Self-learning leveler presetting method and device suitable for hot-dip galvanizing production line

Abstract

The present invention discloses a self-learning presetting method and device for a hot-dip galvanizing production line, which obtains the data of incoming material parameters of hot-dip galvanized sheet, process parameters of the flattening machine, temperature difference information and quality index of the hot-dip galvanized sheet; the data is input into the self-learning presetting model of the flattening machine to obtain the pre-set values ​​of the rolling force, inlet tension and outlet tension of the flattening machine; the self-learning presetting model of the flattening machine is obtained by neural network training. The present invention can ensure both the mechanical performance index of the hot-dip galvanized sheet and the surface quality of the hot-dip galvanized sheet.

Description

Technical Field

[0001] The present invention relates to a setting technology of a leveler in a hot-dip galvanizing unit, and more specifically to a self-learning leveler presetting method and device suitable for a hot-dip galvanizing production line. Background Art

[0002] The continuous hot-dip galvanizing production line galvanizes the surface of the strip steel, and the flattening process is the last step of the continuous production line. The flattening process forms the required roughness on the zinc layer surface, thereby improving the adhesion of the hot-dip galvanized sheet coating. At the same time, the flattening process can improve the surface quality of the zinc layer and reduce the impact of defects such as zinc slag on subsequent use. By controlling the elongation of the flattening machine, the yield platform can be eliminated or less obvious, which can prevent the appearance of slip lines during stretching or deep drawing.

[0003] The annealing furnace temperature, leveler rolling force and tension of the hot-dip galvanizing production line play an important role in the elongation of the hot-dip galvanized sheet. The annealing furnace temperature, zinc pot temperature, leveler rolling force and leveler roll roughness of the hot-dip galvanizing production line have an important direct effect on the surface quality of the hot-dip galvanized sheet. In the traditional leveler rolling force and tension setting method, it is mainly considered that the leveler incoming material is cold-rolled sheet or annealed cold-rolled sheet. In the leveler rolling force and tension setting method for hot-dip galvanized sheet, the production situation of the previous process of the leveler is not considered. Therefore, the operator needs to intervene based on experience to ensure the product quality of the hot-dip galvanized sheet.

[0004] The production process of the existing hot-dip galvanizing production line is as follows: Figure 1 As shown, the flattening machine is located in the post-process of the hot-dip galvanizing production line and plays an important role in the quality of hot-dip galvanizing products. Figure 1 In s , W, h, R as They represent the yield strength, width, thickness and surface roughness of the raw steel coil respectively; S1 represents the soaking section temperature of the continuous annealing furnace; S2 represents the final cooling section temperature of the continuous annealing furnace; R a0 , L, v, ε, T0, P, T1 represent the original roughness of the skin-pass mill roll, rolling length, skin-passing speed, elongation of the hot-dip galvanized sheet, inlet hot-dip galvanized sheet tension, skin-passing rolling force, and outlet hot-dip galvanized sheet tension, respectively; R a Indicates the surface roughness of hot-dip galvanized sheet; the surface roughness of hot-dip galvanized sheet is tested at the hot-dip galvanizing exit inspection station.

[0005] The flattening process of existing hot-dip galvanizing production lines is as follows: Figure 2 As shown in the figure, T0, P, and T1 represent the entrance hot-dip galvanized sheet tension, the flat rolling force, and the exit hot-dip galvanized sheet tension, respectively. By accurately presetting T0, P, and T1, a good hot-dip galvanized sheet elongation can be obtained.

[0006] Traditional leveling machine pre-setting technology considers the following functional relationship:

[0007] ε=f1(σ s ,W,h,P,T0,T1,v) Formula (1)

[0008] The surface roughness of the traditional flattened strip is considered to have the following functional relationship:

[0009] R a =f2(R as ,R a0 ,P,L) Formula (2)

[0010] In existing patent applications, such as publication number CN105404155A, a comprehensive optimization setting technology for plate surface quality control parameters suitable for hot-dip galvanizing leveler is introduced. This patent does not consider the leveler pre-setting technology for hot-dip galvanized sheet elongation control.

