A self-learning method for coiling temperature model adapted to fast-paced rolling

Abstract

The present invention relates to a self-learning method for a coiling temperature model that is adapted to fast-paced rolling, and belongs to the technical field of hot rolling methods in the metallurgical industry. The technical solution of the present invention is: a water-cooled self-learning target point of the coiling temperature model is configured, and when the self-learning target point passes through a pyrometer, the actual temperature data is obtained, and immediately compared with the predicted temperature, and a water-cooled self-learning coefficient is calculated. After upper and lower limits and data smoothing processing, the coefficient is updated to the model database for coiling temperature model calculation. The beneficial effect of the present invention is that it avoids the original design of updating the model self-learning coefficient only after the hot-rolled strip completely passes through the pyrometer before coiling, and in fast-paced production, the updated model self-learning coefficient cannot be used for the setting calculation of the next piece of strip, resulting in abnormal temperature control of multiple coils of strip.

Description

Technical Field

[0001] The invention relates to a coiling temperature model self-learning method adapted to fast-paced rolling, belonging to the technical field of hot rolling methods in the metallurgical industry. Background Art

[0002] In the field of hot rolling, ensuring the mechanical properties of hot-rolled strip requires precise control of the strip's temperature at the coiling entrance, also known as the coiling temperature. Typically, a coiling temperature control model calculates and sets the amount of water the strip passes through the layer cooling zone. To ensure control accuracy, coiling temperature control models often incorporate a self-learning function. This function uses the actual coiling temperature data of the previous strip to modify the calculated settings for the current strip, achieving satisfactory control results.

[0003] During the production process, the above control method was found to have the following problems:

[0004] The coiling temperature model self-learning update occurs after the coiler's pre-pyrometer test is complete, that is, after the strip has completely passed the pyrometer. The coiling temperature model is last calculated after the first stand of the finishing mill is loaded. When production is fast, the previous strip may not have completely passed the pyrometer before the current strip enters the finishing mill (i.e., the first stand is loaded). At this point, the previous strip's actual data has not yet been self-learned and updated. In this case, the previous strip's model self-learning cannot correct the current strip's coiling temperature model calculation. Inappropriate self-learning parameters can result in deviations in coiling temperature control for two or more consecutive strips, causing significant losses to the company. Summary of the Invention

[0005] The purpose of the present invention is to provide a self-learning method for a coiling temperature model that is adapted to fast-paced rolling. By configuring a water-cooling self-learning target point of the coiling temperature model, when the self-learning target point passes through a pyrometer, the actual temperature data is obtained, and immediately compared with the predicted temperature. The water-cooling self-learning coefficient is calculated, and after upper and lower limits and data smoothing processing, it is updated to the model database for coiling temperature model calculation. This avoids the original design in which the model self-learning coefficient is updated only after the hot-rolled strip has completely passed through the pyrometer before coiling. In fast-paced production, the updated model self-learning coefficient cannot be used for the setting calculation of the next piece of strip, resulting in abnormal temperature control of multiple coils of strip. This effectively solves the above-mentioned problems existing in the background technology.

[0006] The technical solution of the present invention is: a coiling temperature model self-learning method adapted to fast-paced rolling, comprising the following steps:

[0007] (1) Configure the parameters related to the water cooling self-learning position point, including the minimum length L min and the minimum length ratio coefficient e;

[0008] (2) Calculate the water cooling self-learning target point, track and calculate the length of the strip passing through the pyrometer before coiling through the control system, when the passing length is greater than the minimum length L min Or when the ratio of the passing length to the total length of the strip is greater than the minimum length ratio coefficient e, the current position is determined to be the water-cooling self-learning target point;

[0009] (3) Obtain the actual temperature of the water-cooling self-learning target point and calculate the water-cooling self-learning coefficient by comparing it with the predicted temperature;

[0010] (4) The upper and lower limits of the water cooling self-learning coefficient are judged and smoothed. After the upper and lower limits are judged again, they are updated to the model database for the next coiling temperature model calculation.

