A deformation resistance self-learning optimization method

Abstract

The present application relates to a kind of deformation resistance self-learning optimization method, comprising the following steps: creating deformation temperature and strain rate as index two-dimensional self-learning coefficient table with steel race as unit;Predicted deformation temperature and strain rate are obtained by finish rolling pre-computation model, determine center cell index;The interpolation influence value of adjacent cell (same temperature or strain rate) and diagonal cell (temperature and strain rate are adjacent) to target point is calculated respectively, and the self-learning coefficient of center cell, adjacent cell and diagonal cell is fused by weighted average, and finally generate optimized self-learning value. Center cell weight is highest, adjacent cell is second, and diagonal cell is lowest. The present application effectively reduces the deviation caused by layer jump of self-learning coefficient by multidimensional interpolation and dynamic weighting strategy, significantly improves the accuracy and stability of deformation resistance model, so as to improve the accuracy of rolling force setting, and guarantee the quality and efficiency of hot-rolled strip production.

Description

Technical Field

[0001] This patent application belongs to the field of hot-rolled strip steel production technology in the metallurgical industry, and more specifically, it relates to a self-learning optimization method for deformation resistance. Background Technology

[0002] In the hot-rolled strip steel production process of the metallurgical industry, the accuracy of the rolling force model directly determines the roll gap setting, strip shape control, and strip threading stability, making it a core technical link to ensure product quality and production efficiency. Among these, the deformation resistance model, as a key component of the rolling force model, is crucial to the reliability of rolling force calculations. With increasing competition in the steel market, downstream users have increasingly stringent requirements for product dimensional accuracy and surface quality. Simultaneously, the production model of small batches, multiple varieties, and multiple specifications is becoming increasingly common, posing greater challenges to the adaptability and stability of the deformation resistance model.

[0003] Traditional deformation resistance self-learning methods typically operate on a steel grade basis, establishing a two-dimensional index table using predefined deformation temperature and strain rate layers. The self-learning coefficients are dynamically adjusted for the rolling parameters of each strip coil. However, in actual production, upstream process fluctuations (such as furnace temperature deviations and roughing mill exit thickness variations) and dynamic adjustments to the load distribution across the finishing mill stands often cause the predicted deformation temperature and strain rate to deviate from the preset layer range. This deviation triggers "layer jumps" in the self-learning coefficients, meaning the same steel grade frequently switches to different index layers during continuous rolling due to parameter fluctuations. Since the self-learning coefficients of adjacent layers may differ significantly, these jumps can lead to abrupt changes in the rolling force setpoint, resulting in production problems such as strip shape fluctuations, thickness deviations, and even strip breakage, severely hindering the achievement of high-precision rolling.

[0004] Furthermore, existing algorithms are mostly limited to the current layer and fail to fully coordinate the multidimensional relationships between central, adjacent, and diagonal units. This simplification can easily lead to uneven transitions in self-learning values ​​when parameters change continuously, further exacerbating the instability of the model output. Therefore, how to achieve smooth optimization of self-learning coefficients under complex working conditions and reduce the interference of jumps on model accuracy has become a pressing technical challenge to improve the production control level of hot-rolled strip steel. Summary of the Invention

[0005] The technical problem to be solved by this invention is to propose a deformation resistance self-learning optimization method that integrates multidimensional interpolation and dynamic weighting strategies. The aim is to fully coordinate the multidimensional correlation between the central unit, adjacent units and diagonal units through refined data processing and weight allocation mechanism, significantly improve the model adaptability, reduce the interference of layer jump on the model accuracy, and thus meet the needs of modern steel industry for high-precision and high-stability rolling control.

[0006] To solve the above problems, the technical solution adopted by the present invention is as follows:

[0007] A self-learning optimization method for deformation resistance includes the following steps:

[0008] S1. Create a self-learning coefficient table for deformation resistance based on steel type. The table is indexed by deformation temperature and strain rate in two dimensions to save the self-learning coefficients of deformation resistance for different layers.

