Collaborative integrated plate shape presetting control method, device and system and storage medium

Abstract

The invention discloses a collaborative integrated plate shape presetting control method, device and system, and a storage medium. The method comprises the following steps: establishing a preset table model by using historical production big data; establishing a preset machine learning tree model according to the incoming material and process information of the current rolled steel coil; through a preset table model and a preset machine learning tree model, a final preset value is output through collaborative integration based on a 3sigma principle. By adopting the technical scheme of the invention, accurate plate shape control with anti-interference capability is realized.

Description

Technical Field

[0001] The present invention belongs to the technical field of cold-rolled strip steel, and in particular relates to a collaborative integrated plate shape presetting control method and device, system, and storage medium. Background Art

[0002] Shape is an important quality indicator of cold-rolled strip products. As the requirements for product quality become more and more stringent, improving the accuracy of shape control has received more and more attention. With the development of industrial big data and artificial intelligence technology, shape control based on data-driven and artificial intelligence algorithms has become an important research direction. Cold-rolled shape control includes preset control and closed-loop feedback control. Preset control is an important means to ensure the quality of the leading shape, and it is also an important basis for ensuring that each shape adjustment method quickly reaches the stable feedback stage.

[0003] The preset parameters of rolling are generally selected according to the convexity of the incoming material. At present, there are almost no corresponding convexity meters in domestic cold rolling mills, and the convexity of the incoming material cannot be detected. Therefore, most cold rolling mills currently analyze and set preset values ​​based on their own incoming material characteristics and manual experience. The preset values ​​are not standardized enough, which will lead to plate shape defects, and the historical data of the unit has not been fully mined and utilized. The plate shape preset system needs to be further improved and optimized. The existing preset models can be roughly divided into two categories: traditional mathematical analytical models and intelligent models that apply machine learning. Due to the complexity of the cold rolling process and the many variables involved, traditional mathematical analytical models often need to be greatly simplified. However, the simplified model may not guarantee the required accuracy, which may lead to unsatisfactory results in practical applications. Summary of the invention

[0004] The technical problem to be solved by the present invention is to provide a collaborative integrated plate shape pre-setting control method and device, system, and storage medium to achieve precise plate shape control with anti-interference capability.

[0005] To achieve the above object, the present invention adopts the following technical solution:

[0006] A collaborative integrated plate shape presetting control method, comprising:

[0007] Use historical production big data to build pre-set table models;

[0008] Establish a pre-set machine learning tree model based on the incoming material and process information of the current rolled steel coil;

[0009] Through the preset table model and the preset machine learning tree model, the final preset value is output through collaborative integration based on the 3sigma principle.

[0010] Preferably, DPC density peak clustering is used to establish a preset table model using historical production big data.

[0011] As a preference, the output final preset value WRB is:

[0012] WRB=α1WRB1+α2WRB2

[0013] Among them, WRB1 refers to the preset value obtained by querying from the preset model based on production history data, WRB2 refers to the preset value predicted by the preset machine learning tree model considering the characteristics of incoming materials, α1 represents the weight of the preset value queried from the preset model based on production history data, α2 represents the weight of the preset value predicted by the preset machine learning tree model, α1+α2=1.

[0014] The present invention also provides a collaborative integrated plate shape presetting control device, comprising:

[0015] The first processing module is used to establish a preset table model using historical production big data;

[0016] The second processing module is used to establish a preset machine learning tree model according to the incoming material and process information of the current rolled steel coil;

[0017] The third processing module is used to output a final preset value through collaborative integration based on the 3sigma principle through a preset table model and a preset machine learning tree model.

[0018] Preferably, the first processing module uses DPC density peak clustering to establish a preset table model using historical production big data.

[0019] As a preference, the output final preset value WRB is:

[0020] WRB=α1WRB1+α2WRB2

[0021] Among them, WRB1 refers to the preset value obtained by querying from the preset model based on production history data, WRB2 refers to the preset value predicted by the preset machine learning tree model considering the characteristics of incoming materials, α1 represents the weight of the preset value queried from the preset model based on production history data, α2 represents the weight of the preset value predicted by the preset machine learning tree model, α1+α2=1.

[0022] An embodiment of the present invention also provides a collaborative integrated plate shape preset control system, comprising: a memory and a processor, wherein the memory stores a computer program executed by the processor, and the computer program executes a collaborative integrated plate shape preset control method when executed by the processor.

