Method for improving width precision of first coil steel with changed specifications

Abstract

The invention relates to a method for improving width precision of first coil steel with changed specifications, which comprises the following steps: if a layer to which a strip steel rough rolling broadening coefficient belongs is not updated within a set time or after a vertical roll is changed, the strip steel rough rolling broadening coefficient of the layer is not suitable for the current working condition; and the rough rolling broadening coefficient of the layer of strip steel is readjusted. The method has the advantages that the problem that the adaptive capacity to working conditions is poor due to low self-learning frequency when rough rolling broadening coefficient layer division is fine is solved; through comparison between the time difference and the set time, whether the rough rolling broadening coefficient of the layer has the requirement of adjusting through divergence self-learning or not can be determined; when a proper reference layer cannot be found for the layer, the rough rolling broadsiding coefficient can be finely adjusted by means of the global correction coefficient; after the layers which are not rolled for a long time and the vertical rollers are replaced, the width control situation in the first few days is obviously improved, and the situation that the through plate is ultra-wide and narrow is greatly reduced due to no equipment failure or incoming material reasons.

Description

Technical Field

[0001] The invention relates to the technical field of metallurgical process control, and in particular to a method for improving the width accuracy of a first steel roll after specification change. Background Art

[0002] One of the most basic requirements for plates is "dimensional accuracy". Low dimensional accuracy will affect product quality. Therefore, improving the width accuracy of hot-rolled plates and reducing the production of unqualified width rolled products are important issues in plate hot rolling production control. In hot rolling production, the rough rolling mill is used to control the width of the plate in addition to obtaining the thickness of the intermediate billet suitable for finishing. The stability of rough rolling width control is directly related to self-learning. If the width control accuracy is not high, it will eventually cause large-scale quality problems.

[0003] The fineness of the layer division of the rough rolling width expansion coefficient has always been a dilemma. The coarser the layer division, the higher the self-learning frequency, but at the same time, the wider the adaptation of the rough rolling width expansion coefficient, the worse the applicability to strip steel of different steel grades and specifications within the layer, and thus the worse the calculation accuracy of the width model. The finer the layer division, the more accurate the rough rolling width expansion coefficient, and the better the applicability to strip steel of different steel grades and specifications within the layer, but the relative learning frequency is lower. The rough rolling width expansion coefficient layer division of a hot rolling line model of a steel plant is relatively fine, with more than 10,000 layers. The self-learning frequency of most layer parameters is low. At the time of writing, there are 10,785 layers, of which 10,148 layers have not been produced in the past 20 days. Since the stability of the rough rolling condition is relatively low, vertical roll calibration, wear, air temperature, furnace condition, rolling rhythm, etc. have a great impact on the rough rolling condition. The width control of the strip is mainly completed in the rough rolling section. Therefore, the adaptability of the rough rolling width coefficient to the working condition is very important. The rough rolling width coefficient that has not been self-learned for more than 20 days cannot adapt to the current working condition. For example, the rough rolling width coefficient corresponding to a steel grade that has not been produced for 6 months is -0.058. If it is not adjusted, the final result may cause the overall width of the strip to be about 10mm too wide, which is urgently needed to be improved. Summary of the invention

[0004] The purpose of the present invention is to provide a method for improving the width accuracy of the first coil of steel when changing specifications, thereby improving the self-learning efficiency of rough rolling width expansion and improving the width control accuracy when changing specifications.

[0005] To achieve the above object, the present invention is implemented through the following technical solutions:

[0006] A method for improving the width accuracy of the first coil of steel after specification change. If the layer to which the strip rough rolling width expansion coefficient belongs has not been updated within the set time or after the vertical roll is changed, the strip rough rolling width expansion coefficient of this layer is not suitable for the current working conditions. Through divergent self-learning, the strip rough rolling width expansion coefficient of this layer is readjusted.

