CN111318579A - Roughing camber control method based on data driving - Google Patents

Roughing camber control method based on data driving Download PDF

Info

Publication number
CN111318579A
CN111318579A CN202010149682.8A CN202010149682A CN111318579A CN 111318579 A CN111318579 A CN 111318579A CN 202010149682 A CN202010149682 A CN 202010149682A CN 111318579 A CN111318579 A CN 111318579A
Authority
CN
China
Prior art keywords
data
value
camber
pass
head
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Granted
Application number
CN202010149682.8A
Other languages
Chinese (zh)
Other versions
CN111318579B (en
Inventor
徐冬
刘洋
杨荃
王晓晨
孙友昭
王化彬
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
University of Science and Technology Beijing USTB
Original Assignee
University of Science and Technology Beijing USTB
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by University of Science and Technology Beijing USTB filed Critical University of Science and Technology Beijing USTB
Priority to CN202010149682.8A priority Critical patent/CN111318579B/en
Publication of CN111318579A publication Critical patent/CN111318579A/en
Application granted granted Critical
Publication of CN111318579B publication Critical patent/CN111318579B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Images

Classifications

    • B—PERFORMING OPERATIONS; TRANSPORTING
    • B21—MECHANICAL METAL-WORKING WITHOUT ESSENTIALLY REMOVING MATERIAL; PUNCHING METAL
    • B21B—ROLLING OF METAL
    • B21B37/00—Control devices or methods specially adapted for metal-rolling mills or the work produced thereby
    • B21B37/68—Camber or steering control for strip, sheets or plates, e.g. preventing meandering

Abstract

The invention provides a rough rolling camber control method based on data driving, which can effectively reduce operation intervention and realize effective control of camber quality at the head and tail of an intermediate billet. The method comprises the following steps: establishing a data-driven rough rolling intermediate billet head and tail camber prediction model; collecting process data related to rough rolling and camber; forming a plurality of groups of input data by taking the set precision as a step length of the leveling value of the pass operation work within the allowable range of the leveling value; predicting a head camber value YHc and a tail camber value YTc according to the formed input data and the established prediction model; determining a target function value corresponding to each pass operation work leveling value according to the obtained YHc and YTc; and taking the sum of the leveling value of the current pass of the operation work and the intervention value thereof when the target function takes the minimum value as the comprehensive leveling value of the current pass of the operation work, issuing the comprehensive leveling value to the rough rolling basic automatic control system, and executing the inclination adjustment of the roller. The invention relates to the technical field of plate strip rolling.