[0011] For example, patent applications No. 200510024683, 200710185706, 200710185706, 201310031951, 201310276037, and 201511028097 use different methods to set rolling force and tension. The above patents mainly consider that the incoming material of the leveler is cold-rolled sheet or annealed cold-rolled sheet, and do not consider that the incoming material of the leveler is hot-dip galvanized sheet. Summary of the invention

[0012] In view of the above-mentioned defects existing in the prior art, the purpose of the present invention is to provide a self-learning leveling machine pre-setting method and device suitable for a hot-dip galvanizing production line, which can not only ensure the mechanical performance indicators of the hot-dip galvanized sheet, but also ensure the surface quality of the hot-dip galvanized sheet.

[0013] To achieve the above object, the present invention adopts the following technical solution:

[0014] The present invention provides a self-learning leveler presetting method suitable for a hot-dip galvanizing production line:

[0015] Obtain data on incoming hot-dip galvanized sheet material parameters, leveler process parameters, temperature difference information, and quality indicators of hot-dip galvanized sheet;

[0016] Inputting the data into a self-learning preset model of a leveling mill to obtain preset values ​​of rolling force, inlet tension, and outlet tension of the leveling mill;

[0017] The self-learning preset model of the leveling machine is obtained through neural network training.

[0018] Preferably, the neural network is a BP neural network.

[0019] The data is used as training samples of the BP neural network, and the BP neural network is trained to obtain the self-learning preset model of the leveling machine.

[0020] Preferably, the BP neural network adopts an error back propagation algorithm for learning and training.

[0021] Preferably, the incoming material parameters of the hot-dip galvanized sheet include:

[0022] Yield strength of raw steel coil σ s , width W, thickness h; and / or

[0023] The process parameters of the leveling machine include:

[0024] Leveling speed v, rolling length L of the roll, original roughness R of the roll a0 ; and / or

[0025] The temperature difference information includes:

[0026] Annealing furnace temperature difference S1-S2, zinc pot and final cooling temperature difference S3-S2; and / or

[0027] The quality indicators of hot-dip galvanized sheet include:

[0028] Elongation ε of hot-dip galvanized sheet and roughness R of finished product a .

[0029] Preferably, the input layer includes 10 units, the middle layer includes 18 units, and the output layer includes 3 units.

[0030] The present invention also provides a self-learning leveler presetting device suitable for a hot-dip galvanizing production line, comprising:

[0031] Collection module: used to obtain the data of incoming parameters of hot-dip galvanized sheet, process parameters of leveling machine, temperature difference information and quality index of hot-dip galvanized sheet;

[0032] Processing module: used for inputting the data into the self-learning preset model of the leveling machine to obtain the preset value of the leveling machine, and the self-learning preset model of the leveling machine is obtained through neural network training.

[0033] Preferably, the neural network is a BP neural network.

[0034] The processing module uses the data as training samples of the BP neural network, performs learning and training on the BP neural network, and obtains the self-learning preset model of the leveling machine.

[0035] Preferably, the BP neural network adopts an error back propagation algorithm for learning and training.

[0036] BP neural network is a three-layer neural network structure, including input layer, middle layer and output layer.

[0037] Preferably, the incoming material parameters of the hot-dip galvanized sheet include: the yield strength σ of the raw steel coil s , width W, thickness h; and / or

[0038] The process parameters of the leveling machine include: leveling speed v, rolling length L of the roller, original roughness R of the roller a0 ; and / or

[0039] The temperature difference information includes: annealing furnace temperature difference S1-S2, zinc pot and final cooling temperature difference S3-S2; and / or

[0040] The quality indicators of hot-dip galvanized sheet include: elongation ε of hot-dip galvanized sheet, roughness R of finished product a .

[0041] Preferably, the input layer includes 10 units, the middle layer includes 18 units, and the output layer includes 3 units.

[0042] The present invention provides a self-learning leveler presetting method and device suitable for a hot-dip galvanizing production line, which can ensure that the elongation of the hot-dip galvanized sheet meets the standard and the surface quality of the hot-dip galvanized sheet, and can also reduce the manual intervention of the hot-dip galvanizing leveler operator. BRIEF DESCRIPTION OF THE DRAWINGS

[0043] Figure 1 It is a schematic diagram of the process of the existing hot-dip galvanizing production process;

[0044] Figure 2 This is a schematic diagram of the flattening process principle of the existing hot-dip galvanizing production line;

[0045] Figure 3 It is a flow chart of a method for presetting a self-learning leveler for a hot-dip galvanizing production line according to the present invention;

[0046] Figure 4 It is a structural schematic diagram of a self-learning pre-setting model of a flattener in a self-learning flattener pre-setting method applicable to a hot-dip galvanizing production line of the present invention. DETAILED DESCRIPTION

[0047] In order to better understand the above technical solution of the present invention, the technical solution of the present invention is further described below in conjunction with the accompanying drawings and embodiments.