[0007] Divergent self-learning, the steps are as follows:

[0008] S1, obtain the rough rolling width of the strip by self-learning level;

[0009] S2. Retrieve the 'Last Update Time' parameter in the rough rolling width self-learning layer of the strip steel, and confirm the time difference between the last self-learning of the rough rolling width self-learning layer of this strip steel and the current time. If the time difference is higher than the set time, then enter the divergent self-learning;

[0010] S3, finding a self-learning level of rough rolling width of similar strip steel, if there is no self-learning level of rough rolling width of similar strip steel, then the rough rolling width expansion coefficient of the strip steel of this level is adjusted by using the global correction coefficient of rough rolling width expansion; if there is a self-learning level of rough rolling width of similar strip steel, then the self-learning level of rough rolling width of similar strip steel is used to adjust the rough rolling width expansion coefficient of the strip steel of this level;

[0011] S4. The strip rough rolling width expansion coefficient is written into the layer strategy table after being adjusted by divergent self-learning, and the modification time of the layer strategy table is refreshed to the current time.

[0012] In S1, the rough rolling width self-learning layer index of the strip includes the target thickness, target width, slab width, steel family and rolling mode of the strip.

[0013] In S3, the self-learning layers of the rough rolling width of similar strip steels are: the hardness layers are the same, the width is within one adjacent layer, the width reduction is within two adjacent layers, and the production date is within the set time.

[0014] In S3, the rough rolling width expansion global correction coefficient adjusts the rough rolling width expansion coefficient of the strip of this layer. The calculation formula is as follows:

[0015] CORWID0_new=iCORWID0+ADDCORWID ①

[0016] In formula ①, CORWID0_new represents the rough rolling width expansion coefficient (CORWID0) value after adjustment; iCORWID0 represents the rough rolling width expansion coefficient (CORWID0) value before adjustment; ADDCORWID represents the rough rolling width expansion global correction coefficient value.

[0017] In S3, the rough rolling width of the adjacent strip steel is self-learned to adjust the rough rolling width expansion coefficient of the strip steel of the layer, which includes two parts, as follows:

[0018] The width of the first part is within an adjacent layer, the width reduction is within an adjacent layer, the production date is within the set time, and the self-learning values ​​of the adjacent layers in the first part are averaged to obtain the average rough rolling width expansion coefficient of the adjacent layers of the first part;

[0019] The width of the second part is within one adjacent layer, the width reduction is within two adjacent layers, and the production date is within the set time. The self-learning values ​​of the adjacent layers in the second part are averaged to obtain the average rough rolling width expansion coefficient of the adjacent layers in the second part;

[0020] If both parts have data, the first part accounts for 0.8, otherwise the part with data accounts for 1.

[0021] In S3, the rough rolling width of the similar strip is self-learned at the level to adjust the rough rolling width expansion coefficient of the strip at that level. The calculation formula is as follows:

[0022] CORWID0_new= a*CORWID0_1+(1-a)*CORWID0_2 ②

[0023] In formula ②, CORWID0_new represents the adjusted rough rolling width expansion coefficient (CORWID0) value; CORWID0_1 represents the average rough rolling width expansion coefficient of adjacent layers in the first part; CORWID0_2 represents the average rough rolling width expansion coefficient of adjacent layers in the second part; a represents the proportion coefficient of the first part. If CORWID0_1 has no value, a is forced to be 0. If CORWID0_2 has no value, a is forced to be 1. If both parts have data, a takes the value of 0.8.

[0024] The width of each layer spans 850~950mm, 950~1050mm...greater than 1065mm, and the width reduction of each layer spans <-35mm,...-5~5mm, 5~15mm...,>55mm.

[0025] Compared with the prior art, the present invention has the following beneficial effects:

[0026] 1. If the rough rolling width expansion coefficient of the strip steel of the layer is not suitable for the current working condition, the rough rolling width expansion coefficient of the strip steel of the layer is readjusted through divergent self-learning to solve the problem that when the rough rolling width expansion coefficient is divided into fine layers, the adaptability to the working condition is poor due to the low self-learning frequency;

[0027] 2. Retrieve the 'Last Update Time' parameter in the rough rolling width self-learning layer of the strip, and confirm the time difference between the last self-learning of the rough rolling width self-learning layer of this strip and the current time. If the time difference is higher than the set time, then enter the divergent self-learning. Divergent self-learning is used to solve the problem of low self-learning frequency and poor adaptability to working conditions. By comparing the time difference with the set time, it can be determined whether the rough rolling width expansion coefficient of this layer needs to be adjusted through divergent self-learning;

