current iteration result; wherein, the loss value for each level is calculated using the loss function. The iteration result with the minimum loss value is determined, and the number of intervals contained in each gear is taken as the optimal solution for the number of intervals contained in each gear under the current gear number.

Images