[0009] S2. Design a pre-calculation model for the finishing mill, and obtain the predicted deformation temperature of each stand in the finishing mill unit through the pre-calculation model. With strain rate The coordinate point ( , This is called point P; based on the deformation temperature The deformation temperature index j is determined by the range of the interval, based on the strain rate. The range of the interval determines the strain rate index i value, and the corresponding layer is determined by the two-dimensional index.

[0010] S3. Determine k(i,j) as the central unit according to the layer where point P is located, and find the adjacent units corresponding to the central unit. The adjacent units are those where one of the strain rate index i or the deformation temperature index j is the same as the central unit, and the other is 1 different from the central unit index. There are a total of 4 adjacent units, among which there are 2 units with the same deformation temperature index j: k(i-1,j) and k(i+1,j), and there are also 2 units with the same strain rate index i: k(i,j-1) and k(i,j+1).

[0011] S4. Calculate the value of point P using the self-learning value of adjacent cells. , The self-learning influence value of the unit is used to obtain the self-learning coefficient of the adjacent unit.

[0012] S5. Using k(i,j) as the central unit, find the corresponding diagonal units. The diagonal units are units whose two-dimensional index values ​​differ from the central unit by 1. There are 4 diagonal units in total: k(i-1,j-1), k(i-1,j+1), k(i+1,j-1), k(i+1,j+1).

[0013] S6. Calculate the value at point P using the self-learning value of the diagonal unit. , The self-learning influence value of the diagonal unit is used to obtain the self-learning coefficient.

[0014] S7. Calculate the interpolation influence of adjacent elements (with the same temperature or strain rate) and diagonal elements (with adjacent temperatures and strain rates) on the target point, and then use a weighted average to fuse the self-learning coefficients of the central element, adjacent elements and diagonal elements according to different weights to finally generate the optimized self-learning value.

[0015] Further: Step S4 is specifically performed as follows:

[0016] S41. First, calculate the same deformation temperature. Different strain rates At point P Directional self-learning value The calculation is performed using interpolation as shown in equation (1):

[0017]

[0018] S42, Next, calculate at the same strain rate Different deformation temperatures At point P Directional self-learning value The calculation is performed using interpolation as shown in equation (2):

[0019]

[0020] S43. Calculate the results of the above two items. and Calculate the average value , as the self-learning coefficient of adjacent units;

[0021] .

[0023] Further: Step S6 is specifically performed as follows:

[0024] S61, Firstly, at the deformation temperature... and Under the condition of different strain rates at point P The self-learning values ​​for the directions are respectively and The interpolation method is used to calculate equations (4) and (5);

[0025]

[0026]

[0027] S62, Deformation temperature and Temperature calculated under the given conditions and At the same strain rate Different deformation temperatures Perform the difference calculation again to obtain the diagonal element in ( , Self-learning coefficient :

[0028] .

[0029] Further: Step S7 specifically refers to:

[0030] The self-learning value of the central unit k(i,j) and the self-learning coefficients of adjacent units are integrated. and diagonal unit self-learning coefficient The calculation is performed by weighting the values ​​according to different weights, and the value of point P is calculated. , Self-learning value The weight coefficient of the central unit is The weight coefficient of adjacent units is The weighting coefficient of the diagonal unit is , > > In other words, the central unit has the highest weight, followed by adjacent units, and the diagonal units have the lowest weight. When an adjacent unit or diagonal unit has a self-learning coefficient of 1 (initial value), the weight coefficients need to be dynamically adjusted. and The corresponding weight value is set to 0, that is, the influence of this item on the central unit is ignored;

[0031] at the same time, , , The relationship is shown in equation (7), the self-learning value As shown in equation (8)

[0032]

[0033] .

[0034] further: .

[0035] Due to the adoption of the above technical solution, the beneficial effects achieved by this invention are:

[0036] In hot rolling production, fluctuations in upstream processes (such as furnace temperature deviations and changes in roughing mill exit thickness) and dynamic adjustments to the load distribution of each stand in the finishing mill often cause the predicted deformation temperature and strain rate to deviate from the preset layer range. This deviation can trigger "layer jumps" in the self-learning coefficients, meaning that the same steel grade frequently switches to different index layers during continuous rolling due to parameter fluctuations. To address this issue, a deformation resistance self-learning optimization method integrating multidimensional interpolation and dynamic weighting strategies is proposed. This method fully coordinates the multidimensional correlations between central elements, adjacent elements, and diagonal elements, reducing the interference of layer jumps on model accuracy.

