A method for controlling the width of the roughing zone of hot-rolled transition material

Abstract

This invention relates to a method for controlling the width of the roughing zone of hot-rolled transition steel. The method includes the following steps: S1, determining whether a strip is a transition steel based on its classification changes, and assigning initial values ​​to the long-term learning coefficients; S2, adjusting the self-learning pace step and confidence level based on the information change; S3, adjusting the width of the strip based on the long-term learning to complete the strip width design; S4, performing short-term self-learning calculations based on the information of the preceding and following strips, and modifying the long-term learning coefficients; S5, performing width correction compensation based on the rolling deviation of each pass. This technical solution, by utilizing the rolling characteristics and planning features of transition steel in the roughing zone, and through a model width setting method, adaptive learning method, and specific width correction method, reduces operational intervention and achieves high-quality rolling of transition steel.

Description

Technical Field

[0001] This invention relates to a control method, specifically a method for controlling the width of the roughing zone of hot-rolled transition material, belonging to the field of model production control technology for hot-rolled roughing. Background Technology

[0002] During hot rolling production, due to changes in market contracts, transitional strips are needed to bridge the gap between different contracts, taking into account variations in rolling conditions such as temperature and specifications. These strips typically have short planned lengths, and during rolling, variations in site conditions and control requirements can easily lead to width discrepancies, either exceeding the set width or falling too narrow. Furthermore, ordinary adaptive correction functions cannot achieve short-term width correction, impacting hot rolling yield and production stability. Before the implementation of this solution, the rolling of this type of steel was generally done manually by operators, which resulted in the following problems: firstly, control delays, as operators could only intervene in the next strip after the previous one was finished; secondly, intervention errors, as the correction could not identify whether consecutive strips were rolled in the same batch, and errors could occur due to changes in classification. Therefore, a new solution is urgently needed to address these technical problems. Summary of the Invention

[0003] This invention addresses the problems existing in the prior art by providing a method for controlling the width of the roughing zone of hot-rolled transition materials. This technical solution reduces operational intervention and achieves high-quality rolling of transition materials by utilizing the rolling characteristics and planning properties of transition materials in the roughing zone, through a model width setting method, an adaptive learning method, and a specific width correction method.

[0004] To achieve the above objectives, the technical solution of the present invention is as follows: a method for controlling the width of the roughing zone of hot-rolled transition material, the method comprising the following steps:

[0005] S1. Based on the changes in strip steel classification, determine whether it is a transitional material, and assign initial values ​​to the long-term learning coefficients.

[0006] S2. Adjust the self-learning pace and confidence level based on the amount of information change.

[0007] S3. Based on long-term learning, adjust the width of the strip to complete the strip width design.

[0008] S4. Perform short-time self-learning calculations based on the information from the front and rear strips, and modify the long-time learning coefficients.

[0009] S5. Based on the rolling deviation of each pass, perform width correction compensation.

[0010] In step S1, based on the changes in strip steel classification, it is determined whether the material is a transitional material, and initial values ​​are assigned to the long-term learning coefficients, as detailed below:

[0011] To determine whether a strip is a transitional material, the hot rolling control system can be used to identify variations in strip classification, primarily focusing on large deviations in width, thickness, final rolling temperature, and type classification, as shown in Table 1.

[0012]

[0013]

[0014] Table 1: Strip Steel Classification

[0015] When determining the transition material, the classification of steel grades is compared with the classification of the previous strip steel. For each change, +1 is added, and the change is accumulated. When the accumulated change value SC exceeds 2, it is considered to be the transition material.

[0016] When controlling transition strip production, long-term and short-term genetic learning coefficients are introduced. Since transition strip production is interspersed between various plans, with irregular and lengthy intervals, the traditional method of using the adaptive coefficient of the previous strip for step-by-step control is insufficient for self-learning control. Long-term learning is introduced by saving this type of strip according to the previously defined classification criteria, which can then be directly referenced in later production. The purpose of introducing short-term genetic coefficients is for temporary self-learning during continuous production of transition strips. Through short-term real-time modifications, rapid correction and adjustment of the strip are ensured.

[0017] A long-term learning genetic coefficient table for the roughing rolling zone is plotted using classification information (see Table 2). During the first rolling pass, the initial long-term genetic coefficient, which is also the short-term learning coefficient of the strip, is defined as ranging from 0.9 to 1.18 based on on-site production conditions. To prevent narrow strip width from affecting contracts, an initial short-term genetic coefficient of 1.014 is used based on experience. After one rolling pass, the genetic coefficient table stores the corresponding classification information and genetic coefficient; this genetic coefficient is then referred to as the long-term learning coefficient.

