Method for improving cross section of first coil of hot rolling

Abstract

The present application relates to the technical field of hot rolling steel control, and especially relates to a method for improving cross section defect of first coil in hot rolling, which comprises rolling plan preparation, first coil steel identification, model setting and pre-rolling gap leveling; the model setting comprises model parameter initialization, construction of a target function with minimum cross section thickness deviation and minimum temperature deviation as targets, introduction of deviation from default value penalty term and regression value penalty term, adoption of gradient descent method to solve optimal model parameters, and correction of the parameters including rolling force prediction parameters, speed setting parameters and temperature prediction parameters of each stand, and model output; the present application improves first coil steel quality index precision, reduces non-planned repair cutting loss, reduces unnecessary steel stacking accident loss, and satisfies user quality demand by optimizing rolling plan preparation, accurately identifying first coil steel, adaptively adjusting model parameters and accurately leveling pre-rolling gap.

Description

Technical Field

[0001] This invention relates to the field of hot-rolled steel control technology, and in particular to a method for improving the defective cross-section of the first coil in hot rolling. Background Technology

[0002] In hot rolling production, the rolling of the first coil is a typical non-steady-state rolling process. On average, the finishing mill rolls are changed 4-6 times daily. After each roll change, the roll diameter of the new rolls needs to be re-entered, and zero adjustment is required. Due to interruptions in preceding production, the mill equipment status and rolling environment differ significantly from steady-state rolling. Furthermore, problems such as the lack of specificity in traditional rolling plans, inaccurate identification of the first coil, poor adaptability of model parameter settings, and inadequate pre-roll gap leveling easily lead to cross-sectional defects in the first coil, such as being thicker in the middle and thinner at the edges, thicker at the edges and thinner in the middle, tilted cross-sections, and uneven thickness.

[0003] In existing technologies, rolling planning relies heavily on manual experience, making it difficult to fully consider the unsteady rolling characteristics of the first coil. This leads to unreasonable rolling connections during specification switching and steel grade changes, further exacerbating cross-sectional defects. First coil identification mainly relies on manual marking or simple batch switching signals, failing to accurately distinguish between the true first coil and normal steel changes within a batch, resulting in the misuse or omission of targeted control measures. Regarding model setting, the pre-setting, post-calculation, and self-learning mechanisms of traditional hot rolling models are difficult to adapt to the unsteady working conditions of the first coil. Model parameters often use default initial values ​​or empirical values, failing to accurately predict key parameters such as deformation resistance and heat conduction during the rolling process. Pre-rolling roll gap leveling often uses fixed mechanical leveling methods, failing to consider the influence of the first coil's incoming material deviation and the hot elastic deformation of the rolls, resulting in deviations between the actual and theoretical roll gaps, ultimately affecting cross-sectional quality.

[0004] With the increasing demand for multi-variety, small-batch, and customized production, the production frequency of the first coil of steel rolled has increased significantly. Defective cross-sections have reduced the product qualification rate, with over 85% of coils exceeding 150 IU in straightness, requiring leveling and rework, increasing process costs, affecting delivery timelines, and potentially impacting the stability of subsequent rolling processes. For hot-rolled strip steel plants, the performance indicators for the first coil cannot be guaranteed due to a lack of continuous control parameter feedback. Accurate assessment of equipment status, production line conditions, and slab process parameters under new rolling conditions is impossible. Consequently, the precision control indicators for straightness, thickness, crown, and wedge shape are lower than those for conventionally rolled strip steel. Therefore, a comprehensive improvement method integrating optimized rolling plan preparation, accurate first coil identification, adaptive model setting, and precise pre-roll roll gap leveling is urgently needed to solve the technical challenge of defective cross-sections in the first coil of hot-rolled steel. Summary of the Invention

[0005] This invention provides a method for improving the cross-sectional defects of the first coil in hot rolling. By optimizing the rolling plan, accurately identifying the first coil, adaptively adjusting model parameters, and precisely leveling the pre-roll gap, the method improves the accuracy of the first coil's quality indicators, reduces unplanned rework and cutting losses, reduces unnecessary steel stacking accidents, and meets user quality requirements.

