JPH0373365B2 - - Google Patents

Description

[Detailed Description of the Invention] [Field of Application of the Invention] The present invention relates to a control method for controlling the shape of a rolled material in a hot rolling mill, and particularly to a control method for controlling the shape of a rolled material by learning control of setup calculation when controlling the shape of a rolled material. Related to improving control accuracy.

[Background of the invention]

Generally, the quality of a rolled material is evaluated by its shape in the rolling direction (hereinafter abbreviated as shape), plate thickness, and plate width. Among these, the shape of the rolled material 3 is as shown in Fig. 7.
It is expressed by the steepness λ obtained from the waviness (a/l, c/l, b/l) at three points in the width direction (both ends and the center of the rolled material) using the following formula.

λ={(a/l+c/l)×1/2−b/l}×100(%
) It is desirable that the shape of the rolled material be flat (λ = 0). Therefore, in order to provide high-quality products, it is necessary to perform optimal control so that the shape of the rolled material is flat. .

As a control to ensure this prescribed quality,
Conventionally, DDC (Direct Digital Calculation) was used to set up the rolling line (setting the roll gap, rolling speed, work roll bending force, etc.), and control (including feedback control and operator manual intervention) during rolling was performed. control) and learning control that corrects setup calculations.

The control device starts the DDC after setting up the rolling mill in the rolling line, but if the calculation result of the setup calculation is not an appropriate value for flattening the shape, the DDC operates and uses feedback control. Although the shape is made flat, the quality of the product deteriorates until the shape is flattened, and the product yield decreases.

Therefore, in normal shape control, as shown in FIG. 8, the shape target value λ _ref at the time of setup is
The deviation (λ _ref - _λ r ) from the initial shape value λ _r (shape before the start of DDC) of the rolled material rolled on the rolling line set up using this target value λ _ref is smoothed and the average value is calculated. , this average value is the target value for the next rolled material. It performs so-called learning control.

This initial shape value λ _r is a value directly measured by a shape detector installed on the exit side of the rolling line, and the average value is given as = (1-β)′+β(λ _ref −λ _r ). I'm looking for it. Here, β: smoothing constant (close to 0 when noise is large, usually about 0.3),
λ′: Previous learning value.

However, only non-contact optical type shape detectors for hot-rolled materials can be used due to the following problems, and detection accuracy is poor. That is, although the hot rolled material is rolled at a high temperature of 900 to 1100° C., it is impossible to make measurements by bringing the detector into contact with the rolled material due to the thermal stress applied to the detector.

In addition, the method of detecting the shape of cold-rolled materials, which applies tension to the rolled material and vibrates it to detect the stress distribution, has a large influence on the width accuracy of hot-rolled materials due to the tension ( In other words, it is impossible because the rolled material is a non-magnetic material. For these reasons, currently only non-contact optical detectors can be used as shape detectors for hot-rolled materials.

The measurement principle of the optical shape detector is to shine light on three locations in the width direction of the rolled material shown in FIG. 7, and determine the shape of the rolled material from the movement of the light reflected by the rolled material. The operation of the shape detector will be explained in detail using FIG. Under the assumption that the shape of the rolled material is a sine wave (sine wave approximation), the shape detector converts it into waviness using the method described below.

Now, consider the case where the shape of the rolled material is as shown in Fig. 9, y=a/2sin2π/lx (1). In the shape detector, a signal is created by differentiating equation (1), and a signal is obtained as follows: y'=π _a /lcos2π/lx (2). This is the waveform shown in . This signal is full-wave rectified, and the sampling time t
The calculations are displayed by integrating the interval and calibrating the correspondence with the degree of waviness. (x=vt, v: speed of rolled material) F=∫ ^t ₀ |y'|dt=k・(a/l)・t...(3) where K: constant, output of shape detector The value is a/l=F/K·t...(4).

The output value of the shape detector is updated every sampling time t as shown in . The above is the detection principle.

Furthermore, considering the case where mechanical vibration Z(t) is applied to the rolled material, the integral value of equation (3) becomes F=K・(a/l)・t+Z(t)−Z(o), and Z An error of (t)-Z(o) occurs. Therefore, when an optical shape detector is subjected to mechanical vibration, it becomes extremely difficult to accurately determine the shape. Note that mechanical vibration refers to, for example, cracks in a rolled material due to rolling.

Due to the problems with the shape detector described above, the accuracy of detecting the initial shape λ _r of the rolled material deteriorates.
In hot rolling lines, there was a problem in that the effects of learning control were slow to appear, making it difficult to improve the accuracy of setup calculations.