[0011] In the step (1), the parameters related to the water-cooling self-learning target point also include the upper and lower limits of the current value of the water-cooling self-learning coefficient, the smoothing coefficient and the upper and lower limits of the current value; the minimum length L min Less than the distance from the pyrometer at the finishing rolling exit to the coiler; the minimum length proportional coefficient e means the ratio of the length from the target point position to the strip head to the total length of the strip, and the value range is 0.1 to 0.3; the upper and lower limits of the current value of the water-cooling self-learning coefficient are used to judge and process the water-cooling self-learning coefficient calculated using the actual coiling temperature specifications, and the value range is 0.6 to 3.5; the new upper and lower limits of the water-cooling self-learning coefficient are used to judge and process the water-cooling self-learning coefficient calculated using the actual coiling temperature specifications, and the value range is 0.6 to 3.5; the smoothing coefficient range is 0.2 to 0.8, which indicates the proportion of the self-learning value calculated based on the current data during self-learning update. The larger the value, the faster the self-learning update rate.

[0012] In the step (2), when the hot-rolled strip coiling temperature is controlled, the strip is divided into several sample segments in the length direction, and the length L of each sample segment is S The range is 1 to 5 meters. The target point for calculating the water cooling self-learning is the minimum length L. min and the minimum length ratio coefficient e to obtain the sequence number i of the sample segment;

[0013] The calculation method of the sequence number i of the water-cooled self-learning target point sample segment is: use the minimum length L min Divide by the sample length Ls, round up and add 1, recorded as i1; multiply the minimum length proportional coefficient e by the calculated total length of the strip, divide by the sample length Ls, round up and add 1, recorded as i2; if i1<i2, then i=i1, otherwise i=i2.

[0014] In the step (3), when the water-cooled self-learning target point, that is, the i-th sample segment passes through the pyrometer before coiling, the average value T of the actual measurement data of the i-th sample segment on the pyrometer is obtained. m , according to the original water-cooled self-learning coefficient zold , actual strip speed v, strip thickness, actual layer cooling inlet temperature and actual layer cooling water volume to calculate the predicted coiling temperature T p , that is, the function for predicting the coiling temperature is:

[0015] T p =f(z,v,h,FDT,flow) (1)

[0016] Where: z is the water cooling self-learning coefficient, v is the actual speed, h is the actual thickness of the strip, FDT is the actual temperature of the layer cooling inlet, and flow is the actual water volume of the layer cooling;

[0017] Td is the self-learning update dead zone, with a value range of 3 to 5°C. p With T m The absolute value of the difference is less than T d When the control deviation is small, the water cooling self-learning value is not updated, that is, Z n =Z old Otherwise, based on the coiling temperature calculation function of formula (1), according to the actual average temperature T of the water-cooling self-learning target point m Calculate the water cooling self-learning coefficient z of the current strip c value.

[0018] In the step (4), the calculated water cooling self-learning coefficient z c Perform upper and lower limits and smoothing processing, and update to the model database; Z c The upper and lower limits are denoted as z c-up and z c-ll , for z c To judge the upper and lower limits, the calculation formula is:

[0019]

[0020] z c Perform smoothing and calculate the new water cooling self-learning coefficient z n , the calculation formula is:

[0021] z n =(1-β)×z old +β×z c (3)

[0022] Among them, β is the smoothing coefficient, and its value range is 0.2 to 0.8;

[0023] For the new water cooling self-learning coefficient z n To judge the upper and lower limits, the calculation formula is:

[0024]

[0025] Among them, zn-up For z n The upper limit value of z n-ll For z n The lower limit value of the water cooling self-learning coefficient z n Update to the model database for use in setting the temperature model for the next strip coiling to improve calculation accuracy.

[0026] The beneficial effects of the present invention are as follows: by configuring the water-cooling self-learning target point of the coiling temperature model, when the self-learning target point passes through the pyrometer, the actual temperature data is obtained, and it is immediately compared with the predicted temperature, and the water-cooling self-learning coefficient is calculated. After the upper and lower limits and data smoothing processing are performed, it is updated to the model database for use in the coiling temperature model calculation, avoiding the original design in which the model self-learning coefficient is updated only after the hot-rolled strip completely passes through the pyrometer before coiling. In fast-paced production, the updated model self-learning coefficient cannot be used for the setting calculation of the next piece of strip, resulting in abnormal temperature control of multiple coils of strip. BRIEF DESCRIPTION OF THE DRAWINGS

[0027] Figure 1 It is the workflow diagram of the present invention. DETAILED DESCRIPTION

[0028] In order to make the purpose, technical solutions and advantages of the invention implementation cases clearer, the technical solutions in the invention implementation cases will be clearly and completely described below in conjunction with the drawings in the implementation cases. Obviously, the implementation cases described are only a small part of the implementation cases of the present invention, rather than all the implementation cases. Based on the implementation cases in the present invention, all other implementation cases obtained by ordinary technicians in this field without making creative work are within the scope of protection of the present invention.