[0028] 3. The rough rolling width expansion coefficient of the strip steel of this layer is adjusted by using the rough rolling width expansion global correction coefficient. When a suitable reference layer cannot be found in this layer, the rough rolling width expansion coefficient can be fine-tuned by relying on the global correction coefficient;

[0029] 4. Use the rough rolling width self-learning layer of similar strip steel to adjust the rough rolling width expansion coefficient of the strip steel of this layer. The similar strip steel layer found according to this scheme has similar materials and specifications with this layer, and has a high similarity with the rough rolling width expansion coefficient of this layer. On the other hand, the similar strip steel layer found according to this scheme has undergone self-learning recently and can better match the current working conditions.

[0030] 5. After the layers that had not been rolled for a long time and the vertical rollers were replaced, the width control situation in the first few days was significantly improved. It was not due to equipment failure or incoming material reasons, which greatly reduced the situation of excessive width and narrow size of the plate.

[0031] 6. Through the application of divergent self-learning function, the influence of vertical roller replacement, temperature and other on-site working conditions on width control can be alleviated to a certain extent. The influence of inaccurate calculation of vertical roller wear on strip width control can also be alleviated to a certain extent. Standardized processes avoid excessive manual intervention and reduce the workload of maintenance personnel. BRIEF DESCRIPTION OF THE DRAWINGS

[0032] Figure 1 It is a logical flow chart for improving the width accuracy of the first coil of steel when changing specifications.

[0033] Figure 2 It is a flow chart calculated according to similar levels.

[0034] Figure 3 It is a schematic diagram for storing the global correction coefficient of rough rolling width expansion.

[0035] Figure 4 It is the actual control width curve.

[0036] Figure 5 It is the intercepted log record of the rough rolling model calculation. DETAILED DESCRIPTION

[0037] The present invention is described in detail below in conjunction with the accompanying drawings, but it should be noted that the implementation of the present invention is not limited to the following embodiments.

[0038] The following examples are implemented on the premise of the technical solution of the present invention, and provide detailed implementation methods and specific operation processes, but the protection scope of the present invention is not limited to the following examples. The methods used in the following examples are conventional methods unless otherwise specified.

[0039] [Example 1]

[0040] See Figure 1 ,See Figure 2 , a method for improving the width accuracy of the first coil of steel after specification change. If the layer to which the strip rough rolling width expansion coefficient belongs has not been updated within the set time or after the vertical roll is changed, then the strip rough rolling width expansion coefficient of this layer is not suitable for the current working conditions. Through divergent self-learning, the strip rough rolling width expansion coefficient of this layer is readjusted; divergent self-learning, the steps are as follows:

[0041] S1. Obtain the target thickness, target width, slab width (calculate the width reduction), steel family and rolling mode (the number of passes in which each stand participates in the rough rolling process) of the strip, and determine the self-learning level of the rough rolling width of the strip based on the above information.

[0042] S2. Retrieve the 'last update time' parameter in the rough rolling width self-learning layer of the strip, and confirm the time difference between the last self-learning of the rough rolling width self-learning layer of this strip and the current time. If the time difference is greater than twenty days, then enter divergent self-learning; divergent self-learning refers to a design method in which similar layers have linkage self-learning behaviors in multi-layer self-learning calculations to improve self-learning efficiency. Divergent self-learning is only called when the rough rolling furnace is set to avoid multiple adjustments to the self-learning parameters, resulting in uncontrollable parameter adjustment range.

