JPH0688059B2 - Crown learning control method - Google Patents

Description

Detailed Description of the Invention [Industrial applications]

The present invention relates to a crown learning control method applied to crown setup control, and particularly when operating in a rolling command cycle in which specifications such as plate thickness, plate width, material, etc. of rolled material change significantly every few rolling schedules. The present invention relates to a crown learning control method that enables stable and highly accurate rolling by rolling.

[Prior art]

Learn roll profile based on crown error,
A technique for obtaining a stable rolling schedule is disclosed in, for example, JP-A-63-72426. Generally, in the computer crown setup control,
Rolling mill Work roll surface profile model (work roll profile model) is used to predict the work roll profile, and the already determined rolling conditions (rolling load, strip width, strip thickness, etc.) The plate crown predicted value on the entry side and the predicted value of the work roll profile are applied to the plate crown model, and in order to achieve the target crown, the set value of the crown control operation end such as the work roll bender or the intermediate roll shift is set. It is determined. At that time, in order to obtain a predetermined crown control accuracy, the work roll profile model and the plate crown model used in the crown setup control are
It is adjusted so that the error from the actual data is reduced by the lead rod pinching test method or aluminum plate pressing test method, or by using the plate profile meter set on the exit side of the finishing stand. However, at the stage of performing crown setup control, if there is a prediction error in the rolling state, the work roll profile, and the plate crown model itself, and the predicted plate crown on the exit side of the final stand that was predicted before the rolling mill was passed. A crown error will occur between the actually measured plate crown. Therefore, in order to reduce this error as much as possible, generally, the sheet crown model is learned while comparing the sheet crown calculation value obtained by calculation based on the rolling results after the sheet passing and the sheet crown actual measurement value. . A simple example of the learning method of the plate crown model will be described below. Now, regarding the Fi rolling machine, the plate crown on the exit side is Cr
i, Cri _-1 for the plate crown on the entry side, Ri for the work roll profile model, Y for the work roll elastic deformation model
i, strip width Wi Wi rolling load, bending load P _B i, HC
Let −δ be HCδi, the above-mentioned plate crown Cr on the output side
i can be expressed by the following equation (1). In this equation (1), Cr _OFS i is an error learning term (offset term). Cri = f {Cri ₋₁ , Yi (W, Pi, P _B i, HCδi), Ri} + Cr _OFS i (1) The offset term (Cr _OFS i) is obtained as follows. First, the actual measured value of the plate crown by the plate profile meter Cr
_The predicted crown error Cr _E , which is the difference between A and the calculated plate crown value Cr _C calculated from the rolling results after passing the rolling mill, is calculated by the following equation (2). Cr _E = Cr _A −Cr _C (2) Using the above crown error Cr _{E, according} to the following equation (3), the plate crown error Crj at the final standout side at the stage where the number of rolling coils is j from the roll change point is j Ask for. Crj = (1−G) · Crj ₋₁ + G · Cr _E (3) Here, G is an exponential smoothing gain. Now, for the sake of convenience, consider a case where the formula (1) is applied to a hot strip mill having seven stands. And Cr
_OFS1 = Cr _OFS2 = ... = Cr _OFS7 , and assuming that a plate crown on the inlet side of the Fi rolling mill ≈ Fi _{-1 on} the outlet side of the rolling mill is established, the error learning term is the following (4). Given by the formula. Cr _OFS i = Crj / g (4) Here, g is a distribution law for distributing the plate crown error on the exit side of the F7 rolling mill to each stand. When performing the crown learning control based on the expressions (1) to (4) described above, the error learning term Cr _OFS i of the expression (1) is updated every time one rolled material is rolled, and the roll change is performed. The subsequent predicted crown error Cr _E will be reflected in the crown setup control as a time constant determined by the exponential smoothing gain G.

[Problems to be Solved by the Invention]