Description

Roughing camber control method based on data driving
Technical Field
The invention relates to the technical field of plate strip rolling, in particular to a rough rolling camber control method based on data driving.
Background
In recent years, with the development of ferrous metallurgy technology in China, the requirements for high precision, high added value and high technology are higher and higher, and the strip shape problem is one of the main problems which puzzle the production efficiency and the product quality of high-precision strip steel in China. Due to asymmetric influence factors, the intermediate billet is easy to have the defects of camber and the like in the production process of hot rolling and rough rolling, so that accidents such as roller tightening, rolling jamming, belt breakage, tearing, steel piling and the like occur in subsequent rolling. In the rough rolling process, the intermediate billet needs to be accurately controlled in real time, otherwise the camber defect is easily generated, and the accurate control of the intermediate billet is the basis for ensuring the quality of the strip steel and the smooth downstream rolling.
Currently, the study of camber mechanism-based control has been advanced, but the field complexity adds many difficulties to the mathematical modeling of precise control of the intermediate blank camber. A large number of nonlinear influence factors exist in the process of rough rolling of the intermediate blank, the existing camber mechanism control model is difficult to accurately describe the nonlinear characteristics, and the traditional and conventional camber control methods are difficult to meet the requirements of camber control in the hot continuous rolling field.
Disclosure of Invention
In view of the above, the invention provides a rough rolling camber control method based on data driving, which can train a camber prediction model by using offline data, calculate camber values of a head part and a tail part corresponding to different leveling values by an exhaustion method, finally form a comprehensive leveling value by searching a leveling value corresponding to a minimum value of an objective function and combining with an intervention value of the leveling value, and send the comprehensive leveling value to an automatic control system based on rough rolling to execute roll inclination adjustment, thereby realizing a control target based on data driving and considering the head and tail quality of an intermediate billet.
In order to achieve the technical purpose, an embodiment of the present invention provides a rough rolling camber control method based on data driving, including:
step 1-1: according to process data related to the camber of the intermediate blank in each rough rolling pass in a historical database, a data-driven rough rolling intermediate blank head camber and tail camber prediction model is established;
step 1-2: collecting process data which are finished by the rough rolling and related to the camber, preset and calculated in the current pass;
step 1-3: setting the leveling value delta S of the pass operation within the allowable range of the leveling value by taking the set precision as a step length, and combining the collected process data in the step 1-2 to form a plurality of groups of input data XOc;
step 1-4: predicting a head camber value YHc and a tail camber value YTc according to the composed input data and the established rough rolling intermediate billet head and tail camber prediction models;
step 1-5: calculating an objective function value corresponding to each pass operation work flattening value delta S according to the obtained predicted head camber value YHc and tail camber value YTc, wherein the objective function is as follows:
F(ΔS)＝|λYHc(ΔS)|+|(1+λ)YTc(ΔS)|
wherein λ represents a coefficient; YHc (Δ S) represents a value of YHc when the current pass operation leveling value is Δ S; YTc (Δ S) represents a value of YTc when the current pass operation leveling value is Δ S; step 1-6: calculating the minimum value of F (delta S), and recording the operation work leveling value delta S of the current pass corresponding to the minimum value;
step 1-7: obtaining an intervention value delta S of the manual leveling value delta S of the current pass0；
Step 1-8: the comprehensive leveling value ((Delta S + Delta S) of the pass operator0) To the rough rolling base automation control system to execute roll inclination adjustment.
After the step 1-1 of the roughing mill is executed, only the step 1-2-the step 1-7 need to be executed for the camber control of each pass of each intermediate billet;
further, the prediction model in step 1-1 is built by the following steps:
step 2-1: acquiring process data related to camber of the intermediate blank in each rough rolling pass in a historical database, wherein the process data are used as training samples and comprise N input sample data XO, N head bending output sample data YHO and N tail bending output sample data YTO;
the input sample data includes the variables: the method comprises the following steps of (1) adjusting a previous pass operation work leveling value, an operation work leveling value of the current pass, an average rolling force of an operation side of the previous pass, an average rolling force of a transmission side of the previous pass, the longitudinal rigidity of a rolling mill at the operation side, the longitudinal rigidity of a rolling mill at the transmission side, the width of an intermediate blank, the plastic deformation coefficient of a plate blank of the current pass, the outlet thickness of an intermediate blank of the previous pass, the outlet thickness of the intermediate blank of the current pass and the head bending amount of an outlet of;
the head curvature output sample data variables are: the bending amount of the outlet head of the channel is measured;
the tail-bent output sample data variables are: the bending amount of the tail part of the outlet of the current pass;
step 2-2: performing dimensionless standardization processing on the acquired sample data to form N input sample data X, N head bending output sample data YH and N tail bending output sample data YT;