[0029] A self-learning method for a coiling temperature model adapted to fast-paced rolling includes the following steps:

[0030] (1) Configure the parameters related to the water cooling self-learning position point, including the minimum length L min and the minimum length ratio coefficient e;

[0031] (2) Calculate the water cooling self-learning target point, track and calculate the length of the strip passing through the pyrometer before coiling through the control system, when the passing length is greater than the minimum length L min Or when the ratio of the passing length to the total length of the strip is greater than the minimum length ratio coefficient e, the current position is determined to be the water-cooling self-learning target point;

[0032] (3) Obtain the actual temperature of the water-cooling self-learning target point and calculate the water-cooling self-learning coefficient by comparing it with the predicted temperature;

[0033] (4) The upper and lower limits of the water cooling self-learning coefficient are judged and smoothed. After the upper and lower limits are judged again, they are updated to the model database for the next coiling temperature model calculation.

[0034] In the step (1), the parameters related to the water-cooling self-learning target point also include the upper and lower limits of the current value of the water-cooling self-learning coefficient, the smoothing coefficient and the upper and lower limits of the current value; the minimum length L min Less than the distance from the pyrometer at the finishing rolling exit to the coiler; the minimum length proportional coefficient e means the ratio of the length from the target point position to the strip head to the total length of the strip, and the value range is 0.1 to 0.3; the upper and lower limits of the current value of the water-cooling self-learning coefficient are used to judge and process the water-cooling self-learning coefficient calculated using the actual coiling temperature specifications, and the value range is 0.6 to 3.5; the new upper and lower limits of the water-cooling self-learning coefficient are used to judge and process the water-cooling self-learning coefficient calculated using the actual coiling temperature specifications, and the value range is 0.6 to 3.5; the smoothing coefficient range is 0.2 to 0.8, which indicates the proportion of the self-learning value calculated based on the current data during self-learning update. The larger the value, the faster the self-learning update rate.

[0035] In the step (2), when the hot-rolled strip coiling temperature is controlled, the strip is divided into several sample segments in the length direction, and the length L of each sample segment is S The range is 1 to 5 meters. The target point for calculating the water cooling self-learning is the minimum length L. min and the minimum length ratio coefficient e to obtain the sequence number i of the sample segment;

[0036] The calculation method of the sequence number i of the water-cooled self-learning target point sample segment is: use the minimum length L min Divide by the sample length Ls, round up and add 1, recorded as i1; multiply the minimum length proportional coefficient e by the calculated total length of the strip, divide by the sample length Ls, round up and add 1, recorded as i2; if i1<i2, then i=i1, otherwise i=i2.

[0037] In the step (3), when the water-cooled self-learning target point, that is, the i-th sample segment passes through the pyrometer before coiling, the average value T of the actual measurement data of the i-th sample segment on the pyrometer is obtained. m , according to the original water-cooled self-learning coefficient z old , actual strip speed v, strip thickness, actual layer cooling inlet temperature and actual layer cooling water volume to calculate the predicted coiling temperature T p , that is, the function for predicting the coiling temperature is:

[0038] T p =f(z,v,h,FDT,flow) (1)

[0039] Where: z is the water cooling self-learning coefficient, v is the actual speed, h is the actual thickness of the strip, FDT is the actual temperature of the layer cooling inlet, and flow is the actual water volume of the layer cooling;

[0040] Td is the self-learning update dead zone, with a value range of 3 to 5°C. p With T m The absolute value of the difference is less than T d When the control deviation is small, the water cooling self-learning value is not updated, that is, Z n =Z old Otherwise, based on the coiling temperature calculation function of formula (1), according to the actual average temperature T of the water-cooling self-learning target point m Calculate the water cooling self-learning coefficient z of the current strip c value.