[0006] To achieve the above objectives, the present invention employs the following technical solution: A method for improving the defective cross-section of the first coil in a hot-rolled die includes the following steps: S1. Rolling plan preparation: Select the easiest-to-roll variety, thickness and width in the design outline as the first coil of steel, and clarify the slab specification matching, steel rolling compatibility, equipment load limit, rolling temperature window and specification transition gradient between the first coil of steel and subsequent coils. S2. First Coil Identification: Accurately identify the first coil after the finishing work rolls are replaced, distinguishing it from other coils except the first coil. S3. Model setting, which includes the following process: S3.1 Model parameter initialization: Retrieve steady-state rolling data with similar steel grades and specifications to the first coil from the historical rolling database, calculate carbon equivalent matching degree and specification similarity to screen similar coil groups, and use the weighted average method to calculate the deformation resistance coefficient, thermal conductivity coefficient and friction factor of similar coil groups as initial parameters. The weights are dynamically allocated according to the rolling time distance and cross-sectional qualification rate of similar coils, and the weight of qualified coils within the current 90 days or the previous 500 coils is not less than 0.6. S3.2 Construct an objective function with the goal of minimizing cross-sectional thickness deviation and temperature deviation, introduce a penalty term for deviation from default value and a penalty term for regression value, and use the gradient descent method to solve for the optimal model parameters. The corrected parameters include the rolling force prediction parameters, speed setting parameters and temperature prediction parameters for each stand. S3.3 Model Output: Input the corrected model parameters into the process control system to generate a personalized rolling program for the first coil of steel, and set a dynamic adjustment threshold for the parameters. When the cross-sectional deviation is detected to exceed 0.05mm during the rolling process, the model parameters are triggered for secondary correction. S4. Pre-rolling roll gap leveling: Collect the hydraulic pressure on the operating side and drive side of each stand in the finishing mill, and display the hydraulic pressure difference on the operation page. The operator adjusts the roll gap according to the average pressure difference of the normal steel coil in the previous cycle, so that the pressure difference value of each stand of the first coil after the current roll change meets the preset requirements.

[0007] Furthermore, the thickness of the first coil of steel is limited to 4.0mm to 5.0mm, the width is limited to 1200 to 1400mm, and the steel grade is Q235B or SPHC series low carbon steel.

[0008] Furthermore, the thickness difference between the first coil of steel and the second coil of steel shall not exceed 15%, and the width difference shall not exceed 10%.

[0009] Furthermore, the personalized rolling procedure includes the set values ​​for the reduction amount, rolling speed, bending roll force, and cooling water volume for each stand.

[0010] Furthermore, the accurate identification of the first coil of rolling mill is based on the following criteria: when the diameter of the finishing mill rolls F5 to F7 changes, the next piece of steel is considered to be the first coil of rolling mill.

[0011] Furthermore, in step S3.2, the objective function introduces a penalty term for deviation from the roll changing condition, expressed as: minF=α·|h0-h|+β·|T0-T|+γ·ΔS; Where h0 is the target thickness of the first coil of steel, and h is the model-predicted thickness; T0 is the conventional final rolling temperature of the first coil of steel, and T is the model-predicted final rolling temperature; ΔS is the deviation of the roll change condition, which characterizes the combined situation of the thermal expansion of the new roll and the fluctuation of the rolling speed; α, β, and γ are weighting coefficients, and α=0.6, β=0.3, and γ=0.1.

[0012] Furthermore, in step S3.1, the historical rolling database contains ≥1,000 valid rolling data entries, and the update cycle is ≤24h; The screening criteria for similar coils are as follows: the steel grade is the target steel grade or the same type of steel with a carbon equivalent deviation of no more than ±0.03%; the specifications are ≥85% similar to the first coil; the rolling conditions are the same type of finishing mill, steady-state rolling stage, and the cross-sectional qualification rate is ≥95%; and abnormal coil data with thickness fluctuations or equipment failures during the rolling process are excluded.

[0013] Furthermore, in step S3.1, the weight allocation rule of the weighted average method is as follows: the weight of similar qualified papers within the past 3 months is 0.7, and the weight of similar qualified papers within the past 3 to 12 months is 0.3.