wherein, for the variable with positive and negative value distinction, the standardization target interval is [ -1,1], and the standardization processing formula is:
Figure BDA0002401984370000021
in the formula, x*For the normalized variable data, x is the raw variable data, xmaxIs the maximum value of the variable data in the current rough rolling process, xminIs the minimum value of variable data in the current rough rolling process;
wherein, for the variable without positive and negative value distinction, the standardization target interval is [0,2], and the standardization processing formula is:
Figure BDA0002401984370000031
in the formula, x*For the normalized variable data, x is the raw variable data, xmaxIs the maximum value of the variable data in the current rough rolling process, xminIs the minimum value of variable data in the current rough rolling process;
step 2-3: at M inputs of input sample dataOf the variables, m variables are selected, and n is established for each variable selectediThe center of each variable is formed by arranging and combining
Figure BDA0002401984370000032
A center of aggregation
Figure BDA0002401984370000033
Wherein k is 1,2, 1., c, i is 1,2, 1., m,
Figure BDA0002401984370000034
is a VkThe ith element in (1);
selecting m variables, establishing n for each variable selectediThe centers of the variables are:
Figure BDA0002401984370000035
wherein j is 1,2i，i＝1,2,...,m；
Step 2-4: calculating the center V of each set of each data pair in the input sample data XkDegree of membership of
Figure BDA0002401984370000036
Wherein the p-th sample data X is calculatedpThe ith selection variable of
Figure BDA0002401984370000037
For the corresponding k-th set center VkThe ith variable center of
Figure BDA0002401984370000038
The method of membership of (2) is as follows:
Figure BDA0002401984370000039
wherein k is 1,2, 1, c, i is 1,2, 1, m, p is 1,2, …, N, σiIs a radius, and the value range is
Figure BDA00024019843700000310
niThe number of variable centers;
according to the formula
Figure BDA00024019843700000311
Determining the p-th input sample data Xp to each set center VkDegree of membership of
Figure BDA00024019843700000312
Step 2-5: outputting sample data YH of head bending according to the obtained membership
Figure BDA0002401984370000041
Calculate head coefficient βH；
According to the determined membership degree
Figure BDA0002401984370000042
Calculating coefficient for the p-th input sample data Xp
Figure BDA0002401984370000043
Figure BDA0002401984370000044
Wherein, k is 1,2,., c, p is 1,2, …, N is the number of input sample data;
according to obtaining
Figure BDA0002401984370000045
Establishing an equation:
Figure BDA0002401984370000046
wherein the content of the first and second substances,
Figure BDA0002401984370000047
is the p-th input sample data Xp, YHpIs head bending outputPth data of sample data YH, βH＝[β10,β11,…,β1M,…,βc0,βc1,…,βcM]TAnd substituting the N pairs of input and output data sets into the established equation to obtain a matrix equation:
YH＝XβH
in the formula, βHIs a (M +1) c × 1 dimensional parameter matrix, YH and X are respectively an N × 1 dimensional matrix, an N × (M +1) c dimensional matrix and a head coefficient βHThe least squares estimate of (d) is:
βH＝(XTX)-1XTYH
wherein, superscript T represents matrix transposition;
step 2-6, replacing the head bending amount of the last pass outlet in the step 2-1 with the tail bending amount of the last pass outlet, replacing the head bending output sample data YH in the step 2-5 as output with the tail bending output sample data YT as output, and calculating a tail coefficient β according to the operations of the steps 2-2 to 2-5T。
Further, the process data in step 1-2 includes the following variables: the method comprises the following steps of adjusting the leveling value of the previous pass operation, the average rolling force of the previous pass operation side, the average rolling force of the previous pass transmission side, the longitudinal rigidity of an operation side rolling mill, the longitudinal rigidity of a transmission side rolling mill, the width of an intermediate blank, the plastic deformation coefficient of the plate blank of the current pass, the outlet thickness of the intermediate blank of the previous pass, the outlet thickness of the intermediate blank of the current pass and the head bending amount of the outlet of the previous pass.
Further, the step 1-4 of predicting a head camber value YHc and a tail camber value YTc according to the composed input data and the established rough rolling intermediate billet head and tail camber prediction models, and the method comprises the following steps:
step 4-1: performing dimensionless normalization on the input data XOc to form input data Xc;
step 4-2: calculating each data pair of each set center V in input data XckDegree of membership ω ofkWherein, k is 1, 2.., c;
step 4-3: calculating coefficients
Figure BDA0002401984370000051
Wherein k is 1, 2.., c;
steps 4-4 according to βHPredicting head camber value YHc:
Figure BDA0002401984370000052
wherein k is 1, 2.., c, [ x ]1,x2,...,xM]Is one piece of input data Xc;
step 4-5, replacing the head bending amount of the upper-pass outlet in the step 1-2 with the tail bending amount of the upper-pass outlet according to βTThe tail camber value is predicted YTc according to steps 4-1 to 4-5.
Compared with the prior art, the invention has the beneficial effects that:
1. the labor intensity is reduced: an optimal leveling value is given in a data driving mode, the generation of sickle curves is reduced, the intervention of operators is reduced, and the labor work is greatly reduced;
2. a control target: the invention can give consideration to the control target of the head and tail quality of the intermediate billet;
3. the control effect is as follows: based on a data driving method, a prediction and control method of the camber of the rough rolling intermediate billet is established, data expression of the nonlinear relation of the process conditions to the camber is formed, and the camber generation process can be more accurately described.
Drawings