[0037] This invention significantly improves the model's adaptability through refined data processing and weight allocation mechanisms, achieving smooth optimization of self-learning coefficients under complex working conditions, thereby meeting the modern steel industry's demand for high-precision and high-stability rolling control. Attached Figure Description

[0038] Figure 1 This is a diagram showing the arrangement of the deformation resistance self-learning coefficients of the present invention.

[0039] Figure 2 This is an example diagram of the deformation resistance self-learning coefficient of the present invention. Detailed Implementation

[0040] The present invention will be further described in detail below with reference to the embodiments.

[0041] A self-learning optimization method for deformation resistance, such as Figure 1 , Figure 2 As shown, Figure 1 The blue circle represents the central unit, the yellow circle represents the adjacent unit, and the green circle represents the diagonal unit; Figure 2 The median value represents the self-learning coefficient of the unit at different positions.

[0042] Specifically, the steps include the following:

[0043] S1. Create a self-learning coefficient table for deformation resistance based on steel type. The table is indexed by deformation temperature and strain rate in two dimensions to save the self-learning coefficients of deformation resistance for different layers.

[0044] S2. Design a pre-calculation model for the finishing mill, and obtain the predicted deformation temperature of each stand in the finishing mill unit through the pre-calculation model. With strain rate The coordinate point ( , This is called point P; based on the deformation temperature The deformation temperature index j is determined by the range of the interval, based on the strain rate. The range of the interval determines the strain rate index i value, and the corresponding layer is determined by the two-dimensional index.

[0045] S3. Determine k(i,j) as the central unit according to the layer where point P is located, and find the adjacent units corresponding to the central unit. The adjacent units are those where one of the strain rate index i or the deformation temperature index j is the same as the central unit, and the other is 1 different from the central unit index. There are a total of 4 adjacent units, among which there are 2 units with the same deformation temperature index j: k(i-1,j) and k(i+1,j), and there are also 2 units with the same strain rate index i: k(i,j-1) and k(i,j+1).

[0046] S4. Calculate the value of point P using the self-learning value of adjacent cells. , The self-learning influence value of the unit is used to obtain the self-learning coefficient of the adjacent unit.

[0047] S5. Using k(i,j) as the central unit, find the corresponding diagonal units. The diagonal units are units whose two-dimensional index values ​​differ from the central unit by 1. There are 4 diagonal units in total: k(i-1,j-1), k(i-1,j+1), k(i+1,j-1), k(i+1,j+1).

[0048] S6. Calculate the value at point P using the self-learning value of the diagonal unit. , The self-learning influence value of the diagonal unit is used to obtain the self-learning coefficient.

[0049] S7. Calculate the interpolation influence of adjacent elements (with the same temperature or strain rate) and diagonal elements (with adjacent temperatures and strain rates) on the target point, and then use a weighted average to fuse the self-learning coefficients of the central element, adjacent elements and diagonal elements according to different weights to finally generate the optimized self-learning value.

[0050] Step S4 is performed as follows:

[0051] S41. First, calculate the same deformation temperature. Different strain rates At point P Directional self-learning value The calculation is performed using interpolation as shown in equation (1):

[0052]

[0053] S42, Next, calculate at the same strain rate Different deformation temperatures At point P Directional self-learning value The calculation is performed using interpolation as shown in equation (2):

[0054]

[0055] S43. Calculate the results of the above two items. and Calculate the average value , as the self-learning coefficient of adjacent units;

[0056] .

[0058] Step S6 is performed as follows:

[0059] S61, Firstly, at the deformation temperature... and Under the condition of different strain rates at point P The self-learning values ​​for the directions are respectively and The interpolation method is used to calculate equations (4) and (5);

[0060]

[0061]

[0062] S62, Deformation temperature and Temperature calculated under the given conditions and At the same strain rate Different deformation temperatures Perform the difference calculation again to obtain the diagonal element in ( , Self-learning coefficient :

[0063] .