[0041] In the step (4), the calculated water cooling self-learning coefficient z c Perform upper and lower limits and smoothing processing, and update to the model database; Z c The upper and lower limits are denoted as z c-up and z c-ll , for z c To judge the upper and lower limits, the calculation formula is:

[0042]

[0043] z c Perform smoothing and calculate the new water cooling self-learning coefficient z n , the calculation formula is:

[0044] z n =(1-β)×z old +β×z c (3)

[0045] Among them, β is the smoothing coefficient, and its value range is 0.2 to 0.8;

[0046] For the new water cooling self-learning coefficient z n To judge the upper and lower limits, the calculation formula is:

[0047]

[0048] Among them, z n-up For z n The upper limit value of z n-ll For z n The lower limit value of the water cooling self-learning coefficient z n Update to the model database for use in setting the temperature model for the next strip coiling to improve calculation accuracy.

[0049] Example:

[0050] The distance from the finishing rolling exit to the coiler of a hot rolling production line is 148.2m. The water cooling self-learning related parameter configuration in the coiling temperature model configuration file.

[0051] The minimum length of the water-cooled self-learning target point is L min Generally, it should be less than the distance from the finishing rolling outlet to the coiler.

[0052] Preferably, L min =120m.

[0053] The minimum length ratio coefficient ranges from 0.1 to 0.3, preferably, e=0.1.

[0054] The current value range of the water cooling self-learning coefficient is 0.6 to 3.5. Preferably, the range is set to 0.6 to 3.0.

[0055] The new value range of the water cooling self-learning coefficient is 0.6 to 3.5, and preferably, the range is set to 0.6 to 3.0.

[0056] The smoothing coefficient β ranges from 0.2 to 0.8. Preferably, the smoothing coefficient β=0.3.

[0057] In the continuous production of steel grade SPHC, with a width of 1250mm, a target thickness of 5.75mm, and a target coiling temperature of 620℃, the sample length is 2.5m. Based on the slab size, the finished strip length is calculated to be 386.4m, and the total number of samples is 154. Calculate the sequence number i of the sample segment of the water-cooled self-learning target point, where

[0058] i1=minimum length L min / sample length + 1 = 120 / 2.5 + 1 ≈ 49

[0059] i2 = total strip length × minimum length ratio coefficient + 1 = 386.4 × 0.1 / 2.5 + 1 ≈ 16

[0060] The sequence number i of the water-cooling self-learning target point sample segment is the minimum value of i1 and i2, i1>i2, so i=i2=16.

[0061] Furthermore, the actual data corresponding to the sample segment number 16 is obtained, and the average value T of the actual measured data is m =634℃, the original water-cooled self-learning coefficient z old =1.34, actual strip speed v = 5.12m / s, strip thickness = 5.75mm, actual layer cooling inlet temperature = 879℃, actual layer cooling water volume = 7586m 3 / h, predicted coiling temperature T p =f(z, v, h, FDT, flow) = 623°C.

[0062] Furthermore, the self-learning updates the dead zone T d The value range is 3~5℃, preferably, T d =5℃, T p With T m The absolute value of the deviation = 11°C, which is greater than the control dead zone T d , then according to Tm, substitute the function f(z, v, h, FDT, flow) to calculate the water cooling self-learning coefficient z of the current strip c The value of z c =1.28.

[0063] Furthermore, the calculated current water cooling self-learning coefficient z c Perform upper and lower limits and smoothing processing, and update to the model database. c The upper and lower limits are 3.0 and 0.6 respectively, z c =1.28 is within the upper and lower limits, so z c =1.28. c Perform smoothing and calculate the new water cooling self-learning coefficient z n , where the smoothing coefficient β = 0.3, the calculation process of zn is:

[0064] z n =(1-β)×z old +β×z c =(1-0.3)×1.34+0.3×1.28=1.322

[0065] Furthermore, for z n Make upper and lower limit judgments, z n The upper and lower limits are 3.0 and 0.6 respectively, z n =1.322 is within the upper and lower limits, so z n =1.322.

[0066] After calculating the new water-cooling self-learning value, it is immediately updated to the model database without waiting for the entire strip to pass through the coiling pyrometer, and used for the coiling temperature model setting calculation of the next strip. This ensures that when the next strip bites the steel in the finishing rolling, the new water-cooling self-learning value is used when the last coiling temperature model calculation is triggered, thereby ensuring the coiling temperature control accuracy during fast-paced production.