[0018] Steel Classification Width Classification Thickness classification Final rolling temperature classification Long-term genetic value Short-term genetic value

[0019] Table 2: Genetic Values ​​Table.

[0020] Specifically, S2 adjusts the self-learning pace (step) and confidence level based on the amount of information change, as detailed below.

[0021] During transition material rolling, due to the short planning time, a fast learning pace is required, thus the model's self-learning pace step needs to be increased. Furthermore, because the intervals between plans are relatively long, some control conditions on-site will change accordingly. Width control needs to consider both short-term and long-term learning, therefore, the long-term learning confidence level is increased. The learning confidence level is determined based on the cumulative value SC of information changes calculated by S1. When SC changes significantly, the long-term learning confidence level increases, and vice versa. Both values ​​are empirical values ​​summarized from on-site rolling, as shown in Table 3 below.

[0022] Rolling interval time Cumulative value of information change SC Self-learning rhythm step confidence level delta >10 days >＝3 0.8 0.9 >3 days >＝2 1.1 0.7 >1 day <2 1.4 0.5

[0023] Table 3: Self-learning pace step and confidence delta configuration table

[0024] The model's self-learning pace (step) adjusts the learning rhythm based on the control method. When the accumulated information change value (SC) is large, the self-learning pace is increased to avoid the problem of insufficient learning due to too short a plan. The confidence level (delta) represents the confidence level between long-term and short-term learning. When the rolling interval is long, the confidence level of long-term learning increases, ensuring the accuracy of width control.

[0025] Specifically, S3, based on long-term learning, adjusts the width of the strip to complete the strip width design, as detailed below.

[0026] The learning coefficient of the strip is calculated by the deviation between the actual width and the calculated width of the strip in the roughing zone. It reflects the width deviation of the strip in the roughing zone and can be used as an important reference for width adjustment in the next rolling.

[0027] Adjust the formula to Formula 1:

[0028] Roughing target width = Roughing target width - [Roughing long-term learning value * confidence delta + Roughing short-term learning value * (1 - delta)] * self-learning pace step; Formula 1;

[0029] Specifically, S4 involves performing short-time self-learning calculations based on information from the preceding and following strips, and then modifying the long-time learning coefficients, as detailed below.

[0030] This step involves updating long-term and short-term self-learning. The self-learning coefficient uses an initial value during the first rolling, and needs to be continuously updated as rolling continues. The update method for short-term learning is to adjust accordingly based on the deviation between the actual width of the strip and the set width. The adjustment method is the same as that for ordinary strip production and will not be described in detail here.

[0031] In the production process of strip steel of the same category, long-term learning is updated based on short-term learning. That is, in the production of the same category, long-term learning adopts the short-term learning of the previous strip steel. When changing categories, long-term learning retains the short-term learning value after the last rolled strip until the next time this category of steel is rolled, and then it is updated.

[0032] Specifically, S5 involves width correction compensation based on the rolling deviation of each pass, as detailed below.

[0033] According to industry surveys, existing hot rolling production processes all utilize reversible mills in the roughing zone. For transition materials, the roughing model incorporates a feature that adjusts for width based on rolling deviations between each pass, moving these deviations to the next pass for width correction. Deviations are gradually eliminated starting from each pass.

[0034] S51. During each rolling pass, the model performs an inter-pass calculation to calculate a target width based on the changes in parameters such as on-site equipment and processes.

[0035] S52. Obtain the average actual width of the model by using the width detector produced on site. Based on the inspection, remove the width anomaly of 15m at the head.

[0036] S53. Obtain the deviation between the setting and the actual value for each pass, and use it in the previously calculated target width value for each pass, see Formula 2. When it is the last pass, the target width between passes is the final roughing target width.

[0037] Target width between tracks = Original target width of track + Track deviation width; Formula 2.

[0038] Compared to existing technologies, this invention has the following advantages: This solution addresses the quality improvement of transition materials in hot-rolled production, a persistent challenge in hot-rolled process control. Existing width control models rely on continuous self-adjustment after repeated rolling to achieve stable quality. However, transition materials have short production cycles and significant specification variations, making high-precision control difficult to achieve using existing technologies. This method utilizes long-term learning, self-learning rhythm adjustment, and confidence levels to inherit and update width learning values. It also leverages the characteristics of inter-pass settings for width compensation and correction, effectively solving the problem of excessively wide transition materials.

[0039] The technical advantages of this method are as follows:

[0040] 1. Classify the strip steel according to the specifications on site, and use the change in specification classification as the criterion for judging transition material and non-transition material. Through this step, the model can be set up in a fully automatic way. Compared with manual judgment, it has a lower error rate and can be predicted in advance, which is convenient for model pre-setting.