Fig. 1 is a schematic flow chart of a rough rolling camber control method based on data driving according to an embodiment of the present invention.
Detailed Description
In order to make the technical problems, technical solutions and advantages of the present invention more apparent, the following detailed description is given with reference to the accompanying drawings and specific embodiments.
As shown in fig. 1, a rough rolling camber control method based on data driving according to an embodiment of the present invention includes:
step 1-1: according to process data related to the camber of the intermediate blank in each rough rolling pass in a historical database, a data-driven rough rolling intermediate blank head camber and tail camber prediction model is established;
step 1-2: collecting process data which are finished by the rough rolling and related to the camber, preset and calculated in the current pass;
step 1-3: setting the leveling value delta S of the pass operation within the allowable range of the leveling value by taking the set precision as a step length, and combining the collected process data in the step 1-2 to form a plurality of groups of input data XOc;
step 1-4: predicting a head camber value YHc and a tail camber value YTc according to the composed input data and the established rough rolling intermediate billet head and tail camber prediction models;
step 1-5: calculating an objective function value corresponding to each pass operation work flattening value delta S according to the obtained predicted head camber value YHc and tail camber value YTc, wherein the objective function is as follows:
F(ΔS)＝|λYHc(ΔS)|+|(1+λ)YTc(ΔS)|
wherein λ represents a coefficient; YHc (Δ S) represents a value of YHc when the current pass operation leveling value is Δ S; YTc (Δ S) represents a value of YTc when the current pass operation leveling value is Δ S; in the embodiment, the value range of lambda is 0.1-0.9; since the head camber quality is more important, preferably λ is 0.7;
step 1-6: calculating the minimum value of F (delta S), and recording the operation work leveling value delta S of the current pass corresponding to the minimum value;
step 1-7: obtaining an intervention value delta S of the manual leveling value delta S of the current pass0；
In this embodiment, the operator performs manual intervention according to the calculated current pass adjustment value Δ S, and records the intervention value Δ S of the current pass adjustment value0；
Step 1-8: the comprehensive leveling value ((Delta S + Delta S) of the pass operator0) To the rough rolling base automation control system to execute roll inclination adjustment.
After the step 1-1 of the roughing mill is executed, only the step 1-2-the step 1-7 need to be executed for the camber control of each pass of each intermediate billet;
in this embodiment, the prediction model in step 1-1 is established by the following steps:
step 2-1: acquiring process data related to camber of the intermediate blank in each rough rolling pass in a historical database, wherein the process data are used as training samples and comprise N input sample data XO, N head bending output sample data YHO and N tail bending output sample data YTO;
in this embodiment, N is 20000, and the process data of the near term 20000 sets of intermediate blanks is taken as sample data;
the input sample data includes the variables: the method comprises the following steps of (1) adjusting a previous pass operation work leveling value, an operation work leveling value of the current pass, an average rolling force of an operation side of the previous pass, an average rolling force of a transmission side of the previous pass, the longitudinal rigidity of a rolling mill at the operation side, the longitudinal rigidity of a rolling mill at the transmission side, the width of an intermediate blank, the plastic deformation coefficient of a plate blank of the current pass, the outlet thickness of an intermediate blank of the previous pass, the outlet thickness of the intermediate blank of the current pass and the head bending amount of an outlet of;
the head curvature output sample data variables are: the bending amount of the outlet head of the channel is measured;
the tail-bent output sample data variables are: the bending amount of the tail part of the outlet of the current pass;
step 2-2: performing dimensionless standardization processing on the acquired sample data to form N input sample data X, N head bending output sample data YH and N tail bending output sample data YT;
wherein, for the variables (such as operation leveling value and bending amount) with positive and negative values, the standardized target interval is [ -1,1]The normalization processing formula is:
Figure BDA0002401984370000071
in the formula, x*For the normalized variable data, x is the raw variable data, xmaxIs the maximum value of the variable data in the current rough rolling process, xminIs the minimum value of variable data in the current rough rolling process;
in this embodiment, the variables that are differentiated by positive and negative values include: the method comprises the following steps of (1) adjusting a previous pass operation work leveling value, an operation work leveling value of the current pass, a head bending amount of an outlet of the previous pass, a tail bending amount of an outlet of the previous pass, a head bending amount of an outlet of the current pass and a tail bending amount of an outlet of the current pass;
for variables without positive and negative value distinction, the standardized target interval is [0,2], and the standardized processing formula is as follows:
Figure BDA0002401984370000072
in the formula, x*For the normalized variable data, x is the raw variable data, xmaxIs the maximum value of the variable data in the current rough rolling process, xminIs the minimum value of variable data in the current rough rolling process;
in this embodiment, the variables without positive and negative value distinction include: the average rolling force of the previous operation side, the average rolling force of the previous transmission side, the longitudinal rigidity of an operation side rolling mill, the longitudinal rigidity of a transmission side rolling mill, the width of an intermediate billet, the plastic deformation coefficient of the plate blank of the current pass, the outlet thickness of the intermediate billet of the previous pass and the outlet thickness of the intermediate billet of the current pass;