However, the exponential smoothing gain G cannot be increased in order to suppress overcompensation of the crown learning control. Therefore, in the conventional crown learning control based on the equations (1) to (4), the strip crown error gradually changes for each rolled material like the strip crown error generated due to the prediction error of the work roll profile. However, if the error of the plate crown model itself due to the specifications of the rolled material (plate thickness, plate width, material, etc.) and the work roll profile prediction error are mixed, On the other hand, a sufficient change could not be exerted due to a large change in plate crown error for each rolled material. This is specifically shown in the diagrams shown in FIGS. 4 and 5. That is, FIG. 5 (a) shows plate thickness (mm) and plate width (mm) according to rolling, as shown in FIGS.
Shows the plate crown control accuracy when the above-described conventional crown learning control method is applied to a coffin cycle in which is gradually changed, and FIG. 7B shows the plate crown model accuracy at that time. In FIGS. 5A and 5B, the solid line indicated by the symbol A is the target crown at the time of computer setting, and the symbol B is
The broken line indicated by ∘ indicates the measured crown, and the solid line indicated by the reference C indicates the actual calculated value of the plate crown.
The horizontal axis is the number of rolling coils. In addition, FIGS. 6 (A) and 6 (B) show FIGS. 6 (C) and 6 (D).
As shown in FIG. 5 respectively, the results corresponding to FIGS. 5 (A) and 5 (B) above when the conventional crown learning control method is applied to the cycle in which the plate thickness and the plate width greatly change are shown. ing. From FIG. 5 described above, when the plate thickness and the plate width gradually change as the number of rolling coils increases, the control accuracy is 1σ≈, although the model accuracy is gradually decreased. 10 μm
And good results have been obtained. On the other hand, according to FIG. 6, when the plate thickness and the plate width change greatly, the control accuracy is 1
It is as low as σ≈26 μm, and it can be seen that the control error for each rolling coil cannot be reduced. The present invention has been made to solve the conventional problems, plate thickness, plate width, the error of the plate crown model itself due to the specifications of the rolled material such as material and the prediction error of the work roll profile An object of the present invention is to provide a crown learning control method capable of exhibiting a sufficient effect even when mixed.

According to the present invention, a first offset term and a second offset term are provided as error learning terms of a plate crown model equation applied to rolling crown learning control.
The offset term is set, the first offset term learns the plate crown error caused by the prediction error of the work roll profile, and the learned content is reflected in the crown setup control of the next rolled material, and the second offset term is The above object is achieved by storing a plate crown error caused by specifications of rolled material for each rolled material and reflecting the stored content in the crown setup control of rolled material of the same lot next time. .

[Action and effect]

In the present invention, from the first offset term, the plate crown that gradually changes mainly between the roll change and the next roll change mainly due to the prediction error of the work roll profile, as in the conventional crown learning control method. It becomes possible to learn the error and reflect it in the crown setup control of the rolled material next time, and due to the specifications of the rolled material such as plate thickness, plate width, material, etc. due to the second offset term above. The strip clan error, that is, the strip crown error from which the strip crown error caused by the above-mentioned prediction error of the work roll profile is removed from the strip crown error is stored for each rolled material, and is stored next time in the crown setup control for the strip rolled in the same lot. Can be reflected in. Therefore, for each rolled material, there are a learning term for the plate crown error caused by the work roll profile prediction error, and a learning term for the plate crown error caused by the specifications of the rolled material (plate thickness, plate width, material, etc.). It is possible to separate and control learning by schedule-free rolling, and the control accuracy can be improved even when performing the computer setup control of the crown for the rolling cycle in which the strip thickness, strip width and steel type change frequently. It is possible to maintain a stable and high state, and as a result, it is possible to stabilize the crown quality to high quality.