[0014] Compared with the prior art, the beneficial effects of the present invention are: By optimizing rolling schedule preparation, accurately identifying the first coil of steel, adaptively adjusting model parameters, and precisely leveling the roll gap before rolling, the accuracy of the first coil's quality indicators has been improved, reducing unplanned rework and cutting losses, minimizing unnecessary steel stacking accidents, and meeting user quality requirements. Roll gap leveling has been effectively improved, and the flatness of the first coil of steel has been significantly reduced, averaging below 100 IU, eliminating the need for leveling rework. The rework rate of the first coil has been reduced from 85% to below 24%, lowering rework costs, accelerating contract delivery, and reducing the proportion of discrepancies in sheet shape. Detailed Implementation

[0015] The specific embodiments of the present invention will be further described below: This invention provides a method for improving the defective cross-section of the first coil in a hot-rolled roll, comprising the following steps: S1. Rolling plan preparation: Select the easiest-to-roll steel variety, thickness, and width from the design outline as the first coil of steel, and clarify the slab specification matching, steel rolling compatibility, equipment load limit, rolling temperature window, and specification transition gradient between the first coil of steel and subsequent coils; taking a 1780mm hot rolling production line as an example: the rolling thickness specification after roll change is limited to 4.0mm~5.0mm, the width is between 1200~1400mm, and the grades are low carbon steel series such as Q235B and SPHC.

[0016] S2. First Coil Identification: Accurately identify the first coil of steel after rolling to distinguish it from other normally rolled products; when the diameter of the finishing mill F5 to F7 rolls changes, the next piece of steel is considered to be the first coil of steel after rolling.

[0017] S3. Model Setting: A separate, precisely controlled set of model parameters is used to avoid the self-learning coefficients interfering with the self-learning settings of normal products. This specifically includes the following process: S3.1 Model parameter initialization: Retrieve steady-state rolling data with similar steel grades and specifications to the first coil from the historical rolling database, calculate carbon equivalent matching degree and specification similarity to screen similar coil groups, and use the weighted average method to calculate the deformation resistance coefficient, thermal conductivity coefficient and friction factor of similar coil groups as initial parameters. The weights are dynamically allocated according to the rolling time distance and cross-sectional qualification rate of similar coils, and the weight of the 500 qualified coils rolled within the current 90 days or before is not less than 0.6. Taking the first coil of Q235B steel, 4.0mm×1250mm as an example, the search dimensions and filtering conditions are set as follows: From the hot rolling production history database (containing ≥1,000 valid rolling data, with an update cycle of ≤24h), similar coil data are searched with "steel grade matching, similar specifications, and compatible rolling conditions" as the core dimensions. The screening criteria are as follows: 1) Steel type: Q235B or low-carbon steel with a carbon equivalent deviation of ≤ ±0.03%; 2) Specifications: target thickness 4.0~5.0mm, target width 1200~1400mm (similarity to the specifications of this first coil ≥ 85%); 3) Rolling conditions: same type of finishing mill, steady-state rolling stage (not the first coil after roll change), cross-sectional pass rate ≥ 95%; 4) Exclusion criteria: abnormal coil data with thickness hit rate less than 95% or equipment failure during the rolling process. Considering that recent rolling data is closer to the current equipment status (characteristics of the new roll after roll replacement), the initial model parameters are calculated using a dynamic weighted average method. The weight allocation rules are as follows: 1) Similar qualified rolls within the last 3 months: weight 0.7 (a total of 580 rolls were selected, accounting for 33.29%); 2) Similar qualified rolls within the last 3 to 12 months: weight 0.3 (a total of 1162 rolls were selected, accounting for 66.71%). The initial core parameters are obtained through weighted calculation. Deformation resistance coefficient K: 1180MPa (basic value of room temperature deformation resistance of Q235B low carbon steel, corrected for rolling temperature). Thermal conductivity coefficient λ: 46 W / (m·K) (suitable for thermal conductivity characteristics in the heating temperature range of 1200~1250℃). Friction factor f: 0.31 (the surface roughness of the new work roll is relatively high, and the friction factor is slightly higher than the conventional value of 0.28 to 0.30 in steady-state rolling). The mill stiffness coefficient C is 2200 kN / mm (based on the calibration value of the same type of finishing mill equipment parameters).