[0043] S3. Find the self-learning level of rough rolling width of similar strip steel. The standards of the self-learning level of rough rolling width of similar strip steel are the same hardness level, the width is within one adjacent level, the width reduction is within two adjacent levels, and the production date is within 20 days; the width span of one level is 100mm, and the width reduction span of one level is 10mm;

[0044] If there is no similar strip roughing width self-learning layer, then the roughing width expansion coefficient of the strip of this layer is adjusted by using the roughing width expansion global correction coefficient. The calculation formula is as follows:

[0045] CORWID0_new=iCORWID0+ADDCORWID ①

[0046] In formula ①, CORWID0_new represents the adjusted rough rolling width global correction coefficient (CORWID0); iCORWID0 represents the rough rolling width global correction coefficient (CORWID0) before adjustment; ADDCORWID represents the global correction value;

[0047] If there is a self-learning level of rough rolling width of similar strip steel, then the rough rolling width expansion coefficient of the strip steel of this level is adjusted by using the self-learning level of rough rolling width of similar strip steel, which includes two parts as follows:

[0048] The width of the first part is within one adjacent layer, the width reduction is within one adjacent layer, and the production date is within ten days. The self-learning values ​​of the adjacent layers in the first part are averaged to obtain the average rough rolling width expansion coefficient of the adjacent layers in the first part;

[0049] The width of the second part is within one adjacent layer, the width reduction is within two adjacent layers, and the production date is within 20 days. The self-learning values ​​of the adjacent layers in the second part are averaged to obtain the average rough rolling width expansion coefficient of the adjacent layers in the second part;

[0050] If both parts have data, the first part accounts for 0.8, otherwise the part with data accounts for 1;

[0051] The rough rolling width of similar strip steel is adjusted by self-learning layer to adjust the rough rolling width expansion coefficient of the strip steel of this layer. The calculation formula is as follows:

[0052] CORWID0_new= a*CORWID0_1+(1-a)*CORWID0_2 ②

[0053] In formula ②, CORWID0_new represents the adjusted rough rolling width expansion coefficient (CORWID0) value; CORWID0_1 represents the average rough rolling width expansion coefficient of adjacent layers in the first part; CORWID0_2 represents the average rough rolling width expansion coefficient of adjacent layers in the second part; a represents the proportion coefficient of the first part. If CORWID0_1 has no value, a is forced to be 0. If CORWID0_2 has no value, a is forced to be 1. If both parts have data, a takes the value of 0.8.

[0054] S4. The strip rough rolling width expansion coefficient is adjusted by divergent self-learning and written into the layer strategy table. The modification time of the layer strategy table is refreshed to the current time, which again avoids the divergent self-learning from being called multiple times.

[0055] [Example 2]

[0056] In this embodiment, a method for improving the width accuracy of the first steel roll after specification change is the same as that in Embodiment 1, and a global correction value calculation process is added thereto.

[0057] When a strip is rolled, it is determined whether the strip is the first 100 strips after the vertical roll is replaced. If so, the global correction value is calculated, otherwise it is not calculated. The formula is as follows:

[0058] ADDCORWID=ADDCORWID_old*(n-1) / n+CORWID_DEV / n ③

[0059] In formula ③, ADDCORWID represents the global correction value, ADDCORWID_old represents the original value of the global correction value, n represents the number of blocks after the roll change, and CORWID_DEV represents the self-learning deviation of this block of steel;

[0060] Write the calculation results into the corresponding table in the database. Figure 3 .

[0061] The global correction value is reset when the vertical roller is replaced.

[0062] If divergent self-learning is not used, the rough rolling width expansion coefficient used for the rough rolling layer of a strip steel is 98-3-4-8, and the last rolling of this layer was 121 days ago. The working conditions may have changed greatly during this period. Using this rough rolling width expansion coefficient setting, the final total side pressure of the rough rolling is 176.62mm.

[0063] Through the divergent self-learning function, a total of 77 similar layers were found, of which 42 were in line with the first part and 35 were in line with the second part, see Table 2.

[0064] Table 2 is a list of similar strata.

[0065]

[0066] In the first part, the rough rolling width self-learning parameters of similar layers are set as AW 101 ~AW 142 , n1 is the number of blocks, which is 42 in this example, then the average value is calculated as follows:

[0067] CORWID0_1=ΣAW 1i / n1 ④

[0068] According to the above data, CORWID0_1=0.163308.

[0069] Similarly, the rough rolling width self-learning parameters of the second part of similar layers are set to AW 201 ~AW 235 , n2 is the number of blocks, which is 35 in this example.

[0070] The formula for calculating the average value is:

[0071] CORWID0_2=ΣAW 2i / n2 ⑤

[0072] According to the above data, CORWID0_2=0.145027.