[0023] An embodiment of the present invention further provides a storage medium, on which a computer program is stored, and the computer program executes the collaborative integrated plate shape presetting control method when running.

[0024] The present invention uses a clustering method for the cold rolling process and utilizes historical production big data to establish a preset table model, and the preset value reflects the common information of the rolled steel coils; a preset machine learning tree model is established according to the incoming material and process information of the current rolled steel coils, and the preset value reflects the individual information of each rolled steel coil; the two models are synergistically integrated and output based on the 3sigma principle to output the final preset value, which can simultaneously reflect the commonality contained in the historical production data and the individuality of the specific steel coil, improve the robustness of the preset control, and thus achieve precise plate shape control with anti-interference ability. BRIEF DESCRIPTION OF THE DRAWINGS

[0025] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on the provided drawings without paying creative work.

[0026] Figure 1 This is a flowchart of the collaborative integrated plate shape presetting control method of the present invention.

[0027] Figure 2a It is the DPC clustering decision diagram;

[0028] Figure 2b Scatter plot to find cluster centers;

[0029] Figure 3 Output the final preset value flow chart for the two models based on the 3sigma principle;

[0030] Figure 4 This is a distribution diagram of preset values ​​for a steel coil. DETAILED DESCRIPTION

[0031] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0032] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, the present invention is further described in detail below with reference to the accompanying drawings and specific embodiments.

[0033] Embodiment 1:

[0034] An embodiment of the present invention provides a collaboratively integrated plate shape preset control method, which aims to address the preset control problem of cold-rolled plate shape. It uses a data-driven approach based on historical big data collected on-site to establish a preset table model based on production history data, and a preset prediction model based on incoming material information. The two models are collaboratively integrated to output the final preset value.

[0035] like Figure 1 As shown, an embodiment of the present invention provides a collaborative integrated plate shape presetting control method, comprising:

[0036] Step 1: Use clustering methods to build a pre-set table model using historical production big data.

[0037] Use DPC density clustering to calculate the local density ρ of WRB sample points under each strategy number i and the relative distance δ i Perform calculations.

[0038] The local density is ρ i The calculation formula is as follows:

[0039]

[0040] Among them, the data point x i To data point x j The distance is d ij ;d c For data point x i The neighborhood cutoff distance threshold is the percentage of the total data volume, which is generally 1% or 2%. In this paper, considering the size of the sample data, d c =2%.

[0041] Relative distance σ i Refers to sample x i The minimum distance to other points with higher density. For the sample with the highest density, the relative distance is calculated as follows:

[0042]

[0043] For other data points, the relative distance calculation formula is as follows:

[0044]

[0045] In order to find the density peak, the decision value γ is defined as follows:

[0046] γ i =ρ i ×δ i

[0047] Will γ i By sorting from large to small, the first k data points can be automatically selected as k cluster centers. In this paper, the value of the maximum density peak point is selected as the set value, that is, one cluster center is selected. Figure 2a As shown, select the largest decision value in the upper right corner as the cluster center; find the cluster center scatter plot as shown Figure 2b As shown, the horizontal axis corresponding to the cluster center is the desired WRB preset value.

[0048] The obtained preset values ​​are collated to form a preset table model established using historical production big data.

[0049] Step 2: Establish a preset machine learning tree model based on the incoming material and process information of the current rolled steel coil.

[0050] Since the purpose of modeling this preset model is to reflect the individual information of the current steel coil, the features are selected according to the factory experience, and the tree model itself has the characteristics of feature screening, irrelevant features will not cause too much interference to the results. The input features are the incoming material information and process information of the current steel coil. The output is the normalized work roll bending force WRB. The input features are in normalized form, with a maximum value of 1, corresponding to a true value of 1000kN.

[0051] The training results of the machine learning tree model are shown in Table 1. The prediction error MAE of the working roll bending force is 0.0253, and the prediction error of the intermediate roll bending force is 0.0324.

[0052] Table 1

[0053] Prediction object MSE RMSE MAE Work roll bending force WRB 0.0015 0.0388 0.0253

[0054] This model is used as a preset machine learning tree model based on the incoming material and process information of the current rolled steel coil.