[0018] S3.2 Construct an objective function with the goal of minimizing cross-sectional thickness deviation and temperature deviation, introduce a penalty term for deviation from default value and a penalty term for regression value, and use the gradient descent method to solve for the optimal model parameters. The corrected parameters include the rolling force prediction parameters, speed setting parameters and temperature prediction parameters for each stand. The objective function incorporates a penalty term for deviations in roll-changing conditions to ensure that the parameters are adapted to unsteady rolling. The expression is as follows: minF=α·|h0-h|+β·|T0-T|+γ·ΔS; Where h0 = 4.0 mm, h0 is the target thickness of the first coil of steel; h is the model-predicted thickness; T0 = 880℃ is the conventional final rolling temperature of the first coil of Q235B steel; T is the model-predicted final rolling temperature; ΔS is the deviation of the roll change condition, which characterizes the combined effect of the thermal expansion of the new roll and the fluctuation of the rolling speed; α, β, and γ are weighting coefficients, and α = 0.6, prioritizing the quality of the cross-section; β = 0.3, ensuring the performance of the steel grade; γ = 0.1, adapting to the roll change condition; The optimal model parameters are solved using the gradient descent method, with a convergence accuracy of 10. -4 By substituting the initial parameters and the roll change compensation coefficient into the objective function and solving iteratively, the corrected core parameters are obtained: Deformation resistance coefficient K: 1210MPa, an increase of 2.5% compared to the initial value, to compensate for the fluctuation of rolling force after roll change; Thermal conductivity coefficient λ: 44 W / (m·K), a decrease of 4.3% from the initial value, to accommodate the temperature loss caused by the heat absorption of the new roller; Friction factor f: 0.33, an increase of 6.5% from the initial value, adapted to the surface roughness characteristics of the new roller; Correction coefficients for reduction rate of each frame: The reduction rate of frames F1 to F7 is reduced by 3% to 8% respectively compared with the initial allocation value. The reduction rate gradient of the first coil head is optimized to avoid stress concentration.

[0019] S3.4 Input the corrected model parameters into the process control system to generate a personalized rolling program for the first coil of steel. The personalized rolling program includes set values ​​for the reduction amount, rolling speed, bending roll force, cooling water volume, etc. of each stand, and sets dynamic adjustment thresholds for the parameters. Specifically, the thresholds are set as follows: cross-sectional thickness deviation threshold 0.05mm, rolling temperature deviation threshold ±5℃, and rolling force fluctuation threshold ±3%.

[0020] Simultaneously, secondary correction rules and coefficients for model parameters are set: when a cross-sectional deviation exceeding 0.05mm, a rolling temperature deviation exceeding ±5℃, or a rolling force fluctuation exceeding ±3% is detected during the rolling process, the secondary correction of model parameters is immediately triggered; the secondary correction coefficients are set as follows: deformation resistance coefficient correction coefficient 1.02, thermal conductivity coefficient correction coefficient 0.98, and friction factor correction coefficient 1.01. During the secondary correction, the core parameters after the initial correction are multiplied by the corresponding secondary correction coefficients to obtain the parameters after the secondary correction. At the same time, combined with the real-time collected hydraulic pressure difference and rolling speed data, the reduction rate of each stand is synchronously fine-tuned, with a fine-tuning range of ±1% to ±2%, to ensure the steady state of the rolling process and avoid cross-sectional dimensions exceeding tolerance.

[0021] S4. Pre-rolling roll gap leveling: Collect the hydraulic pressure on the operating side and drive side of each stand in the finishing mill, and display the hydraulic pressure difference on the operation page. The page displays the average pressure difference of the third coil of steel in the previous six cycles. The operator adjusts the roll gap according to the average pressure difference provided on the page so that the pressure difference of each stand of the first coil of steel rolled after the current roll change meets the target value within ±2.

[0022] The above embodiments are implemented based on the technical solution of the present invention, providing detailed implementation methods and specific operation processes. However, the scope of protection of the present invention is not limited to the above embodiments. Unless otherwise specified, the methods used in the above embodiments are conventional methods.