[0048] Please combine Figure 3 As shown, the present invention provides a self-learning leveling machine presetting method suitable for a hot-dip galvanizing production line:

[0049] S001, obtain the data of incoming material parameters of hot-dip galvanized sheet, process parameters of leveling machine, temperature difference information and quality index of hot-dip galvanized sheet;

[0050] S002, inputting data into the self-learning preset model of the leveling machine;

[0051] S003, obtaining the preset value of the leveling machine.

[0052] Among them, the self-learning preset model of the leveling machine is obtained through BP neural network training.

[0053] The present invention uses the incoming material parameters of hot-dip galvanized sheet, the process parameters of the leveling machine, the temperature difference information and the quality index of the hot-dip galvanized sheet as the input of the leveling machine self-learning preset model, and the leveling machine preset value is used as the output of the leveling machine self-learning preset model. The structure of the leveling machine self-learning preset model is as follows: Figure 4 As shown, when establishing a three-layer leveling machine self-learning preset model with a corresponding relationship between model input and model output, an intermediate layer of 18 units is set, that is, the three-layer leveling machine self-learning preset model has 10 units in the input layer, 18 units in the intermediate layer, and 3 units in the output layer.

[0054] Among them, the incoming parameters of hot-dip galvanized sheet include: the yield strength of the raw steel coil σ s , width W, thickness h.

[0055] The process parameters of the leveling machine include: leveling speed v, rolling length L of the roll, original roughness R of the roll a0 .

[0056] The temperature difference information includes: annealing furnace temperature difference S1-S2, zinc pot and final cooling temperature difference S3-S2.

[0057] The quality indicators of hot-dip galvanized sheet include: elongation ε of hot-dip galvanized sheet, roughness R of finished product a .

[0058] The preset values ​​of the leveling mill include: the rolling force P of the leveling mill, the tension of the hot-dip galvanized sheet strip at the leveling mill entrance T0, and the tension of the hot-dip galvanized sheet strip at the leveling mill exit T1.

[0059] For a certain type of hot-dip galvanized steel sheet, collect and record the incoming material parameters and elongation ε requirements of each roll of hot-dip galvanized steel sheet; collect and record the measured values ​​of the hot-dip galvanizing process parameters of each roll of hot-dip galvanized steel sheet before the leveling mill; collect and record the original roughness R of the roll after the leveling mill roll is changed. a0; When the elongation ε of each roll of hot-dip galvanized sheet meets the standard, collect and record the length value of the hot-dip galvanized sheet at different leveling speeds v and the corresponding rolling force P of the leveling machine, the hot-dip galvanized sheet strip tension T0 at the leveling machine entrance, and the hot-dip galvanized sheet strip tension T1 at the leveling machine exit; collect and record the total length of each roll of hot-dip galvanized sheet and the number of rolls passing through the leveling machine, so as to calculate the leveling rolling length; after the hot-dip galvanized sheet after leveling enters the inspection table of the unit, the roughness of the hot-dip galvanized sheet at the length position of the hot-dip galvanized sheet with the elongation ε meeting the standard is checked, and the finished product roughness R is collected and recorded. a . Repeat the above steps several times.

[0060] The obtained multiple groups of data are used as training samples of the BP neural network, and the error back propagation algorithm is used to train the BP neural network to obtain the self-learning preset model of the skin pass mill for the hot-dip galvanized steel sheet.

[0061] After the trained self-learning preset model of the leveler is put into the control system of the hot-dip galvanizing leveler, the relevant process parameters and product quality requirements of the hot-dip galvanizing production line are input to obtain good preset values ​​of the hot-dip galvanizing leveler.

[0062] Due to the particularity of the hot-dip galvanizing unit process, the performance and surface characteristics of the hot-dip galvanized sheet are not only related to the incoming material characteristics and the process parameters of the leveler, but also to the temperature of the continuous annealing furnace and the zinc pot.