[0041] 2. By adding long-term and short-term self-learning methods, and setting confidence levels and self-learning steps based on parameters such as classification changes and rolling intervals, both the rolling continuity of transition materials and rapid on-site response are ensured. This method can be applied to various manufacturing industries, and specific parameter settings can be tailored to the specific characteristics of each industry.

[0042] 3. For each additional rolling pass, width correction compensation is performed. This method can eliminate process deviations during rolling, such as measurement system deviations and equipment deviations, so as to achieve early control and early benefits, avoid the accumulation of deviations that cannot be corrected, and reduce subsequent control costs. Detailed Implementation

[0043] To enhance understanding of the present invention, the present invention will be described in detail below with reference to embodiments.

[0044] Example 1: A method for controlling the width of the roughing zone of hot-rolled transition material, the method comprising the following steps:

[0045] S1. Based on the changes in strip steel classification, determine whether it is a transitional material, and assign initial values ​​to the long-term learning coefficients.

[0046] S2. Adjust the self-learning pace and confidence level based on the amount of information change.

[0047] S3. Based on long-term learning, adjust the width of the strip to complete the strip width design.

[0048] S4. Perform short-time self-learning calculations based on the information from the front and rear strips, and modify the long-time learning coefficients.

[0049] S5. Based on the rolling deviation of each pass, perform width correction compensation.

[0050] In step S1, based on the changes in strip steel classification, it is determined whether the material is a transitional material, and initial values ​​are assigned to the long-term learning coefficients, as detailed below:

[0051] To determine whether a strip is a transitional material, the hot rolling control system can be used to identify variations in strip classification, primarily focusing on large deviations in width, thickness, final rolling temperature, and type classification, as shown in Table 1.

[0052] Steel Classification Width Classification Thickness classification Final rolling temperature classification Layer range 1-20 1-8 1-7 1-10

[0053] Table 1: Strip Steel Classification

[0054] When determining the transition material, the classification of steel grades is compared with the classification of the previous strip steel. For each change, +1 is added, and the change is accumulated. When the accumulated change value SC exceeds 2, it is considered to be the transition material.

[0055] When controlling transition strip production, long-term and short-term genetic learning coefficients are introduced. Since transition strip production is interspersed between various plans, with irregular and lengthy intervals, the traditional method of using the adaptive coefficient of the previous strip for step-by-step control is insufficient for self-learning control. Long-term learning is introduced by saving this type of strip according to the previously defined classification criteria, which can then be directly referenced in later production. The purpose of introducing short-term genetic coefficients is for temporary self-learning during continuous production of transition strips. Through short-term real-time modifications, rapid correction and adjustment of the strip are ensured.

[0056] A long-term learning genetic coefficient table for the roughing rolling zone is plotted using classification information (see Table 2). During the first rolling pass, the initial long-term genetic coefficient, which is also the short-term learning coefficient of the strip, is defined as ranging from 0.9 to 1.18 based on on-site production conditions. To prevent narrow strip width from affecting contracts, an initial short-term genetic coefficient of 1.014 is used based on experience. After one rolling pass, the genetic coefficient table stores the corresponding classification information and genetic coefficient; this genetic coefficient is then referred to as the long-term learning coefficient.

[0057] Steel Classification Width Classification Thickness classification Final rolling temperature classification Long-term genetic value Short-term genetic value

[0058] Table 2: Genetic Values ​​Table.

[0059] Specifically, S2 adjusts the self-learning pace (step) and confidence level based on the amount of information change, as detailed below.

[0060] During transition material rolling, due to the short planning time, a fast learning pace is required, thus the model's self-learning pace step needs to be increased. Furthermore, because the intervals between plans are relatively long, some control conditions on-site will change accordingly. Width control needs to consider both short-term and long-term learning, therefore, the long-term learning confidence level is increased. The learning confidence level is determined based on the cumulative value SC of information changes calculated by S1. When SC changes significantly, the long-term learning confidence level increases, and vice versa. Both values ​​are empirical values ​​summarized from on-site rolling, as shown in Table 3 below.

[0061] Rolling interval time Cumulative value of information change SC Self-learning rhythm step confidence level delta >10 days >＝3 0.8 0.9 >3 days >＝2 1.1 0.7 >1 day <2 1.4 0.5

[0062] Table 3: Self-learning pace step and confidence delta configuration table

[0063] The model's self-learning pace (step) adjusts the learning rhythm based on the control method. When the accumulated information change value (SC) is large, the self-learning pace is increased to avoid the problem of insufficient learning due to too short a plan. The confidence level (delta) represents the confidence level between long-term and short-term learning. When the rolling interval is long, the confidence level of long-term learning increases, ensuring the accuracy of width control.