[0055] Step 3: The two models are synergistically integrated based on the 3sigma principle to output the final preset value.

[0056] The set value of the pre-set model based on production history data is generated based on historical big data statistics, which is general and universal; the set value of the pre-set machine learning tree model is generated based on the prediction of a specific strip steel situation of a roll, which can reflect the personality and characteristics of the strip steel roll. Using the idea of ​​weighted average method in collaborative ensemble learning, the two models are collaboratively integrated, and the final output value can simultaneously reflect the generality of historical statistical data and the personality of a specific roll.

[0057] It can be found that the set value data has a law that when the amount of data is sufficient, the set value approaches the normal distribution. According to the "3sigma principle" of normal distribution, the probability of the value being distributed in (μ-σ, μ+σ) is about 0.6826, the probability of the value being distributed in (μ-2σ, μ+2σ) is about 0.9544, the probability of the value being distributed in (μ-3σ, μ+3σ) is about 0.9974, and the probability of the value falling outside (μ-3σ, μ+3σ) is less than 0.003, which is considered an outlier. Among them, μ represents the mean, that is, the symmetry axis of the normal distribution, and σ represents the data standard deviation, which describes the degree of dispersion of the data.

[0058] The idea of ​​synergistic integration of the two models is as follows Figure 3 As shown in the figure, when the amount of data in the data set is greater than 1000, the data is considered to be close to normal distribution, and the value of the preset model considering the characteristics of the incoming materials is compared with the μ value of the normal distribution. If the predicted data falls within (μ-σ,μ+σ), a weight of 0.25 is assigned to the preset model value based on the production history data, and a weight of 0.75 is assigned to the preset machine learning tree model prediction value, so as to reflect the personality of the data as much as possible; if the predicted data falls within (μ-2σ,μ-σ)∪(μ+σ,μ+2σ), a weight of 0.5 is assigned to the preset model value based on the production history data, and a weight of 0.75 is assigned to the preset machine learning tree model prediction value. The predicted value of the learning tree model is assigned a weight of 0.5, with the individuality and generality of the data each accounting for half; if the predicted data falls within (μ-3σ,μ-2σ)∪(μ+2σ,μ+3σ), a weight of 0.75 is assigned to the preset model value based on the production history data, and a weight of 0.25 is assigned to the predicted value of the preset machine learning tree model, with the generality of the historical data dominating; if the predicted data falls outside the range of (μ-3σ,μ+3σ), it is considered that the predicted value of the preset machine learning tree model is far beyond the normal data range of the historical statistical data and is regarded as an outlier. The final set value is based on the historical statistical data.

[0059] As shown in Table 2, α1 represents the weight of the preset value queried from the preset model based on production history data, α2 represents the weight of the preset value predicted by the preset machine learning tree model, and α1+α2=1 always holds. The final output preset value is shown in the following formula:

[0060] WRB=α1WRB1+α2WRB2

[0061] Among them, WRB1 refers to the preset value obtained by querying the preset model based on production history data, and WRB2 refers to the preset value predicted by the preset machine learning tree model considering the characteristics of incoming materials.

[0062] Table 2

[0063] Interval number The range of predicted values <![CDATA[α1]]> <![CDATA[α2]]> ① μ-σ,μ+σ 0.25 0.75 ② μ-2σ,μ-σ∪μ+σ,μ+2σ 0.50 0.50 ③ μ-3σ,μ-2σ∪μ+2σ,μ+3σ 0.75 0.25 ④ -∞,μ-3σ∪μ+3σ,+∞ 1.00 0.00

[0064] like Figure 4 As shown in the figure, the prediction result of the tree model for the steel coil is that the working roll bending force WRB is 0.334, and the predicted WRB is within the range of 1sigma. According to the collaborative integration principle shown in Table 2, the final calculated setting value is WRB 0.336.

[0065] The set value of the preset model based on production history data in the embodiment of the present invention is generated based on historical big data statistics, which is general and universal; the set value of the preset machine learning tree model is generated based on the prediction of a specific strip steel situation of a certain roll, which can reflect the personality and characteristics of the strip steel of the roll. The two models are collaboratively integrated using the weighted average idea in collaborative ensemble learning, and the final output value can simultaneously reflect the commonality contained in the historical production data and the personality of the specific steel roll.