[0073] According to the formula of divergent self-learning, the formula is as follows:

[0074] CORWID0_new= a*CORWID0_1+(1-a)*CORWID0_2 ①

[0075] The rough rolling width self-learning parameter CORWID0_new is calculated to be '0.160'.

[0076] In practical applications, in order to make the rough rolling width self-learning parameters adjusted by divergent self-learning more conspicuous, '0.000111' is added to the tail. This value will not affect the final width control accuracy. The final rough rolling width self-learning parameter adjustment result is '0.160111'. Figure 5 .

[0077] By using the adjusted rough rolling width expansion coefficient, the final median side pressure is 165.13mm, the total side pressure is reduced by about 11.5mm, and the overall width hit is good. Figure 4 If produced according to the original rough rolling width expansion coefficient, the strip may be 3mm narrower.

[0078] Generally speaking, if you want the self-learning parameters to be adjusted more finely, you need to refine the self-learning parameter levels. However, if you want the self-learning parameters to be able to adapt to the inevitable changes in working conditions, you need to coarsen the self-learning parameter levels. Self-learning parameters without levels are obviously the most adaptable to changes in working conditions. The technical solution provided by the present invention is a solution that seeks a balance between finely adjusting the rough rolling width expansion coefficient and improving the adaptability of width control to working conditions, which ensures the refinement of the rough rolling width expansion coefficient and greatly improves the adaptability of the parameters to working conditions.

[0079] The idea of ​​this technical solution can also be adopted in the process of optimizing and adjusting other quality self-learning coefficients, but the adjustment range needs to be changed according to different self-learning coefficient properties.

[0080] If the rough rolling width expansion coefficient of the steel strip of a layer is not suitable for the current working condition, the present invention readjusts the rough rolling width expansion coefficient of the steel strip of this layer through divergent self-learning, so as to solve the problem that when the rough rolling width expansion coefficient is divided into finer layers, the adaptability to the working condition is poor due to the low self-learning frequency; the 'last update time' parameter in the rough rolling width self-learning layer of the steel strip is called up to confirm the time difference between the last self-learning of the rough rolling width self-learning layer of this steel strip and the current time; if the time difference is higher than the set time, then the divergent self-learning is entered, and the divergent self-learning is used to solve the problem of poor adaptability to the working condition caused by the low self-learning frequency; by comparing the time difference with the set time, it can be determined whether the rough rolling width expansion coefficient of this layer needs to be adjusted through divergent self-learning. The method is simple, easy to use and easy to promote; the rough rolling width expansion global correction coefficient is used to adjust the rough rolling width expansion coefficient of the steel strip of this layer, and when the layer cannot find a suitable reference layer , relying on the global correction coefficient, the rough rolling width expansion coefficient can be fine-tuned; the rough rolling width self-learning layer of similar strip steel is used to adjust the rough rolling width expansion coefficient of the strip steel of this layer. The similar strip steel layer found according to this scheme has similar materials and specifications with this layer on the one hand, and has a high similarity with the rough rolling width expansion coefficient of this layer. On the other hand, the similar strip steel layer found according to this scheme has undergone self-learning recently and can better match the current working conditions; after the layer that has not been rolled for a long time and the vertical roller has been replaced, the width control situation in the first few days has been significantly improved. The situation of super-wide and narrow size of the plate has been greatly reduced due to non-equipment failure or incoming material reasons; through the application of the divergent self-learning function, the influence of the on-site working conditions such as vertical roller replacement and temperature on the width control can be alleviated to a certain extent, and the influence of inaccurate calculation of vertical roller wear on the strip width control can also be alleviated to a certain extent. Standardized processes avoid excessive manual intervention and reduce the workload of maintenance personnel.

Claims

1. A method for improving the width accuracy of the first roll of steel when changing specifications. It is characterized in that If the layer to which the strip rough rolling width expansion coefficient belongs has not been updated within the set time or after the vertical roll is replaced, the strip rough rolling width expansion coefficient of this layer is not suitable for the current working conditions. Through divergent self-learning, the strip rough rolling width expansion coefficient of this layer is readjusted.

