A strip shape control method based on the DS evidence theory

Abstract

The present invention discloses a strip shape control method based on the DS evidence theory, which includes: analyzing relevant characteristic parameters and collecting rolling process data; analyzing the strip deformation mechanism and constructing a rolling mechanism model; calculating rolling mechanism data based on the rolling mechanism model, and constructing a machine learning coordination model of mechanism fusion data based on the rolling mechanism data and the rolling process data; establishing a strip exit shape prediction model of the mechanism fusion data based on the DS evidence theory; establishing a comprehensive shape control strategy based on the warping limit and the waviness discrimination model. The present invention comprehensively considers the influence of rolling mechanism data and process data on the strip quality, increases the connection between strongly correlated features and the model based on the DS evidence theory, and uses a comprehensive shape control strategy to perform feedback control on the strip shape, realizing the dynamic adjustment of the rolling working conditions to ensure the strip exit shape quality.

Description

Technical Field

[0001] The present invention belongs to the technical field of rolling control, and particularly relates to a strip shape control method based on DS evidence theory. Background Technique

[0002] With the continuous improvement of users' requirements for high-precision strip products, more machine learning and intelligent algorithms are applied to the strip rolling field to solve complex rolling conditions. Strip shape and thickness are important criteria for measuring whether the quality and dimensional accuracy of strip products are qualified. From the perspective of the overall product quality, the control accuracy of both is particularly important. For strip steel, analyzing the rolling mechanism and integrating and optimizing multiple stands are both difficult points in the process. Any fluctuation in the strip process parameters of a single stand will be transmitted to the upstream and downstream of all stands through the temperature, thickness, and cross-sectional shape between the stands. How to improve the control accuracy of strip thickness and shape remains a difficult problem in industrial applications.

[0003] DS evidence theory belongs to the most effective algorithm in collaborative multi-source information. This theory broadens the basic event space to the power set, and the provided combination rules can achieve fusion without prior information, and can more effectively process uncertain information. Applying DS evidence theory to the construction of a machine learning (ML) model, more efficient features are constructed, and intelligent algorithms are added to optimize parameters to improve the model accuracy. However, the model established based on data does not depend on the system mechanism. When the environmental conditions corresponding to the data set change, the model will no longer be able to adapt to the environment and needs to be reconstructed. Therefore, conducting in-depth theoretical exploration and excavation of the rolling mechanism and weakening the dependence of the model on data can further improve the applicability of the model. Summary of the Invention

[0004] The purpose of the present invention is to provide a strip shape control method based on DS evidence theory to solve the problems existing in the above-mentioned prior art.

[0005] To achieve the above purpose, the present invention provides a strip shape control method based on DS evidence theory, including the following steps:

[0006] Analyze relevant characteristic parameters and collect rolling process data;

[0007] Analyze the strip deformation mechanism and construct a rolling mechanism model;

[0008] Calculate rolling mechanism data based on the rolling mechanism model, and construct a machine learning coordination model of mechanism fusion data based on the rolling mechanism data and the rolling process data;

[0009] Establish a strip exit shape prediction model of mechanism fusion data based on DS evidence theory;

[0010] Establish a comprehensive shape control strategy based on the warping limit and waviness discrimination model, obtain rolling process parameters, perform feedback control on the strip shape based on the rolling process parameters, and dynamically adjust the rolling working conditions.

[0011] Optionally, the rolling process data includes rolling process data and exit strip index data;

[0012] The rolling process data includes rolling force, bending roll force, roll gap value, roll shifting amount, rolling speed, inter-stand tension, rolling inlet temperature, and exit temperature;

[0013] The exit strip index data includes the exit width, exit thickness, exit camber, and exit flatness of the strip.

[0014] Optionally, the process of analyzing the strip deformation mechanism and constructing a rolling mechanism model based on the rolling process data includes:

[0015] Based on the shape control mechanism, construct a strip critical warping model;

[0016] Based on the influence of inter-stand tension on warping behavior, construct a warping correction model under tension;

[0017] Based on the influence of the transverse metal flow in the strip on the residual stress, construct a strip metal transverse flow model;

[0018] Based on the influence of inter-stand tension and strip internal stress, construct an inter-stand secondary deformation model.