step 2-3: selecting M variables from M input variables of input sample data, and establishing n for each selected variableiThe center of each variable is formed by arranging and combining
Figure BDA0002401984370000073
A center of aggregation
Figure BDA0002401984370000074
Wherein k is 1,2, 1., c, i is 1,2, 1., m,
Figure BDA0002401984370000075
is a VkThe ith element in (1);
selecting m variables, establishing n for each variable selectediThe centers of the variables are:
Figure BDA0002401984370000076
wherein j is 1,2i，i＝1,2,...,m；
In this embodiment, M is 11, M is 2, and the variables selected are the width of the intermediate billet, the outlet thickness of the intermediate billet of the previous pass, and niWhen c is 9, the geometric center is formed as 3, and the obtained VkThe constituent V is represented as:
Figure BDA0002401984370000081
step 2-4: calculating the center V of each set of each data pair in the input sample data XkDegree of membership of
Figure BDA0002401984370000082
Wherein the p-th sample data X is calculatedpThe ith selection variable of
Figure BDA0002401984370000083
For the corresponding k-th set center VkThe ith variable center of
Figure BDA0002401984370000084
The method of membership of (2) is as follows:
Figure BDA0002401984370000085
wherein k is 1,2, 1, c, i is 1,2, 1, m, p is 1,2, …, N, σiIs a radius, and the value range is
Figure BDA0002401984370000086
niThe number of variable centers;
in the present embodiment, σi＝0.35，i＝1,2,.….,m；
According to the formula
Figure BDA0002401984370000087
Determining the p-th input sample data Xp to each set center VkDegree of membership of
Figure BDA0002401984370000088
Step 2-5: outputting sample data YH of head bending according to the obtained membership
Figure BDA0002401984370000089
Calculate head coefficient βH；
According to the determined membership degree
Figure BDA00024019843700000810
Calculating coefficient for the p-th input sample data Xp
Figure BDA00024019843700000811
Figure BDA00024019843700000812
Wherein, k is 1,2,. …, c, p is 1,2, …, N is the number of input sample data;
according to obtaining
Figure BDA0002401984370000091
Establishing an equation:
Figure BDA0002401984370000092
wherein the content of the first and second substances,
Figure BDA0002401984370000093
is the p-th input sample data Xp, YHpPth data of head bending output sample data YH, βH＝[β10,β11,…,β1M,…,βc0,βc1,…,βcM]TAnd substituting the N pairs of input and output data sets into the established equation to obtain a matrix equation:
YH＝XβH
in the formula, βHIs (M +1) c × 1 dimensional parameter matrix, YH, X areIs a matrix of dimension N × 1, N × (M +1) c, and head coefficients βHThe least squares estimate of (d) is:
βH＝(XTX)-1XTYH
wherein, superscript T represents matrix transposition;
step 2-6, replacing the head bending amount of the last pass outlet in the step 2-1 with the tail bending amount of the last pass outlet, replacing the head bending output sample data YH in the step 2-5 as output with the tail bending output sample data YT as output, and calculating a tail coefficient β according to the operations of the steps 2-2 to 2-5T。
The process data in step 1-2 includes the following variables: the method comprises the following steps of adjusting the leveling value of the previous pass operation, the average rolling force of the previous pass operation side, the average rolling force of the previous pass transmission side, the longitudinal rigidity of an operation side rolling mill, the longitudinal rigidity of a transmission side rolling mill, the width of an intermediate blank, the plastic deformation coefficient of the plate blank of the current pass, the outlet thickness of the intermediate blank of the previous pass, the outlet thickness of the intermediate blank of the current pass and the head bending amount of the outlet of the previous pass.
In the steps 1-4, head camber value YHc and tail camber value YTc are predicted according to the composed input data and the established rough rolling intermediate billet head and tail camber prediction models, and the method comprises the following steps:
step 4-1: performing dimensionless normalization on the input data XOc to form input data Xc, the normalization being performed as described in step 2-2;
step 4-2: calculating each data pair of each set center V in input data XckDegree of membership ω ofkWherein k is 1,2,. …, c, calculated as described in steps 2-4;
step 4-3: calculating coefficients
Figure BDA0002401984370000101
Wherein k is 1,2,. …, c;
steps 4-4 according to βHPredicting head camber value YHc:
Figure BDA0002401984370000102
wherein k is 1,2,. …, c, [ x [ ]1,x2,.….,xM]Is one piece of input data Xc;
step 4-5, replacing the head bending amount of the upper-pass outlet in the step 1-2 with the tail bending amount of the upper-pass outlet according to βTThe tail camber value is predicted YTc according to steps 4-1 to 4-5.
Compared with the prior art, the invention has the beneficial effects that:
1. the labor intensity is reduced: an optimal leveling value is given in a data driving mode, the generation of sickle curves is reduced, the intervention of operators is reduced, and the labor work is greatly reduced;
2. a control target: the invention can give consideration to the control target of the head and tail quality of the intermediate billet;
3. the control effect is as follows: based on a data driving method, a prediction and control method of the camber of the rough rolling intermediate billet is established, data expression of the nonlinear relation of the process conditions to the camber is formed, and the camber generation process can be more accurately described.
While the foregoing is directed to the preferred embodiment of the present invention, it will be understood by those skilled in the art that various changes and modifications may be made without departing from the spirit and scope of the invention as defined in the appended claims.