[0064] Step S7 specifically refers to:

[0065] The self-learning value of the central unit k(i,j) and the self-learning coefficients of adjacent units are integrated. and diagonal unit self-learning coefficient The calculation is performed by weighting the values ​​according to different weights, and the value of point P is calculated. , Self-learning value The weight coefficient of the central unit is The weight coefficient of adjacent units is The weighting coefficient of the diagonal unit is , > > In other words, the central unit has the highest weight, followed by adjacent units, and the diagonal units have the lowest weight. When an adjacent unit or diagonal unit has a self-learning coefficient of 1 (initial value), the weight coefficients need to be dynamically adjusted. and The corresponding weight value is set to 0, that is, the influence of this item on the central unit is ignored;

[0066] at the same time, , , The relationship is shown in equation (7), the self-learning value As shown in equation (8)

[0067]

[0068] .

[0069] Regarding the value, .

[0070] The relevant values ​​for temperature index, temperature range, and strain rate index are shown in the table below.

[0071]

[0072] The following examples illustrate this in detail:

[0073] Example 1: Taking rolled piece 1 as an example; the deformation resistance self-learning coefficient table has been created, and the table is indexed by deformation temperature and strain rate respectively, saving the deformation resistance self-learning coefficients of different layers;

[0074] The predicted deformation temperature of the F1 stand of the finishing mill was obtained using a pre-calculation model of the finishing mill. =962℃ and strain rate =7.213 (1 / sec), this coordinate point ( , Point P is defined by index j=11 based on the range of deformation temperature and index i=7 based on the range of strain rate. The corresponding layer is determined by the two-dimensional index.

[0075] 1. Determine k(7, 11) as the central element based on the layer where point P is located. Find the adjacent elements of the central element. The adjacent elements are those whose strain rate index or deformation temperature index is consistent with the central element. There are a total of 4 adjacent elements, including 2 with the same temperature index: k(6, 11) and k(8, 11), and 2 with the same strain rate index: k(7, 10) and k(7, 12).

[0076] (1) Calculate the value of point P using the self-learning value of adjacent cells. , The self-learning influence value is first calculated at point P under the same deformation temperature and different strain rates. Directional self-learning value The calculation is performed using interpolation as shown in equation (1):

[0077]

[0078]

[0079] (2) Next, calculate the results at point P under the same strain rate and different deformation temperatures. Directional self-learning value The calculation is performed using interpolation as shown in equation (2):

[0080]

[0081]

[0082] (3) Combine the above two calculation results and Calculate the average value , as the self-learning influence value of adjacent units;

[0083]

[0084]

[0085] 2. Using k(i, j) as the center, find the diagonal cells. The diagonal cells are those whose two-dimensional index values ​​differ from the center cell by 1. There are 4 diagonal cells in total: k(i-1, j-1), k(i-1, j+1), k(i+1, j-1), and k(i+1, j+1).

[0086] 1) Calculate the coordinates at the point using the self-learning value of the diagonal unit ( , The value of the position is first determined at the deformation temperature. and Under the condition of different strain rates at point P The self-learning values ​​for the directions are respectively and The interpolation method is used to calculate equations (4) and (5);

[0087]

[0088]

[0089]

[0090] 0.98072

[0091] 2) Deformation temperature and Temperature calculated under the given conditions and At the same strain rate but different deformation temperatures, the difference calculation is performed again to obtain the diagonal element in ( , The self-learning impact value :

[0092]

[0093]

[0094] 3. Integrate the self-learning value of the central unit k(i,j) and the influence value of adjacent units. Influence value of diagonal elements The calculation is performed by weighting the values ​​according to different weights, and the value of point P is calculated. , Self-learning value The weight of the central unit Adjacent cell weights Diagonal unit weights That is, take out ;

[0095]

[0096]

[0097]

[0098] The above embodiments are only used to illustrate and not limit the technical solutions of the present invention. Although the present invention has been described in detail with reference to the above embodiments, those skilled in the art should understand that modifications or equivalent substitutions can still be made to the present invention without departing from the spirit and scope of the present invention. Any modifications or partial substitutions should be covered within the scope of the claims of the present invention.