[0019] Optionally, the construction method of the strip critical warping model is:

[0020]

[0021] where, σ cr is the critical stress for the strip to warp, k cr is the critical stress coefficient, E is the strip elastic modulus, v is the Poisson's ratio, h' is the strip thickness, and B' is the strip width;

[0022] The construction method of the warping correction model is:

[0023]

[0024] where, σ t is the inter-stand tension, k1 is the middle wave correction coefficient, k2 is the edge wave correction coefficient, C h is the pre-rolling proportional camber, C H is the post-rolling proportional camber;

[0025] The construction method of the strip metal transverse flow model is:

[0026]

[0027] Among them, S e is the transverse flow coefficient of the strip steel towards the edge, and S m is the transverse flow coefficient of the strip steel towards the center. Q is the flow coefficient related to the width and thickness of the strip steel, and K se is the flow coefficient of the strip steel edge, and K sm is the flow coefficient of the strip steel center;

[0028] The construction method of the secondary deformation model between stands is as follows:

[0029]

[0030] Among them, k c is the deformation coefficient, θ is the deformation resistance, t is the time parameter, and K T is the temperature influence coefficient, d is the deformation index coefficient, h’(x) is the transverse thickness distribution of the strip steel, and k h / B is the thickness-width ratio influence coefficient, and a is the thickness-width ratio index coefficient.

[0031] Optionally, the process of calculating the rolling mechanism data based on the rolling mechanism model and constructing the machine learning coordination model of the mechanism fusion data based on the rolling mechanism data and the rolling process data includes:

[0032] Calculating the corresponding rolling mechanism data based on the strip steel critical warping model, the metal transverse flow model, and the secondary deformation model between stands; the rolling mechanism data includes the strip steel critical warping ultimate stress value, the transverse flow coefficient values of the strip steel towards the edge and the center respectively, the deformation resistance value of the strip steel during secondary deformation between stands, and the strip steel temperature value;

[0033] Taking the rolling mechanism data and the rolling process data as input features, constructing a machine learning coordination model of the mechanism fusion data.

[0034] Optionally, the process of establishing a strip steel exit shape prediction model of the mechanism fusion data based on the DS evidence theory includes:

[0035] Defining the sample space;

[0036] Grouping the rolling mechanism data and the rolling process data to obtain grouped information sources;

[0037] Calculating the grouped information sources as the training input values of the machine learning coordination model to obtain a classifier, and obtaining an evidence data group based on the classifier and the sample space;

[0038] Synthesizing the evidence data group based on the DS evidence theory and defining the synthesis rule.

[0039] Optionally, the method for defining the sample space is as follows:

[0040] G = {IDS, POS, APS, MPS, TAS}

[0041] Wherein, G is the sample space, IDS is the ideal shape, POS is the potential shape, APS is the apparent shape, MPS is the mixed shape, and TAS is the shape affected by tension.

[0042] Optionally, the process of calculating the grouping information source as the training input value of the machine learning coordination model includes:

[0043] Construct an M-SVR model and perform calculations based on the M-SVR model;

[0044] Optimize the M-SVR model based on NSGA-III;

[0045] Calculate the geometric relationship for shape determination, and the M-SVR model obtains a classifier based on the geometric relationship for shape determination.

[0046] Optionally, the method for synthesizing the evidence data group based on the DS evidence theory is as follows:

[0047]

[0048] The synthesis rule includes:

[0049] If there exists:

[0050] Satisfying And ① m(S1) - m(S2) > ε1 (ε1 > 0); ② m(Θ) < ε2 (ε2 > 0); ③ m(S1) > m(Θ), then S1 is the prediction result, and the values of ε1 and ε2 are adjusted to the optimal based on multiple experiments; where Θ is the frame of discernment, representing the set of all possible propositions.