[0064] Specifically, S3, based on long-term learning, adjusts the width of the strip to complete the strip width design, as detailed below.

[0065] The learning coefficient of the strip is calculated by the deviation between the actual width and the calculated width of the strip in the roughing zone. It reflects the width deviation of the strip in the roughing zone and can be used as an important reference for width adjustment in the next rolling.

[0066] Adjust the formula to Formula 1:

[0067] Roughing target width = Roughing target width - [Roughing long-term learning value * confidence delta + Roughing short-term learning value * (1 - delta)] * self-learning pace step; Formula 1;

[0068] Specifically, S4 involves performing short-time self-learning calculations based on information from the preceding and following strips, and then modifying the long-time learning coefficients, as detailed below.

[0069] This step involves updating long-term and short-term self-learning. The self-learning coefficient uses an initial value during the first rolling, and needs to be continuously updated as rolling continues. The update method for short-term learning is to adjust accordingly based on the deviation between the actual width of the strip and the set width. The adjustment method is the same as that for ordinary strip production and will not be described in detail here.

[0070] In the production process of strip steel of the same category, long-term learning is updated based on short-term learning. That is, in the production of the same category, long-term learning adopts the short-term learning of the previous strip steel. When changing categories, long-term learning retains the short-term learning value after the last rolled strip until the next time this category of steel is rolled, and then it is updated.

[0071] Specifically, S5 involves width correction compensation based on the rolling deviation of each pass, as detailed below.

[0072] According to industry surveys, existing hot rolling production processes all utilize reversible mills in the roughing zone. For transition materials, the roughing model incorporates a feature that adjusts for width based on rolling deviations between each pass, moving these deviations to the next pass for width correction. Deviations are gradually eliminated starting from each pass.

[0073] S51. During each rolling pass, the model performs an inter-pass calculation to calculate a target width based on the changes in parameters such as on-site equipment and processes.

[0074] S52. Obtain the average actual width of the model by using the width detector produced on site. Based on the inspection, remove the width anomaly of 15m at the head.

[0075] S53. Obtain the deviation between the setting and the actual value for each pass, and use it in the previously calculated target width value for each pass, see Formula 2. When it is the last pass, the target width between passes is the final roughing target width.

[0076] Target width between tracks = Original target width of track + Track deviation width; Formula 2. Specific implementation examples:

[0077] Strip steel rolling schedule for 1780 hot rolling line:

[0078]

[0079] The roughing zone control method is as follows:

[0080] Strip steel: 24834400700

[0081] Step S1: Based on the changes in strip steel classification, determine whether it is a transition material and assign initial values ​​to the long-term learning coefficients;

[0082]

[0083]

[0084] Transition material judgment:

[0085] Based on the classification of steel grades and compared with the classification of the previous strip, each change is incremented by 1, and the cumulative change is calculated. When the cumulative change value SC exceeds 2, it is considered a transition material. The cumulative change value SC of this strip is 3, so it is determined to be a transition material.

[0086] Since this type of steel is not being produced for the first time, the long-term learning factor for this strip is 1.4569.

[0087] The short-time learning coefficient is 1.13305.

[0088] The rolling time difference between this steel block and the steel block of the same category is 80 hours;

[0089] Step S2: Adjust the self-learning pace and confidence level based on the amount of information change;

[0090] Because the strip rolling time difference is 80 hours, exceeding three days, and the cumulative value is 3...

[0091] Self-learning rhythm step: 1.1;

[0092] Confidence level delta: 0.7;

[0093] Step S3: Based on long-term learning, adjust the width of the strip to complete the strip width design;

[0094] The genetic coefficients calculated by the model using confidence level and self-learning pace are as follows:

[0095] Heredity coefficient = [Long-term learning value in roughing rolling * confidence level delta + Short-term learning value in roughing rolling * (1-delta)] * self-learning pace step

[0096] = (1.4569 * 0.7 + 1.13305 * 0.3) * 1.1 = 1.5609

[0097] The target width for rough rolling was initially set at 1167.3.

[0098] The target width for rough rolling after compensation is set: 1168.8 mm.

[0099] Step S4: Perform short-time self-learning calculations based on the information of the front and rear strips, and modify the long-time learning coefficients;

[0100] When this strip is rolled, the next strip will have a different category, so the long-term learning coefficient will become the short-term coefficient of the current strip, and the long-term learning coefficient will be changed to 1.3305.