Claims

1. A method for improving the defective cross-section of the first coil in hot rolling, characterized in that, Includes the following steps: S1. Rolling plan preparation: Select the easiest-to-roll variety, thickness and width in the design outline as the first coil of steel, and clarify the slab specification matching, steel rolling compatibility, equipment load limit, rolling temperature window and specification transition gradient between the first coil of steel and subsequent coils. S2. First Coil Identification: Accurately identify the first coil after the finishing work rolls are replaced, distinguishing it from other coils except the first coil. S3. Model setting, which includes the following process: S3.1 Model parameter initialization: Retrieve steady-state rolling data with similar steel grades and specifications to the first coil from the historical rolling database, calculate carbon equivalent matching degree and specification similarity to screen similar coil groups, and use the weighted average method to calculate the deformation resistance coefficient, thermal conductivity coefficient and friction factor of similar coil groups as initial parameters. The weights are dynamically allocated according to the rolling time distance and cross-sectional qualification rate of similar coils, and the weight of qualified coils within the current 90 days or the previous 500 coils is not less than 0.

6. S3.2 Construct an objective function with the goal of minimizing cross-sectional thickness deviation and temperature deviation, introduce a penalty term for deviation from default value and a penalty term for regression value, and use the gradient descent method to solve for the optimal model parameters. The corrected parameters include the rolling force prediction parameters, speed setting parameters and temperature prediction parameters for each stand. S3.3 Model Output: Input the corrected model parameters into the process control system to generate a personalized rolling program for the first coil of steel, and set a dynamic adjustment threshold for the parameters. When the cross-sectional deviation is detected to exceed 0.05mm during the rolling process, the model parameters are triggered for secondary correction. S4. Pre-rolling roll gap leveling: Collect the hydraulic pressure on the operating side and drive side of each stand in the finishing mill, and display the hydraulic pressure difference on the operation page. The operator adjusts the roll gap according to the average pressure difference of the normal steel coil in the previous cycle, so that the pressure difference value of each stand of the first coil after the current roll change meets the preset requirements.

2. The method for improving the defective cross-section of the first hot-rolled coil according to claim 1, characterized in that, The thickness of the first coil of steel is limited to 4.0mm to 5.0mm, and the width is limited to 1200 to 1400mm. The steel grade is Q235B or SPHC series low carbon steel.

3. The method for improving the defective cross-section of the first coil in hot rolling according to claim 1, characterized in that, The thickness difference between the first coil of steel and subsequent coils shall not exceed 15%, and the width difference shall not exceed 10%.

4. The method for improving the defective cross-section of the first coil in hot rolling according to claim 1, characterized in that, The personalized rolling procedure includes the reduction amount, rolling speed, bending roll force, and cooling water volume settings for each stand.

5. The method for improving the defective cross-section of the first coil in hot rolling according to claim 1, characterized in that, The method for accurately identifying the first coil of steel is as follows: when the diameter of the finishing mill rolls F5 to F7 changes, the next piece of steel is considered to be the first coil of steel to be rolled.

6. The method for improving the defective cross-section of the first coil in hot rolling according to claim 1, characterized in that, In step S3.2, the objective function introduces a penalty term for deviation from the roll changing condition, expressed as: minF=α·|h0-h|+β·|T0-T|+γ·ΔS; Where h0 is the target thickness of the first coil of steel, and h is the model-predicted thickness; T0 is the conventional final rolling temperature of the first coil of steel, and T is the model-predicted final rolling temperature; ΔS is the deviation of the roll change condition, which characterizes the combined situation of the thermal expansion of the new roll and the fluctuation of the rolling speed; α, β, and γ are weighting coefficients, and α=0.6, β=0.3, and γ=0.

1.

7. The method for improving the defective cross-section of the first hot-rolled coil according to claim 1, characterized in that, In step S3.1, the historical rolling database contains ≥1,000 valid rolling data entries, and the update cycle is ≤24h. The screening criteria for similar coils are as follows: the steel grade is the target steel grade or the same type of steel with a carbon equivalent deviation of no more than ±0.03%; the specifications are ≥85% similar to the first coil; the rolling conditions are the same type of finishing mill, steady-state rolling stage, and the cross-sectional qualification rate is ≥95%; and abnormal coil data with thickness fluctuations or equipment failures during the rolling process are excluded.

8. The method for improving the defective cross-section of the first coil in hot rolling according to claim 1, characterized in that, In step S3.1, the weighting rules for the weighted average method are as follows: the weight of similar qualified papers within the last 3 months is 0.7, and the weight of similar qualified papers within the last 3 to 12 months is 0.3.