[0051] Optionally, the process of establishing a comprehensive shape control strategy based on the warping limit and the waviness discrimination model includes:

[0052] Take the output that satisfies the synthesis rule as the input of the comprehensive shape control strategy, and calculate the natural outlet shape of the strip based on the influence of the secondary deformation between stands on the strip;

[0053] Calculate the target shape based on the following method:

[0054]

[0055] Wherein, G TC is the target shape, f i *is the target flatness of the i-th stand, K f is the flatness correction coefficient, is the target cross-sectional shape, is the thickness reduction of the strip along the width direction, is the target crown of the i-th stand, is the total control amount of the strip target crown, is the sum of the warping limits of the rolling mill unit, ΔC i is the crown control amount of the i-th stand, K ci is the crown correction coefficient;

[0056] Calculate the required target crown adjustment amount for each stand based on the natural exit strip shape and the target strip shape;

[0057] Check the limit values of the rolling mill equipment and the strip warping limit range, and control the target crown adjustment amount based on the limit values and the strip warping limit range;

[0058] Based on the real-time strip shape detected by the full-automatic X-ray diffraction method residual stress detector, calculate the deviation between the measured value and the target value;

[0059] Determine the corresponding rolling process parameters based on the deviation between the measured value and the target value.

[0060] The technical effect of the present invention is:

[0061] The present invention comprehensively considers the influence of rolling mechanism data and process data on the strip quality, increases the connection between strongly correlated features and the model through the DS evidence theory, uses a residual stress detector to detect the strip shape in real time, and adopts a comprehensive strip shape control strategy for strip shape feedback control to achieve dynamic adjustment of the rolling working conditions to ensure the strip exit strip shape quality. Description of the Drawings

[0062] The drawings constituting a part of this application are used to provide a further understanding of this application. The schematic embodiments of this application and their descriptions are used to explain this application and do not constitute an improper limitation of this application. In the drawings:

[0063] Figure 1 is the schematic diagram of the overall experimental process in the embodiment of the present invention;

[0064] Figure 2 is the schematic diagram of the machine learning coordination model of the rolling mechanism fusion data in the embodiment of the present invention;

[0065] Figure 3 is the working flow chart of the NSGA-III optimized M-SVR in the embodiment of the present invention;

[0066] Figure 4 is the comprehensive strip shape control strategy diagram in the embodiment of the present invention. Detailed implementation manners

[0067] It should be noted that, without conflict, the embodiments in the present application and the features in the embodiments can be combined with each other. The present application will be described in detail below with reference to the drawings and in combination with the embodiments.

[0068] It should be noted that the steps shown in the flowchart of the drawings can be executed in a computer system such as a set of computer-executable instructions, and although the logical order is shown in the flowchart, in some cases, the steps shown or described can be executed in a different order than here.

[0069] Embodiment 1

[0070] As Figures 1-4 shown, a strip shape control method based on the DS evidence theory is provided in this embodiment, including the following steps:

[0071] Step 1: Through the influence of the real-time changes of each state parameter in the characteristic analysis process on the strip shape-strip thickness, collect relevant rolling process data on site, including rolling force, bending roll force, roll gap value, roll shifting amount, rolling speed, inter-stand tension, rolling inlet temperature and outlet temperature, the outlet width, outlet thickness, outlet crown and outlet flatness of the strip, which are used as part of the input characteristic parameters of the model of the present invention.

[0072] Step 2: Study the strip deformation mechanism and establish a rolling mechanism model, including the following steps:

[0073] Step 2.1: According to the strip shape control mechanism, that is, the mechanical condition that the strip does not warp while the cross-sectional shape of the strip changes, establish a strip critical warping model:

[0074]

[0075] Among them, σ cr is the critical stress for the strip to warp, k cr is the critical stress coefficient, E is the strip elastic modulus, v is the Poisson's ratio, h' is the strip thickness, and B' is the strip width;

[0076] Considering the influence of the inter-stand tension on the warping behavior, establish a warping correction model under tension:

[0077]

[0078] Among them, σ t is the inter-stand tension, k1 is the middle wave correction coefficient, k2 is the edge wave correction coefficient, C h is the pre-rolling proportional crown, and C H is the post-rolling proportional crown.