[0066] Embodiment 2:

[0067] The embodiment of the present invention further provides a collaborative integrated plate shape presetting control device, comprising:

[0068] The first processing module is used to establish a preset table model using historical production big data;

[0069] The second processing module is used to establish a preset machine learning tree model according to the incoming material and process information of the current rolled steel coil;

[0070] The third processing module is used to output a final preset value through collaborative integration based on the 3sigma principle through a preset table model and a preset machine learning tree model.

[0071] As an implementation method of an embodiment of the present invention, the first processing module uses DPC density peak clustering to establish a preset table model using historical production big data.

[0072] As an implementation method of the embodiment of the present invention, the final preset value WRB is output as:

[0073] WRB=α1WRB1+α2WRB2

[0074] Among them, WRB1 refers to the preset value obtained by querying from the preset model based on production history data, WRB2 refers to the preset value predicted by the preset machine learning tree model considering the characteristics of incoming materials, α1 represents the weight of the preset value queried from the preset model based on production history data, α2 represents the weight of the preset value predicted by the preset machine learning tree model, α1+α2=1.

[0075] Embodiment 3:

[0076] An embodiment of the present invention also provides a collaborative integrated plate shape preset control system, comprising: a memory and a processor, wherein the memory stores a computer program executed by the processor, and the computer program executes a collaborative integrated plate shape preset control method when executed by the processor.

[0077] Embodiment 4:

[0078] An embodiment of the present invention further provides a storage medium, on which a computer program is stored, and the computer program executes the collaborative integrated plate shape presetting control method when running.

[0079] The embodiments described above are only descriptions of the preferred embodiments of the present invention and are not intended to limit the scope of the present invention. Without departing from the design spirit of the present invention, various modifications and improvements made to the technical solutions of the present invention by ordinary technicians in this field should all fall within the protection scope determined by the claims of the present invention.

Claims

1. A collaborative integrated plate shape presetting control method, characterized in that: include: Use historical production big data to build pre-set table models; Establish a pre-set machine learning tree model based on the incoming material and process information of the current rolled steel coil; Through the preset table model and the preset machine learning tree model, the final preset value is output through collaborative integration based on the 3sigma principle.

2. The collaborative integrated plate shape presetting control method according to claim 1, characterized in that: Use DPC density peak clustering to build a pre-set table model using historical production big data.

3. The collaborative integrated plate shape presetting control method according to claim 2, characterized in that: The final preset output value WRB is: WRB=α1WRB1+α2WRB2 Among them, WRB1 refers to the preset value obtained by querying from the preset model based on production history data, WRB2 refers to the preset value predicted by the preset machine learning tree model considering the characteristics of incoming materials, α1 represents the weight of the preset value queried from the preset model based on production history data, α2 represents the weight of the preset value predicted by the preset machine learning tree model, α1+α2=1.

4. A collaborative integrated plate shape presetting control device, characterized in that: include: The first processing module is used to establish a preset table model using historical production big data; The second processing module is used to establish a preset machine learning tree model according to the incoming material and process information of the current rolled steel coil; The third processing module is used to output a final preset value through collaborative integration based on the 3sigma principle through a preset table model and a preset machine learning tree model.

5. The collaborative integrated plate shape presetting control device according to claim 4, characterized in that: The first processing module uses DPC density peak clustering and uses historical production big data to establish a preset table model.

6. The collaborative integrated plate shape presetting control device according to claim 5, characterized in that: The final preset output value WRB is: WRB=α1WRB1+α2WRB2 Among them, WRB1 refers to the preset value obtained by querying from the preset model based on production history data, WRB2 refers to the preset value predicted by the preset machine learning tree model considering the characteristics of incoming materials, α1 represents the weight of the preset value queried from the preset model based on production history data, α2 represents the weight of the preset value predicted by the preset machine learning tree model, α1+α2=1.

7. A collaborative integrated plate shape presetting control system, characterized in that: include: A memory and a processor, wherein the memory stores a computer program executed by the processor, and when the computer program is executed by the processor, the collaborative integrated plate shape presetting control method as described in any one of claims 1 to 3 is executed.

8. A storage medium, characterized in that: The storage medium stores a computer program, which, when running, executes the collaborative integrated plate shape presetting control method as described in any one of claims 1 to 3.

Images