Claims

1. A self-learning optimization method for deformation resistance, characterized in that... Includes the following steps: S1. Create a self-learning coefficient table for deformation resistance based on steel type. The table is indexed by deformation temperature and strain rate in two dimensions to save the self-learning coefficients of deformation resistance for different layers. S2. Design a pre-calculation model for the finishing mill, and obtain the predicted deformation temperature of each stand in the finishing mill unit through the pre-calculation model. With strain rate The coordinate point ( , This is called point P; based on the deformation temperature The deformation temperature index j is determined by the range of the interval, based on the strain rate. The range of the interval determines the strain rate index i value, and the corresponding layer is determined by the two-dimensional index. S3. Determine k(i,j) as the central element according to the layer where point P is located, and find the adjacent elements corresponding to the central element. The adjacent elements are those where one of the strain rate index i or the deformation temperature index j is the same as the central element, and the other is 1 different from the central element index. There are a total of 4 adjacent elements, of which 2 have the same deformation temperature index j: k(i-1,j) and k(i+1,j), and 2 have the same strain rate index i: k(i,j-1) and k(i,j+1). S4. Calculate the value of point P using the self-learning value of adjacent cells. , The self-learning influence value of the unit is used to obtain the self-learning coefficient of the adjacent unit. S5. Using k(i,j) as the central unit, find the corresponding diagonal units. The diagonal units are units whose two-dimensional index values ​​differ from the central unit by 1. There are 4 diagonal units in total: k(i-1,j-1), k(i-1,j+1), k(i+1,j-1), k(i+1,j+1). S6. Calculate the value at point P using the self-learning value of the diagonal unit. , The self-learning influence value of the diagonal unit is used to obtain the self-learning coefficient. S7. Calculate the interpolation influence of adjacent and diagonal units on the target point, and then use a weighted average to fuse the self-learning coefficients of the central unit, adjacent units, and diagonal units according to different weights to finally generate the optimized self-learning value.

2. The deformation resistance self-learning optimization method according to claim 1, characterized in that: Step S4 is performed as follows: S41. First, calculate the same deformation temperature. Different strain rates At point P Directional self-learning value The calculation is performed using interpolation as shown in equation (1): ； S42, Next, calculate at the same strain rate Different deformation temperatures At point P Directional self-learning value The calculation is performed using interpolation as shown in equation (2): ； S43. Calculate the results of the above two items. and Calculate the average value , as the self-learning coefficient of adjacent units; 。 3. The deformation resistance self-learning optimization method according to claim 2, characterized in that: Step S6 is performed as follows: S61, Firstly, at the deformation temperature... and Under the condition of different strain rates at point P The self-learning values ​​for the directions are respectively and The interpolation method is used to calculate equations (4) and (5); ； ； S62, Deformation temperature and Temperature calculated under the given conditions and At the same strain rate Different deformation temperatures Perform the difference calculation again to obtain the diagonal element in ( , Self-learning coefficient : 。 4. The deformation resistance self-learning optimization method according to claim 2, characterized in that: Step S7 specifically refers to: The self-learning value of the central unit k(i,j) and the self-learning coefficients of adjacent units are integrated. and diagonal unit self-learning coefficient The calculation is performed by weighting the values ​​according to different weights, and the value of point P is calculated. , Self-learning value The weight coefficient of the central unit is The weight coefficient of adjacent units is The weighting coefficient of the diagonal unit is , > > When a self-learning coefficient of 1 occurs in adjacent or diagonal cells, the weighting coefficients need to be dynamically adjusted. and The corresponding weight value is set to 0, that is, the influence of this item on the central unit is ignored; at the same time, , , The relationship is shown in equation (7), the self-learning value As shown in equation (8) ； 。 5. The deformation resistance self-learning optimization method according to claim 1, characterized in that: 。

Images