[0079] Step 2.2: Establish a cross - flow model of strip metal. The cross - flow of metal in the strip has a certain impact on the residual stress and will preferentially flow in the direction that is conducive to reducing the maximum residual stress:

[0080]

[0081] Among them, S e is the cross - flow coefficient of the strip towards the edge, S m is the cross - flow coefficient of the strip towards the center, Q is the flow coefficient related to the width and thickness of the strip, K se is the flow coefficient at the strip edge, K sm is the flow coefficient at the strip center.

[0082] Step 2.3: Establish a secondary deformation model between stands under the action of the tension between stands and the stress in the strip:

[0083]

[0084] Among them, k c is the deformation coefficient, θ is the deformation resistance, t is the time parameter, K T is the temperature influence coefficient, d is the deformation index coefficient, h’(x) is the cross - thickness distribution of the strip, k h / B is the thickness - width ratio influence coefficient, a is the thickness - width ratio index coefficient, T is the strip temperature;

[0085] The deformation resistance calculation model is:

[0086]

[0087] Among them, p B is the rolling force per unit width, L is the contact arc length with the work roll, H is the thickness of the incoming rolled piece, h is the thickness of the outgoing rolled piece;

[0088] The strip temperature calculation formula is:

[0089]

[0090] The initial condition is the strip temperature field at the end of the previous link:

[0091] T| t＝0 =T0(x,y,z)

[0092] Among them, ρ, c and λ are the density, specific heat capacity and thermal conductivity of the strip respectively, x, y and z are the coordinates of the strip in the length, width and thickness directions in the rectangular coordinate system; Q' is the source term considering the heat generated by strip deformation.

[0093] Step 3: Calculate the rolling mechanism data and establish a machine learning coordination model that fuses the mechanism data with the process data, as shown in Figure 2 below, which includes the following steps:

[0094] Step 3.1: According to the strip critical warping model, metal transverse flow model, and inter-stand secondary deformation model established in Step 2, calculate the corresponding rolling mechanism data; the rolling mechanism data includes the strip critical warping ultimate stress value, the transverse flow coefficient values of the strip towards the edge and the center respectively, the deformation resistance value of the strip undergoing secondary deformation between stands, and the strip temperature value;

[0095] Step 3.2: Use the rolling mechanism data and the process data as input features to establish a machine learning coordination model for the mechanism-fused data.

[0096] Step 4: Establish a strip exit shape prediction model for the mechanism-fused data based on the DS evidence theory, including the following steps:

[0097] Step 4.1: The essence of the strip shape problem is the uneven distribution of strip residual stress. The main reasons for this phenomenon are uneven deformation between stands and uneven strip cooling. Define the sample space G = {IDS, POS, APS, MPS, TAS}, where IDS is the ideal shape, POS is the potential shape, APS is the apparent shape, MPS is the mixed shape, and TAS is the shape affected by tension;

[0098] Step 4.2: Divide the rolling mechanism data and the rolling process data into three groups of information sources based on the rolling environment, rolling process, and strip exit state. Among them, the rolling environment data includes the rolling inlet temperature, outlet temperature, and strip temperature; the rolling process data includes the rolling force, bending roll force, roll gap value, roll shifting amount, rolling speed, inter-stand tension, strip critical warping ultimate stress value, the transverse flow coefficient values of the strip towards the edge and the center respectively, the deformation resistance value of the strip undergoing secondary deformation between stands; the strip exit state data includes the strip exit width value, exit thickness value, exit convexity value, exit flatness value, and the strip thickness value at each moment; Use the above data as the training input values of the machine learning coordination model, that is, the M-SVR model, to obtain classifiers, denoted as M1, M2, M3; The specific calculation content is:

[0099] (1) Collect the data of the above three groups of information sources as the input values of the M-SVR model;

[0100] (2) Establish the M-SVR model

[0101] For the multi-output regression problem with M-dimensional input and N-dimensional output, it is usually assumed that the sample data set is:

[0102]

[0103] Among them, x i is the input feature parameter, and y i is the output feature parameter; By introducing a loss function defined on the hypersphere, the purpose of extending the ε-loss function to multiple spatial dimensions is achieved, and its specific expression is defined as:

[0104]

[0105] where L(u) is the loss function; This definition method can weaken the interference of noise data on the calculation results and improve the anti-noise performance of the model. Therefore, this feature is particularly suitable for complex systems such as multivariable, nonlinear, and strong coupling;

[0106] In summary, the M-SVR problem is transformed into the following objective function optimization problem:

[0107]

[0108] where w and b are the normal vector and intercept of the hyperplane respectively, is the non-linear mapping function, C is the penalty factor, and L(·) is the loss function;

[0109] (3) Establish the NSGA-III optimized M-SVR model

[0110] The model constructed by selecting the RBF kernel function is superior to the traditional model composed of other kernel functions in terms of accuracy and root mean square error (RMSE), which not only ensures the generalization ability of the model but also reduces the calculation time. Its mathematical form is as follows:

[0111]

[0112] where x s is the sample data point, x c is the center point of the kernel function, and σ is the kernel function width parameter greater than 0;

[0113] The NSGA-III optimization problem can be described as:

[0114] min[h1(x s ), h2(x s ), …, h l (x s )]

[0115]

[0116] where h l (x s ) is the objective function, a1 and a2 are the upper and lower ranges of the variable x, and A ec×x s = B ec and A×x s <B is the constraint condition for x s where A ec , B ec , both A and B are characteristic matrices of the optimization problem;

[0117] (4) According to the geometric relationship for strip shape determination:

[0118]

[0119] where C ti is the crown of the strip at a certain moment, the average thickness of the strip at a certain moment;

[0120] C ti = h c - h e

[0121]

[0122] where h c is the thickness at the center of the plate, h e is the thickness at the edge of the plate, I is the strip shape index, i, j are the i-th and j-th units of the strip, ΔL ij is the length difference between the i-th unit and the j-th unit, b ij is the lateral distance between the i-th unit and the j-th unit, L is the reference unit length of the strip;

[0123] According to the geometric relationship for strip shape determination mentioned in (4), the strip shape state can be clearly known, namely ideal strip shape (the internal stress of the strip is evenly distributed along the width direction of the strip, and when the external stress on the strip is removed, the strip can still maintain good flatness), potential strip shape (the internal stress of the strip is unevenly distributed along the width direction of the strip, but the uneven internal stress of the strip is not sufficient to cause a change in flatness; when the external force is removed, the strip can still maintain flatness; however, when the strip is longitudinally cut, the potential internal stress causes irregular changes in the strip shape, resulting in warping), apparent strip shape (due to the uneven distribution of the internal stress of the strip in the width direction and the internal stress is not sufficient to resist the change in the flatness of the strip, causing warping deformation in local areas, and both removing the external force and longitudinally cutting the strip will exacerbate the apparent strip shape of the strip), mixed strip shape (the strip shape forms of different parts of the strip are different, such as part of the strip has a potential strip shape while other parts show an apparent strip shape) and strip shape affected by tension (if the internal stress generated by the tension is large enough to reduce the overall compressive stress to the level that can convert the apparent strip shape into a potential strip shape), the M-SVR model can automatically classify the three groups of information source data collected into three classifiers according to this geometric determination relationship, denoted as M1, M2, M3;

[0124] Step 4.3: Use the classifier described in Step 4.2 to classify the rolling mechanism data and process data according to the sample space described in Step 4.1, convert the output into a posterior probability, and form a basic probability assignment function with the parameter combination of the confusion matrix as evidence E1, E2, and E3. The calculation formula for the basic probability assignment function is:

[0125] m(S) = F(S) * H(S)

[0126] where F(S) is the converted posterior probability, and H(S) is the support degree correctly predicted as class S in the confusion matrix;

[0127] Apply the DS evidence theory to synthesize E1, E2, and E3. The evidence synthesis formula is:

[0128]

[0129] Define that the synthesis rule satisfies the following (taking S1 and S2 as examples):

[0130] If there exists:

[0131] Satisfy And there are ① m(S1) - m(S2) > ε1 (ε1 > 0); ② m(Θ) < ε2 (ε2 > 0); ③ m(S1) > m(Θ), then S1 is the prediction result, and the values of ε1 and ε2 are adjusted to the optimal according to multiple experiments. Among them, Θ is the frame of discernment, representing the set of all possible propositions.