[0101] Step S5: Perform width correction compensation based on the rolling deviation of each pass;

[0102] Meigang's hot rolling roughing area has two reversible mills, R1 and R2, both of which are reversible. Their rolling pattern is: three passes for R1 + three passes for R2. Meigang only has width detection equipment at the exits of R1 and R2; width is not collected during the first pass of R1. Therefore:

[0103]

[0104]

[0105] Therefore, the final target width is the original target width - the width deviation of R21 = 1168.8 + 4.5 = 1173.4.

[0106] It should be noted that the above embodiments are not intended to limit the scope of protection of the present invention. Equivalent transformations or substitutions made based on the above technical solutions all fall within the scope of protection of the claims of the present invention.

Claims

1. A method for controlling the width of the roughing zone of hot-rolled transition material, characterized in that, The method includes the following steps: S1. Based on the changes in strip classification, determine whether it is a transitional material. In the hot rolling control system, the changes in strip classification mainly focus on large deviations in width, thickness, and final rolling temperature, as well as changes in type classification, as shown in Table 1: Table 1: Strip Steel Classification When determining the transition material, the classification of steel grades is compared with the classification of the previous strip steel. For each change, +1 is added, and the change is accumulated. When the accumulated change value SC exceeds 2, it is considered to be the transition material. When controlling the transition material, long-term and short-term genetic learning coefficients are introduced. Initial values ​​were assigned to the long-term and short-term genetic learning coefficients. S2. Based on the cumulative value SC of the change, adjust the self-learning pace (step) and confidence level. S3. Based on long-term learning, adjust the width of the strip to complete the strip width design, as detailed below. The strip learning coefficient is calculated by the deviation between the actual width and the calculated width of the strip in the roughing zone. It reflects the width deviation of the strip in the roughing zone and serves as an important reference for width adjustment in the next rolling process. Adjust the formula to Formula 1: After compensation, the target width of the roughing roll is calculated as follows: Target width of the roughing roll - [Long-term learning value of the roughing roll * Confidence delta + Short-term learning value of the roughing roll * (1 - delta)] * Self-learning pace step; Formula 1; S4. Perform short-time self-learning calculations based on the information from the front and rear strips, and modify the long-time learning coefficients. S5. Based on the rolling deviation of each pass, perform width correction compensation.

2. The method for controlling the width of the roughing zone of hot-rolled transition material according to claim 1, characterized in that, In step S1, A long-term learning genetic coefficient table for the roughing rolling zone was created using classification information, as shown in Table 2. During the first rolling pass, the initial long-term genetic coefficient for the roughing rolling zone is equivalent to the short-term learning coefficient of the strip. Based on the actual production conditions, the range is defined as 0.9–1.18, and the initial short-term genetic coefficient is 1.

014. After one rolling pass, the genetic coefficient table will contain the corresponding classification information and genetic coefficients; these genetic coefficients at this point are referred to as the long-term learning coefficients. Table 2: Genetic Values ​​Table.

3. The method for controlling the width of the roughing zone of hot-rolled transition material according to claim 2, characterized in that, S2. Based on the information change, adjust the self-learning pace (step) and confidence level. Specifically, during transition material rolling, the learning confidence level is determined based on the cumulative information change value (SC) calculated in S1. When SC changes significantly, the long-term learning confidence level increases, and vice versa. These two values ​​are based on empirical values ​​summarized from on-site rolling, as shown in Table 3 below: Table 3: Self-learning pace step and confidence delta configuration table.

4. The method for controlling the width of the roughing zone of hot-rolled transition material according to claim 3, characterized in that, S4. Perform short-time self-learning calculations based on the information from the front and rear strips, and modify the long-time learning coefficients as follows. In the production process of strip steel of the same category, long-term learning is updated based on short-term learning. That is, in the production of the same category, long-term learning adopts the short-term learning of the previous strip steel. When changing categories, long-term learning retains the short-term learning value after the last rolled strip until the next time this category of steel is rolled, and then it is updated.

5. The method for controlling the width of the roughing zone of hot-rolled transition material according to claim 4, characterized in that, S5. Based on the rolling deviation of each pass, perform width correction compensation, as detailed below. S51. During each rolling pass, the model performs an inter-pass calculation to calculate a target width based on the changes in parameters such as on-site equipment and processes. S52. Obtain the average actual width of the model by using the width detector produced on site. Based on the inspection, remove the width anomaly of 15m at the head. S53. Obtain the deviation between the setting and the actual value for each pass, and use it in the previously calculated target width value for each pass, see Formula 2. When it is the last pass, the target width between passes is the final roughing target width. Target width between tracks = Original target width of track + Track deviation width; Formula 2.

Images