【Example】

Hereinafter, embodiments of the present invention will be described in detail with reference to the drawings. FIG. 1 is a schematic configuration diagram showing a main part of a rolling line applied to a crown learning control method of one embodiment according to the present invention, and FIG. 2 is a diagram for explaining an error learning term used in this embodiment. A diagram and FIG. 3 are explanatory views showing the outline of the method of this example. The present embodiment is a crown learning control method applied to a hot strip mill having seven stands, as in the case of the conventional crown learning control method described above. As shown in FIG. 1, this hot slip top mill allows control signals to be input from the learning control device 10 to each of the first to seventh stands, which will be described later with respect to the control device 10. By setting the first offset term and the second offset term, it is possible to appropriately control the rolling conditions of each stand. First, in carrying out the crown learning control method of the present embodiment, a plate crown model formula used for plate crown setup control is constructed as shown in the following formula (5). Cri = f {Cri _-1 , Yi (W, Pi, P _B i, HCδi), Ri} + Cr _OFS i + Cr _OFS i (5) where Cr _OFS i is the first offset term and Cr ' _OFS i Is the second
Each offset term is shown, and each is an error learning term. Definitions of other parameters and the like are the same as those in the above formula (1). The first offset term is synonymous with the error learning term given by the equation (4) applied to the conventional crown learning control, and is, so to speak, a term for learning the average change of the error. The second offset term is an error learning term introduced by the present invention, and the deviation of the actually measured plate crown from the conventional learning value is determined by factors other than the work roll profile prediction error, that is, the plate pressure. It is treated as a plate crown error that occurs due to the specifications of the rolled material such as plate width and material. FIG. 2 is a diagram corresponding to FIG. 5 (B) or FIG. 6 (B) described above, in which a symbol a corresponds to the first offset term, and a symbol b corresponds to the second offset term. Is equivalent to. In addition, in the drawing, a circle with a black dot indicates a measured value of a plate crown of a rolled material of the same steel type and the same size. Next, a method of calculating the second offset term will be described. First, the deviation ΔCr from the conventional learning value is obtained from the difference between the crown error Cr _E given by the equation (2) and the plate crown error Crj given by the equation (3) ((6) below).
formula). ΔCr = Cr _E −Crj (6) This deviation ΔCr is the prediction accuracy (1
Although the range of change differs depending on σ), the dead zone of whether or not the present embodiment is applied to the deviation ΔCr is set as follows in order to prevent overcompensation. If | ΔCr | <C (constant), Cr ′ _OFS i = 0. When | ΔCr | ≦ C (constant), ΔCr in the above equation (6) is distributed to each stand as the second offset term represented by the following equation (7). Cr ′ _OFS i = ΔCr / g (7) Here, g is synonymous with the case of the above formula (4), and is a distribution law of the plate crown error on the exit side of the F7 rolling mill. In this embodiment, as shown in FIG. 3, crown learning calculation is performed.
12 is executed and the second offset term given by the above formula (7) is stored for each rolled material, and the data file regarding the specifications of the rolled material and the second offset term for the past several dozen rolled materials. Create 14. On the other hand, the crown learning calculation 12 is executed according to the conventional crown learning control method, and the first offset term is stored in the finish learning table 16. At the stage of setting up the crown on the computer,
The data of the rolled material of the same lot as the target rolled material is found from the second offset term data file 14. If there is data on rolled material of the same lot, the first offset term related to the rolled material most recently rolled and the first offset term learned from the previous rolled material extracted from the finish learning table 16 are used. The above formula (5) is constructed by using the above formula, and this is referred to as a plate crown model formula. Using this plate crown model formula, a finish setting calculation 18 is executed, and the calculation result is used for the control device 10 at the time of crown setup control.
Set to (Calculator). At that time, the set value of the crown control operation end of the work roll bender or the like in each rolling mill can be calculated by replacing the Cri term with the target crown in the above formula (5). As described above in detail, according to the present embodiment, the crown learning control that has been effective only for the gradually changing plate crown error is, in the rolling environment of several tens of the latest rolling, Since the learning effect is exhibited for highly reproducible strip crown error caused by specifications (sheet thickness, strip width, material, etc.), the strip crown control accuracy during schedule-free rolling is stable and high. It is possible to maintain. Next, the effect of the present embodiment will be clarified with a concrete example. Fig. 4 shows that the work roll profile changes intricately due to work roll thermal expansion and friction.
Width reversal is 100 when rolled up to 100 rolled materials
~ 250mm rolling, is a graph plotting the crown error between the target crown and the measured value of the plate crown by the plate profile meter on the exit side of the finishing final stand rolling machine set in the calculator at that time, A) shows the result of applying the crown learning control method of the present embodiment, and FIG. 7B shows the result of applying the conventional crown learning control method. As shown in FIG. 4B, in the conventional learning control, the deviation of the error occurs depending on the plate width, and the control accuracy is 1
σ is as low as 32 μm. On the other hand, as shown in FIG. 4 (A), by applying the crown learning control method of the present embodiment, the deviation of the error due to the plate width is eliminated, and the control accuracy is significantly improved. I understand. Although the present invention has been specifically described in detail above, it is needless to say that the present invention is not limited to the above-mentioned embodiments and various modifications can be made without departing from the scope of the invention. For example, in the above-mentioned embodiment, the hot slit top mill having seven stands has been described, but the number of stands can be arbitrarily changed, and the present invention is not limited to the hot strip mill. In the crown learning control method of the present invention, for example, the number of stands is 1.
It can also be effectively applied to thick plate mills.

[Brief description of drawings]

FIG. 1 is a schematic configuration diagram showing a main part of a rolling line applied to a crown learning control method of an embodiment according to the present invention, and FIG. 2 is a line for explaining an error learning term used in the above embodiment. FIG. 3 is an explanatory view showing the outline of the above-mentioned embodiment method, FIG. 4 is a graph showing the effect of the above-mentioned embodiment, and FIGS. 5 and 6 are problems of the conventional crown learning control method. It is a diagram for explaining. A: Target crown when computer is set, B: Actually measured crown, C: Actual calculated value of plate crown, 10: Learning control device.

Claims (1)

[Claims] 1. A first offset term and a second offset term are set as error learning terms of a plate crown model formula applied to rolling crown learning control, and the first offset term is a plate caused by a work roll profile prediction error. The crown error is learned, and the learning content is reflected in the crown setup control of the rolled material next time. The second offset term stores the plate crown error caused by the specifications of the rolled material for each rolled material, and the memory is stored. A crown learning control method, characterized in that the contents are reflected next time in the crown setup control of rolled material of the same lot.