Claims (10)

1. A rough rolling camber control method based on data driving is characterized by comprising the following steps:
step 1-1: according to process data related to the camber of the intermediate blank in each rough rolling pass in a historical database, a data-driven rough rolling intermediate blank head camber and tail camber prediction model is established;
step 1-2: collecting process data which are finished by the rough rolling and related to the camber, preset and calculated in the current pass;
step 1-3: setting the leveling value delta S of the pass operation within the allowable range of the leveling value by taking the set precision as a step length, and combining the collected process data in the step 1-2 to form a plurality of groups of input data XOc;
step 1-4: predicting a head camber value YHc and a tail camber value YTc according to the composed input data and the established rough rolling intermediate billet head and tail camber prediction models;
step 1-5: calculating an objective function value corresponding to each pass operation work flattening value delta S according to the obtained predicted head camber value YHc and tail camber value YTc, wherein the objective function is as follows:
F(ΔS)＝|λYHc(ΔS)|+|(1+λ)YTc(ΔS)|
wherein λ represents a coefficient; YHc (Δ S) represents a value of YHc when the current pass operation leveling value is Δ S; YTc (Δ S) represents a value of YTc when the current pass operation leveling value is Δ S;
step 1-6: calculating the minimum value of F (delta S), and recording the operation work leveling value delta S of the current pass corresponding to the minimum value;
step 1-7: obtaining an intervention value delta S of the manual leveling value delta S of the current pass0；
Step 1-8: the comprehensive leveling value ((Delta S + Delta S) of the pass operator0) To the rough rolling base automation control system to execute roll inclination adjustment.
2. The rough rolling camber control method based on data driving of claim 1, wherein the prediction model in the step 1-1 is established by the following steps:
step 2-1: acquiring process data related to camber of the intermediate blank in each rough rolling pass in a historical database, wherein the process data are used as training samples and comprise N input sample data XO, N head bending output sample data YHO and N tail bending output sample data YTO;
step 2-2: performing dimensionless standardization processing on the acquired sample data to form N input sample data X, N head bending output sample data YH and N tail bending output sample data YT;
step 2-3: selecting M variables from M input variables of input sample data, and establishing n for each selected variableiThe center of each variable is formed by arranging and combining
Figure FDA0002401984360000011
A center of aggregation
Figure FDA0002401984360000012
Wherein k is 1,2, 1., c, i is 1,2, 1., m,
Figure FDA0002401984360000021
is a VkThe ith element in (1);
step 2-4: calculating the center V of each set of each data pair in the input sample data XkDegree of membership of
Figure FDA0002401984360000022
Step 2-5: outputting sample data YH of head bending according to the obtained membership
Figure FDA0002401984360000023
Calculate head coefficient βH；
Step 2-6, replacing the head bending amount of the last pass outlet in the step 2-1 with the tail bending amount of the last pass outlet, replacing the head bending output sample data YH in the step 2-5 as output with the tail bending output sample data YT as output, and calculating a tail coefficient β according to the operations of the steps 2-2 to 2-5T。
3. The roughing camber control method based on data driving of claim 1 wherein the process data in step 1-2 includes the following variables: the method comprises the following steps of adjusting the leveling value of the previous pass operation, the average rolling force of the previous pass operation side, the average rolling force of the previous pass transmission side, the longitudinal rigidity of an operation side rolling mill, the longitudinal rigidity of a transmission side rolling mill, the width of an intermediate blank, the plastic deformation coefficient of the plate blank of the current pass, the outlet thickness of the intermediate blank of the previous pass, the outlet thickness of the intermediate blank of the current pass and the head bending amount of the outlet of the previous pass.
4. The rough rolling camber control method based on data driving according to claim 1, wherein the steps 1-4 are used for predicting a head camber value YHc and a tail camber value YTc according to the composed input data and the established rough rolling intermediate billet head and tail camber prediction models, and the method comprises the following steps:
step 4-1: performing dimensionless normalization on the input data XOc to form input data Xc;
step 4-2: calculating each data pair of each set center V in input data XckDegree of membership ω ofkWherein, k is 1, 2.., c;
step 4-3: calculating coefficients
Figure FDA0002401984360000024
Wherein k is 1, 2.., c;
steps 4-4 according to βHPredicting head camber value YHc:
Figure FDA0002401984360000025
wherein k is 1, 2.., c, [ x ]1,x2,...,xM]Is one piece of input data Xc;