[0132] Step 5: Establish a comprehensive flatness control strategy based on the warping limit and waviness discrimination model, as Figure 4 shown, including the following steps:

[0133] Step 5.1: Considering the secondary deformation of the strip between stands, the metal transverse flow model, the warping limit, and the flatness control ability of the finishing mill group simultaneously, establish a comprehensive flatness control strategy based on the warping limit and waviness discrimination model:

[0134]

[0135] Among them, G TC is the target flatness, f i * is the target flatness of the i-th stand, K f is the flatness correction coefficient, is the target cross-sectional shape, is the thickness reduction of the strip in the width direction, is the target crown of the i-th stand, is the total regulation amount of the strip target crown, is the sum of the warping limits of the rolling mill group, ΔCi is the crown control amount of the i-th stand, K ci is the crown correction coefficient;

[0136] Step 5.2: Use the output that meets the synthesis decision rule in Step 4.3 as the input of the comprehensive strip shape control strategy; consider the influence of the secondary deformation between stands on the strip, and calculate the natural exit strip shape of the strip; consider the law of transverse metal flow of the strip, and calculate the target strip shape value; according to the natural exit strip shape and the target strip shape of the strip, calculate the required target crown adjustment amount for each stand; check the limit value of the rolling mill equipment and the limit range of strip warping to ensure that the crown control amount of each stand is within the limit range of strip warping; use a full-automatic X-ray diffraction residual stress detector to detect the real-time strip shape, calculate the deviation between the measured value and the target value, and determine the corresponding rolling process parameters, such as bending roll force, roll gap value, rolling speed, and cooling water volume, etc., to complete the strip shape feedback control and realize the dynamic adjustment of the rolling working conditions to ensure the strip shape quality of the strip at the exit.

[0137] As mentioned above, it is only the preferred specific implementation manner of this application, but the protection scope of this application is not limited thereto. Any changes or substitutions that can be easily thought of by those skilled in the art within the technical scope disclosed in this application should be covered within the protection scope of this application. Therefore, the protection scope of this application should be subject to the protection scope of the claims.

Claims

1. A strip shape control method based on the DS evidence theory, characterized in that, It includes the following steps: Analyze relevant characteristic parameters and collect rolling process data; Analyze the strip deformation mechanism and construct a rolling mechanism model; Calculate rolling mechanism data based on the rolling mechanism model, and construct a machine learning coordination model for mechanism fusion data based on the rolling mechanism data and the rolling process data; Establish a strip exit shape prediction model for mechanism fusion data based on the DS evidence theory; Establish a comprehensive shape control strategy based on the warping limit and the waviness discrimination model, obtain rolling process parameters, perform feedback control on the shape based on the rolling process parameters, and dynamically adjust the rolling working conditions; The process of establishing a comprehensive shape control strategy based on the warping limit and the waviness discrimination model includes: Take the output that meets the synthesis rule as the input of the comprehensive shape control strategy, and calculate the natural exit shape of the strip based on the influence of the secondary deformation between stands on the strip; Calculate the target shape based on the following method: Among them, is the target sheet shape, is the target flatness of the i-th stand, is the flatness correction coefficient, is the target cross-sectional shape, is the thickness reduction of the strip in the width direction, is the target crown of the i-th stand, is the total regulation amount of the strip target crown, is the sum of the warping limits of the rolling mill set, is the crown regulation amount of the i-th stand, is the crown correction coefficient, is the thickness of the rolled piece at the exit; Calculate the required target crown adjustment amount for each stand based on the natural exit shape of the strip and the target shape; Check the limit values of the rolling mill equipment and the strip warping limit range, and control the target crown adjustment amount based on the limit values and the strip warping limit range; Detect the real-time shape of the strip based on a full-automatic X-ray diffraction method residual stress detector, and calculate the deviation between the measured value and the target value; Determine the corresponding rolling process parameters based on the deviation between the measured value and the target value.