[0063] The elongation ε of hot-dip galvanized sheet and the yield strength σ of raw steel coil s , width W, thickness h, temperature difference s1-s2 between the soaking section and the final cooling section of the annealing furnace, speed v of the leveling mill, rolling length L of the leveling mill roll, rolling force P of the leveling mill, tension T0 of the hot-dip galvanized sheet strip at the leveling mill entrance, and tension T1 of the hot-dip galvanized sheet strip at the leveling mill exit, which can be described by the following function:

[0064] ε=f1(σ s ,W,h,P,T0,T1,v,L,s1-s2) Formula (3)

[0065] After the raw steel coil is galvanized, the raw material roughness does not affect the surface roughness of the hot-dip galvanized sheet. The surface roughness of the hot-dip galvanized sheet is related to the temperature difference between the final cooling section of the annealing furnace and the zinc pot, the original roughness of the skin-pass mill roll, the skin-pass mill rolling force P and the skin-pass mill roll rolling length L, which can be described by the following function:

[0066] R a =f2(s3-s2,R a0 ,P,L) Formula (4)

[0067] In order to obtain satisfactory elongation and surface quality, it is necessary to set reasonable rolling force, entrance tension and exit tension of the skin-pass mill.

[0068] The metal rolling process is a complex system with nonlinearity, multiple variables and strong coupling, and its mathematical model is very difficult to establish. For the traditional temper rolling process, it is difficult to take all influencing factors into account with a simple mathematical model. Therefore, the corresponding model calculation also has certain assumptions and simplifications, which makes it difficult to ensure that it has high calculation accuracy. Similarly, it is difficult to establish an accurate mathematical model for the temper rolling process of the hot-dip galvanizing production line. Therefore, it is necessary to find a temper rolling force and tension calculation model with higher calculation accuracy.

[0069] Artificial neural network refers to the simulation of human neural network from the perspective of information transmission, which combines existing information into different networks in different ways to realize information processing. Artificial neural network calculates the parameters of the model through sample data and propagation algorithm without analyzing the specific relationship principle of the problem.

[0070] Since it is difficult to establish a deterministic functional relationship between formulas (3) and (4), it is impossible to calculate the rolling force P of the leveler, the hot-dip galvanized sheet tension T0 at the leveler inlet, and the hot-dip galvanized sheet tension T1 at the leveler outlet through formulas (3) and (4). Therefore, a leveler presetting method provided by the present invention is used to obtain good performance and surface quality.

[0071] Example

[0072] In this embodiment, the raw material steel grade of the hot-dip galvanized sheet is AP0961E1, and the yield strength σ s The pressure is 600MPa, the thickness h is 0.5mm~1.3mm, the width W is 800mm~1430mm, the elongation ε required by the user is 0.8%~1.5%, and the finished roughness R of the hot-dip galvanized sheet is a The original roughness of the roller is 0.6μm~1.9μm. a It is 2.0μm~2.8μm.

[0073] Collect the elongation index, annealing furnace temperature, zinc pot temperature, roller rolling mileage, and corresponding original roughness of the roller of hot-dip galvanized sheets of different specifications; when the actual value of the elongation meets the standard, collect the actual value of the leveling machine operating speed, inlet tension, outlet tension, rolling force and roughness of the hot-dip galvanized sheet.

[0074] Repeat the above steps to obtain m groups of hot-dip galvanizing production line process parameters, flattening process parameters and hot-dip galvanized sheet product quality of hot-dip galvanized sheets of different specifications of the steel grade. The input and output parameter groups are:

[0075]

[0076] Where:

[0077] i represents the parameter group number;

[0078] σ si represents the yield strength of the i-th parameter group;

[0079] W i represents the strip width of the i-th parameter group;

[0080] h i represents the strip thickness of the i-th parameter group;

[0081] (s1-s2) i represents the temperature difference between the soaking section and the final cooling section of the annealing furnace of the i-th parameter group;

[0082] (s3-s2) i represents the temperature difference between the final cooling section and the zinc pot of the annealing furnace of the i-th parameter group;

[0083] R a0i represents the original roughness of the roll of the i-th parameter group;

[0084] v i represents the running speed of the leveling machine of the i-th parameter group;

[0085] L i represents the roller rolling mileage of the i-th parameter group;

[0086] ε i represents the elongation of the i-th parameter group;

[0087] R ai represents the roughness of the hot-dip galvanized sheet of the i-th parameter group;

[0088] P i represents the actual value of the skin-pass mill rolling force of the i-th parameter group;

[0089] T 0i represents the actual value of the tension at the leveling machine entrance of the i-th parameter group;

[0090] T 1i It represents the actual value of the tension at the exit of the leveling machine of the i-th parameter group.