step 4-5, replacing the head bending amount of the upper-pass outlet in the step 1-2 with the tail bending amount of the upper-pass outlet according to βTThe tail camber value is predicted YTc according to steps 4-1 to 4-5.
5. The rough rolling camber control method based on data driving of claim 1, wherein after the step 1-1 is performed for one rough rolling mill, only the step 1-2-the step 1-7 need to be performed for camber control of each pass of each intermediate billet.
6. The rough rolling camber control method based on data driving of claim 2, wherein in the step 2-1, the input sample data comprises variables: the method comprises the following steps of (1) adjusting a previous pass operation work leveling value, an operation work leveling value of the current pass, an average rolling force of an operation side of the previous pass, an average rolling force of a transmission side of the previous pass, the longitudinal rigidity of a rolling mill at the operation side, the longitudinal rigidity of a rolling mill at the transmission side, the width of an intermediate blank, the plastic deformation coefficient of a plate blank of the current pass, the outlet thickness of an intermediate blank of the previous pass, the outlet thickness of the intermediate blank of the current pass and the head bending amount of an outlet of;
the variables of the head curvature output sample data are: the bending amount of the outlet head of the channel is measured;
the variables of the tail-bent output sample data are: the bending amount of the tail of the outlet of the channel is measured.
7. The rough rolling camber control method based on data driving of claim 2, wherein in the step 2-2, the dimensionless standardization process is as follows:
for the variables with positive and negative value distinction, the standardized target interval is [ -1,1], and the standardized processing formula is as follows:
Figure FDA0002401984360000031
in the formula, x*For the normalized variable data, x is the raw variable data, xmaxIs the maximum value of the variable data in the current rough rolling process, xminIs the minimum value of variable data in the current rough rolling process;
for variables without positive and negative value distinction, the standardization target interval is [0,2], and the standardization processing formula is as follows:
Figure FDA0002401984360000032
in the formula, x*For the normalized variable data, x is the raw variable data, xmaxIs the maximum value of the variable data in the current rough rolling process, xminIs the minimum value of the variable data in the current rough rolling process.
8. The roughing camber control method based on data driving of claim 2 characterized in thatIn step 2-3, m variables are selected, and n is established for each variable selectediThe centers of the variables are:
Figure FDA0002401984360000033
wherein j is 1,2i，i＝1,2,...,m。
9. The rough rolling camber control method based on data driving of claim 2, wherein in the steps 2-4, the p-th input sample data Xp is calculated for each set center VkDegree of membership of
Figure FDA0002401984360000034
The method comprises the following steps:
determining the p-th sample data XpThe ith selection variable of
Figure FDA0002401984360000041
For the corresponding k-th set center VkThe ith variable center of
Figure FDA0002401984360000042
The membership degree is as follows:
Figure FDA0002401984360000043
wherein k is 1,2, 1, c, i is 1,2, 1, m, p is 1,2, …, N, σiIs a radius, σiA value range of
Figure FDA0002401984360000044
niThe number of variable centers;
according to the formula
Figure FDA0002401984360000045
Determining the p-th input sample data Xp to each set center VkDegree of membership of
Figure FDA0002401984360000046
10. The rough rolling camber control method based on data driving of claim 2, wherein in the step 2-5, the head coefficient βHThe method of the calculation method comprises the following steps:
according to the determined membership degree
Figure FDA0002401984360000047
Calculating coefficient for the p-th input sample data Xp
Figure FDA0002401984360000048
Figure FDA0002401984360000049
Wherein, k is 1,2,., c, p is 1,2, …, N is the number of input sample data;
according to obtaining
Figure FDA00024019843600000410
Establishing an equation:
Figure FDA00024019843600000411
wherein the content of the first and second substances,
Figure FDA00024019843600000412
is the p-th input sample data Xp, YHpPth data of head bending output sample data YH, βH＝[β10,β11,…,β1M,…,βc0,βc1,…,βcM]TAnd substituting the N pairs of input and output data sets into the established equation to obtain a matrix equation:
YH＝XβH
in the formula, βHIs a (M +1) c × 1 dimensional parameter matrix, YH and X are respectively an N × 1 dimensional matrix, an N × (M +1) c dimensional matrix and a head coefficient βHIs βH＝(XTX)-1XTYH, in which the superscript T denotes matrix transposition.
CN202010149682.8A 2020-03-06 2020-03-06 Roughing camber control method based on data driving Active CN111318579B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202010149682.8A CN111318579B (en) 2020-03-06 2020-03-06 Roughing camber control method based on data driving