2. The strip shape control method based on the DS evidence theory according to claim 1, wherein, The rolling process data includes rolling process data and exit strip index data; The rolling process data includes rolling force, bending roll force, roll gap value, roll shifting amount, rolling speed, inter-stand tension, rolling inlet temperature and outlet temperature; The exit strip index data includes the exit width, exit thickness, exit crown and exit flatness of the strip.

3. The strip shape control method based on the DS evidence theory according to claim 1, characterized in that, The process of analyzing the strip deformation mechanism and constructing a rolling mechanism model based on the rolling process data includes: Construct a strip critical warping model based on the shape control mechanism; Construct a warping correction model under tension based on the influence of inter-stand tension on the warping behavior; Construct a strip metal transverse flow model based on the influence of the metal transverse flow in the strip on the residual stress; Construct an inter-stand secondary deformation model based on the influence of inter-stand tension and the action of strip internal stress.

4. The strip shape control method based on the DS evidence theory according to claim 3, characterized in that The construction method of the strip critical warping model is: Among them, is the critical stress at which the strip warps, is the critical stress coefficient, is the elastic modulus of the strip, is the Poisson's ratio, is the thickness of the strip, is the width of the strip; The construction method of the warping correction model is: Among them, is the inter-stand tension, is the middle wave correction coefficient, is the edge wave correction coefficient, is the pre-rolling proportional crown, is the post-rolling proportional crown; The construction method of the strip metal transverse flow model is: Among them, is the transverse flow coefficient of the strip steel towards the edge, is the transverse flow coefficient of the strip steel towards the center, is the flow coefficient related to the width and thickness of the strip steel, is the flow coefficient of the strip steel edge, is the flow coefficient of the strip steel center; The construction method of the inter-stand secondary deformation model is: Among them, is the deformation coefficient, is the deformation resistance, is the time parameter, is the temperature influence coefficient, is the deformation index coefficient, is the transverse thickness distribution of the strip steel, is the aspect ratio influence coefficient, is the aspect ratio index coefficient.

5. The strip shape control method based on the DS evidence theory according to claim 1, wherein The process of calculating rolling mechanism data based on the rolling mechanism model and constructing a machine learning coordination model for mechanism fusion data based on the rolling mechanism data and the rolling process data includes: Calculate the corresponding rolling mechanism data based on the strip critical warping model, metal transverse flow model and inter-stand secondary deformation model; the rolling mechanism data includes the strip critical warping limit stress value, the transverse flow coefficient values of the strip towards the edge and the center respectively, the deformation resistance value of the strip undergoing secondary deformation between stands and the strip temperature value; Take the rolling mechanism data and the rolling process data as input features to construct a machine learning coordination model for mechanism fusion data.

6. The strip shape control method based on the DS evidence theory according to claim 1, wherein The process of establishing a strip steel export shape prediction model for mechanism fusion data based on DS evidence theory includes: Defining the sample space; Grouping the rolling mechanism data and the rolling process data to obtain grouped information sources; Calculating the grouped information sources as the training input values of a machine learning coordination model to obtain a classifier, and obtaining an evidence data group based on the classifier and the sample space; Synthesizing the evidence data group based on DS evidence theory and defining the synthesis rule.

7. The strip shape control method based on the DS evidence theory according to claim 6, characterized in that, The method of defining the sample space is: G = {IDS, POS, APS, MPS, TAS} where G is the sample space, IDS is the ideal shape, POS is the potential shape, APS is the apparent shape, MPS is the mixed shape, and TAS is the shape affected by tension.

8. The strip shape control method based on the DS evidence theory according to claim 6, characterized in that, The process of calculating the grouped information sources as the training input values of a machine learning coordination model includes: Constructing an M-SVR model and calculating based on the M-SVR model; Optimizing the M-SVR model based on NSGA-III; Calculating the geometric relationship for shape determination, and the M-SVR model obtaining a classifier based on the geometric relationship for shape determination.

Images