[0091] The m sets of data obtained are used as training samples of the BP neural network, and the BP neural network is trained using the error back propagation algorithm to obtain the self-learning preset model of the leveler for the steel grade. In this embodiment, a total of 168 sets of data are collected for the steel grade AP0961E1 as training samples to train the BP neural network.

[0092] The trained model was used for online application to test the preset values ​​of the skin pass mill of 9 rolls of hot-dip galvanized sheet and strip. The accuracy of the preset model is shown in Table 1.

[0093] Table 1

[0094]

[0095] In summary, the present invention uses hot-dip galvanizing process parameters that can obtain good hot-dip galvanized sheet product quality and measured data of the leveler to train the neural network through the error back propagation algorithm, thereby obtaining a self-learning pre-set model of the leveler of the hot-dip galvanizing unit. After the model is put into the hot-dip galvanizing leveler control system, the rolling force and tension of the hot-dip galvanizing leveler can be accurately set, thereby ensuring the product quality of the hot-dip galvanized sheet. This method can be applied to other domestic hot-dip galvanizing units and has good application prospects.

[0096] Those skilled in the art should recognize that the above embodiments are only used to illustrate the present invention, and are not intended to limit the present invention. As long as they are within the spirit of the present invention, any changes or modifications to the above embodiments will fall within the scope of the claims of the present invention.

Claims

1. A self-learning leveler presetting method suitable for a hot dip galvanizing production line, characterized in that: Obtain data on incoming hot-dip galvanized sheet material parameters, leveler process parameters, temperature difference information, and quality indicators of hot-dip galvanized sheet; Inputting the data into a self-learning preset model of a leveling mill to obtain preset values ​​of rolling force, inlet tension, and outlet tension of the leveling mill; The self-learning preset model of the flattening machine is obtained through neural network training, The neural network is a BP neural network. The data is used as a training sample of the BP neural network, and the BP neural network is trained to obtain a self-learning preset model of the leveling machine. The BP neural network adopts the error back propagation algorithm for learning and training. The hot dip galvanized sheet material parameters include: yield strength of raw steel coil σ s ,width W ,thickness h ; The process parameters of the leveling machine include: leveling speed v , Rolling length of the roller L , Original roughness of the roll R a0 ; The temperature difference information includes: annealing furnace temperature difference S1-S2 , Temperature difference between zinc pot and final cooling S3-S2 ; The quality indicators of hot-dip galvanized sheet include: elongation of hot-dip galvanized sheet ε , Finished product roughness R a .

2. The self-learning skin-pass mill presetting method applicable to a hot-dip galvanizing production line according to claim 1, characterized in that: BP neural network is a three-layer neural network structure, including input layer, middle layer and output layer.

3. The self-learning skin-pass mill presetting method applicable to a hot-dip galvanizing production line according to claim 2, characterized in that: The input layer includes 10 units, the middle layer includes 18 units, and the output layer includes 3 units.

4. A self-learning leveler presetting device suitable for a hot dip galvanizing production line, characterized in that: include: Collection module: used to obtain the data of incoming parameters of hot-dip galvanized sheet, process parameters of leveling machine, temperature difference information and quality index of hot-dip galvanized sheet; Processing module: used for inputting the data into the self-learning preset model of the leveler in the self-learning preset method of the leveler applicable to the hot dip galvanizing production line as claimed in any one of claims 1 to 3 to obtain the preset value of the leveler, wherein the self-learning preset model of the leveler is obtained by neural network training, The neural network is a BP neural network. The processing module uses the data as training samples of the BP neural network, performs learning and training on the BP neural network, and obtains the self-learning preset model of the leveling machine. The BP neural network adopts the error back propagation algorithm for learning and training. BP neural network is a three-layer neural network structure, including input layer, middle layer and output layer; The hot dip galvanized sheet material parameters include: yield strength of raw steel coil σ s ,width W ,thickness h ; The process parameters of the leveling machine include: leveling speed v , Rolling length of the roller L , Original roughness of the roll R a ; The temperature difference information includes: annealing furnace temperature difference S1-S2 , Temperature difference between zinc pot and final cooling S3-S2 ; The quality indicators of hot-dip galvanized sheet include: elongation of hot-dip galvanized sheet ε , Finished product roughness R a .

5. The self-learning leveler presetting device suitable for a hot dip galvanizing production line according to claim 4, characterized in that: The input layer includes 10 units, the middle layer includes 18 units, and the output layer includes 3 units.

Images