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202010149682.8A CN111318579B (en) 2020-03-06 2020-03-06 Roughing camber control method based on data driving

Publications (2)

Publication Number Publication Date
CN111318579A true CN111318579A (en) 2020-06-23
CN111318579B CN111318579B (en) 2020-12-29

Family

ID=71165539

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202010149682.8A Active CN111318579B (en) 2020-03-06 2020-03-06 Roughing camber control method based on data driving

Country Status (1)

Country Link
CN (1) CN111318579B (en)

Cited By (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN112007958A (en) * 2020-09-01 2020-12-01 宝钢湛江钢铁有限公司 Automatic control method for rough rolling camber
CN112958634A (en) * 2021-01-28 2021-06-15 北京科技大学设计研究院有限公司 Pre-leveling method of finish rolling machine frame based on sickle elbow part

Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
JPS61219411A (en) * 1985-03-25 1986-09-29 Kawasaki Steel Corp Method for preventing camber in plate rolling
KR101322120B1 (en) * 2011-08-10 2013-10-28 주식회사 포스코 Method and apparatus for controlling wedge and camber of steel plate
CN105032948A (en) * 2015-08-25 2015-11-11 首钢京唐钢铁联合有限责任公司 Control method for reducing rough rolling camber
CN105234189A (en) * 2015-11-13 2016-01-13 北京首钢自动化信息技术有限公司 Slab sickle bending control system and method used for roughing mill
CN110539024A (en) * 2019-09-02 2019-12-06 安阳钢铁股份有限公司 method for reducing camber of steel plate based on disc shearing edge

Patent Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
JPS61219411A (en) * 1985-03-25 1986-09-29 Kawasaki Steel Corp Method for preventing camber in plate rolling
KR101322120B1 (en) * 2011-08-10 2013-10-28 주식회사 포스코 Method and apparatus for controlling wedge and camber of steel plate
CN105032948A (en) * 2015-08-25 2015-11-11 首钢京唐钢铁联合有限责任公司 Control method for reducing rough rolling camber
CN105234189A (en) * 2015-11-13 2016-01-13 北京首钢自动化信息技术有限公司 Slab sickle bending control system and method used for roughing mill
CN110539024A (en) * 2019-09-02 2019-12-06 安阳钢铁股份有限公司 method for reducing camber of steel plate based on disc shearing edge

Cited By (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN112007958A (en) * 2020-09-01 2020-12-01 宝钢湛江钢铁有限公司 Automatic control method for rough rolling camber
CN112007958B (en) * 2020-09-01 2022-03-18 宝钢湛江钢铁有限公司 Automatic control method for rough rolling camber
CN112958634A (en) * 2021-01-28 2021-06-15 北京科技大学设计研究院有限公司 Pre-leveling method of finish rolling machine frame based on sickle elbow part

Also Published As

Publication number Publication date
CN111318579B (en) 2020-12-29

Similar Documents

Publication Publication Date Title
CN111318579B (en) Roughing camber control method based on data driving
CN105290117B (en) The classification regulation and control method of the ultra-thin cold-strip steel high order flatness defect of big flakiness ratio
CN114329940A (en) Continuous casting billet quality prediction method based on extreme learning machine
CN101376139B (en) Control method for producing conical plate blank using side compression machine of fixed width plate blank
CN108480405B (en) Cold-rolled plate shape regulation and control efficiency coefficient obtaining method based on data driving
CN108655186B (en) Roll-force presetting method based on artificial neural network and mathematical model
CN111250548B (en) Board convexity prediction method based on kernel partial least square combined support vector machine
CN103357669A (en) Plate model prediction control method
CN103506404A (en) Method for controlling roll gap during finish rolling of strip steel
CN106862284B (en) A kind of cold rolled sheet signal mode knowledge method for distinguishing
CN112037209A (en) Steel plate roller wear loss prediction method and system
CN115169453A (en) Hot continuous rolling width prediction method based on density clustering and depth residual error network
CN115608793B (en) Finish rolling temperature regulation and control method for mechanism fusion data
DE102020104286A1 (en) Method and device for process optimization of a production plant
CN115601313A (en) Visual monitoring management system for tempered glass production process
CN117131767A (en) Method for predicting thermal convexity of working roll of hot rolling four-high mill based on random forest algorithm
Ma et al. Effect of strip profile of hot-rolled silicon steel on transverse thickness difference of cold-rolled strip
CN115106384B (en) Thick plate rolling roll gap correction method based on random forest
CN112246880B (en) Twenty-high rolling mill strip shape optimization control method based on feedforward-middle shifting compensation
CN112329198B (en) Wide-thick plate length optimization method based on data driving
CN108067506B (en) Medium and Heavy Plate Rolling passage dynamic becomes setting control method
CN107520255A (en) A kind of self-learning type inlet of rolling mill thickness optimization method
CN113210936A (en) Welding method and system for straight welded pipe and readable medium
CN117531845B (en) Method and device for controlling plane shape of medium plate, storage medium and computer equipment
CN100498805C (en) Roller abrasion mathematical model optimizing method for improving CSP product quality

Legal Events

Date Code Title Description
PB01 Publication
PB01 Publication
SE01 Entry into force of request for substantive examination
SE01 Entry into force of request for substantive examination
GR01 Patent grant
GR01 Patent grant