2. A method for improving the width accuracy of the first steel roll when changing specifications according to claim 1, It is characterized in that The divergent self-learning steps are as follows: S1, obtain the rough rolling width of the strip by self-learning level; S2. Retrieve the 'Last Update Time' parameter in the rough rolling width self-learning layer of the strip steel, and confirm the time difference between the last self-learning of the rough rolling width self-learning layer of this strip steel and the current time. If the time difference is higher than the set time, then enter the divergent self-learning; S3, finding a self-learning level of rough rolling width of similar strip steel, if there is no self-learning level of rough rolling width of similar strip steel, then the rough rolling width expansion coefficient of the strip steel of this level is adjusted by using the global correction coefficient of rough rolling width expansion; if there is a self-learning level of rough rolling width of similar strip steel, then the self-learning level of rough rolling width of similar strip steel is used to adjust the rough rolling width expansion coefficient of the strip steel of this level; S4. The strip rough rolling width expansion coefficient is written into the layer strategy table after being adjusted by divergent self-learning, and the modification time of the layer strategy table is refreshed to the current time.

3. A method for improving the width accuracy of the first steel roll when changing specifications according to claim 2, It is characterized in that In S1, the rough rolling width self-learning layer index of the strip includes the target thickness, target width, slab width, steel family and rolling mode of the strip.

4. A method for improving the width accuracy of the first steel roll when changing specifications according to claim 2, It is characterized in that In S3, the rough rolling width self-learning layers of the similar strip steels are: the hardness layers are the same, the width is within one adjacent layer, the width reduction is within two adjacent layers, and the production date is within the set time.

5. A method for improving the width accuracy of the first steel roll when changing specifications according to claim 2, It is characterized in that In S3, the rough rolling width expansion global correction coefficient is used to adjust the rough rolling width expansion coefficient of the strip steel of the layer, and the calculation formula is as follows: CORWID0_new=iCORWID0+ADDCORWID ① In formula ①, CORWID0_new represents the rough rolling width expansion coefficient (CORWID0) value after adjustment; iCORWID0 represents the rough rolling width expansion coefficient (CORWID0) value before adjustment; ADDCORWID represents the rough rolling width expansion global correction coefficient value.

6. A method for improving the width accuracy of the first steel roll when changing specifications according to claim 2, It is characterized in that In S3, the rough rolling width of the similar strip steel is self-learned at the level to adjust the rough rolling width expansion coefficient of the strip steel at the level, including two parts, as follows: The width of the first part is within an adjacent layer, the width reduction is within an adjacent layer, the production date is within the set time, and the self-learning values ​​of the adjacent layers in the first part are averaged to obtain the average rough rolling width expansion coefficient of the adjacent layers of the first part; The width of the second part is within one adjacent layer, the width reduction is within two adjacent layers, and the production date is within the set time. The self-learning values ​​of the adjacent layers in the second part are averaged to obtain the average rough rolling width expansion coefficient of the adjacent layers in the second part; If both parts have data, the first part accounts for 0.8, otherwise the part with data accounts for 1.

7. A method for improving the width accuracy of the first steel roll when changing specifications according to claim 6, It is characterized in that In S3, the rough rolling width of the similar strip is self-learned at the level to adjust the rough rolling width expansion coefficient of the strip at that level. The calculation formula is as follows: CORWID0_new= a*CORWID0_1+(1-a)*CORWID0_2 ② In formula ②, CORWID0_new represents the adjusted rough rolling width expansion coefficient (CORWID0) value; CORWID0_1 represents the average rough rolling width expansion coefficient of adjacent layers in the first part; CORWID0_2 represents the average rough rolling width expansion coefficient of adjacent layers in the second part; a represents the proportion coefficient of the first part. If CORWID0_1 has no value, a is forced to be 0. If CORWID0_2 has no value, a is forced to be 1. If both parts have data, a takes the value of 0.

8.

8. A method for improving the width accuracy of the first steel roll when changing specifications according to claim 4, It is characterized in that The width of each layer spans 850~950mm, 950~1050mm...greater than 1065mm, and the width reduction of each layer spans <-35mm,...-5~5mm, 5~15mm...,>55mm.

Images