Claims

1. A self-learning method for coiling temperature model adapted to fast-paced rolling, characterized in that The following steps are involved: (1) Configure the parameters related to the water cooling self-learning position point, including the minimum length L min and minimum length ratio coefficient e; minimum length L min Less than the distance from the pyrometer at the finishing exit to the coiler; the minimum length ratio coefficient e means the ratio of the length from the target point to the strip head to the total length of the strip, and the value range is 0.1 to 0.3; (2) Calculate the water cooling self-learning target point, track and calculate the length of the strip passing through the pyrometer before coiling through the control system, when the passing length is greater than the minimum length L min Or when the ratio of the passing length to the total length of the strip is greater than the minimum length ratio coefficient e, the current position is determined to be the water-cooling self-learning target point; (3) Obtain the actual temperature of the water-cooling self-learning target point and calculate the water-cooling self-learning coefficient by comparing it with the predicted temperature; when the water-cooling self-learning target point, that is, the i-th sample segment passes through the pyrometer before coiling, obtain the average value T of the actual measurement data of the i-th sample segment on the pyrometer m , according to the original water-cooled self-learning coefficient z old , actual strip speed v, strip thickness, actual layer cooling inlet temperature and actual layer cooling water volume to calculate the predicted coiling temperature T p , that is, the function for predicting the coiling temperature is: T p =f(z,v,h,FDT,flow)(1) Where: z is the water cooling self-learning coefficient, v is the actual speed, h is the actual thickness of the strip, FDT is the actual temperature of the layer cooling inlet, and flow is the actual water volume of the layer cooling; Td is the self-learning update dead zone, with a value range of 3 to 5°C. p With T m The absolute value of the difference is less than T d When the control deviation is small, the water cooling self-learning value is not updated, that is, the new water cooling self-learning coefficient Z n =Z old Otherwise, based on the coiling temperature calculation function of formula (1), according to the actual average temperature T of the water-cooling self-learning target point m Calculate the water cooling self-learning coefficient z of the current strip c The value of (4) The upper and lower limits of the water cooling self-learning coefficient are judged and smoothed. After the upper and lower limits are judged again, they are updated to the model database for the next coiling temperature model calculation.

2. The method for self-learning a coiling temperature model adapted to fast-paced rolling according to claim 1, characterized in that: In the step (1), the parameters related to the water-cooling self-learning target point also include the upper and lower limits of the current value of the water-cooling self-learning coefficient and the smoothing coefficient; the upper and lower limits of the current value of the water-cooling self-learning coefficient are used to judge and process the water-cooling self-learning coefficient calculated using the actual coiling temperature specification, and the value range is 0.6 to 3.5; the smoothing coefficient has a value range of 0.2 to 0.8, which represents the proportion of the self-learning value calculated according to the current data during the self-learning update, and the larger the value, the faster the self-learning update rate.

3. The method for self-learning a coiling temperature model adapted to fast-paced rolling according to claim 1, characterized in that: In the step (2), when the hot-rolled strip coiling temperature is controlled, the strip is divided into several sample segments in the length direction, and the length L of each sample segment is S The range is 1 to 5 meters. The target point for calculating the water cooling self-learning is the minimum length L. min and the minimum length ratio coefficient e to obtain the sequence number i of the sample segment; The calculation method of the sequence number i of the water-cooled self-learning target point sample segment is: use the minimum length L min Divide by the sample length Ls, round up and add 1, recorded as i1; multiply the minimum length proportional coefficient e by the calculated total length of the strip, divide by the sample length Ls, round up and add 1, recorded as i2; if i1<i2, then i=i1, otherwise i=i2.

4. The method for self-learning a coiling temperature model adapted to fast-paced rolling according to claim 3, characterized in that: In the step (4), the calculated water cooling self-learning coefficient z c Perform upper and lower limits and smoothing processing, and update to the model database; Z c The upper and lower limits are denoted as z c-up and z c-ll , for z c To judge the upper and lower limits, the calculation formula is: z c Perform smoothing and calculate the new water cooling self-learning coefficient z n , the calculation formula is: With n =(1-β)×z old +β×z c (3) Among them, β is the smoothing coefficient, and its value range is 0.2 to 0.8; For the new water cooling self-learning coefficient z n To judge the upper and lower limits, the calculation formula is: Among them, z n-up For z n The upper limit value of z n-ll For z n The lower limit value of the water cooling self-learning coefficient z n Update to the model database for use in setting the temperature model for the next strip coiling to improve calculation accuracy.

Images