To summarize the above, shape detectors that can be used in hot rolling are limited to non-contact type optical shape detectors because the rolled material is non-magnetic. However, the accuracy of the optical shape detector is low because it also recognizes the vibration of the rolled material as the shape. For this reason, there is a problem in that conventional learning control that learns the deviation between the target value of the shape and the output value of the detector is not effective and does not improve the setup accuracy.

In addition, as a shape control method in a 6-layer rolling mill, there is a method disclosed in Japanese Patent Application Laid-Open No. 128311/1983.

Also, JP-A-55-68110, JP-A-55-
No. 42188 and Japanese Patent Application Laid-Open No. 60-54217 disclose a technique for correcting the set-up calculation for the next rolled material using the actual rolling data of the previously rolled material. Correction calculations are performed using the detection values of the shape detector during rolling, and the detection accuracy of the shape detector as described above affects the correction calculations.

[Purpose of the invention]

In shape control in hot rolling, even if the precision of the shape detector itself has a limit and no improvement in precision can be expected, the present invention aims to improve the shape of the rolled material as much as possible within the limits of the precision of the shape detector. The purpose of the present invention is to provide a shape control method that can improve the shape quality of.

[Summary of the invention]

In order to achieve the above object, a shape control method according to the present invention provides a method for controlling a shape in a rolling equipment having a function of feedback controlling the shape of a rolling mill based on a signal from a shape detector installed on the exit side of the rolling mill. The shape of the rolled material being rolled is feedback-controlled based on the detection signal of the detector, and the shape control operation amount at the time when the rolling of the rolled material is completed and the shape influence coefficient on the rolled material due to the operation amount are used. The method is characterized in that the initial shape of the rolled material is calculated.

More specifically, when the absolute value of the difference in waviness between both ends of the rolled material |a/l-c/l| is less than a certain value, the steepness λ is calculated from the output of the shape detector, and the result is The work roll bending force of the final stand is controlled by feedback control recognized in the saturation type shown in FIG. Taking advantage of the fact that the shape of the rolled material is flat when this feedback control is completed, the influence coefficient on the shape of the rolled material calculated from the operation amount at the time of completion of the feedback control and the model formula used for setup is This is characterized by calculating the initial shape λ _r and performing learning control.

[Embodiments of the invention]

Next, a preferred embodiment of the shape control method according to the present invention will be described based on the drawings.

FIG. 1 shows a control device for implementing the shape control method according to the present invention and the relationship between the control device and a rolling line.

In FIG. 1, the rolling line includes a plurality of rolling mill stands (seven in this embodiment) S ₁ , S ₂ ,...
…Constructed by S ₇ . In this rolling line, the shape of the rolled material 3 is determined by the rolling load of each stand _S1 to _S7 , the roll shapes of the reinforcing roll 1 and the work roll 2, and the work roll bending force. Among these factors, the rolling load is determined by the power balance of each stand determined by the product plate thickness, and the roll shape is determined by the rolling schedule. For this purpose, the work roll bending force is determined under the condition that the shape of the rolled material 3 on the outlet side of each stand is flat, and the stands _S1 to _S7 are set up.

Now, as shown in FIG. 2, the control device 5 performs the following three types of processing. First, the rolling condition prediction value (p _o
^* , r _o ^* , b _o ^* ……) each stand S ₁ ~ _{S 7}
Optimum roll bending force of work roll 2 F ₁ ^*
A setup calculation (setup calculation unit 8) that determines ~ _F7 ^* , secondly a DDC process (DDC processing unit 6) that performs feedback control of the roll bending force based on the detection signal of the shape detector 4 during rolling; Thirdly, learning control (learning control section 7) is performed by calculating the steepness value for each rolled material at the end of rolling.

Here, FIG. 3 shows an example of the characteristics of the shape detector 4 installed on the exit side of the final stand _S7 . This characteristic was measured using four types of test materials and changing the moving speed of the test materials. The horizontal axis shows the actual shape of the test material, and the vertical axis shows the output value of the shape detector 4.

As shown in Figure 3, the actual shape and output value are 50%
There are varying degrees of severity. However, when the shape is flat, the output value of the shape detector 4 also becomes a value of 0±0.05%, and the error becomes small. Therefore, in DDC, the work roll bending force may be manipulated by performing feedback control so that the output value of the shape detector 4 becomes zero. That is, the detection signal of the shape detector 4 is used only for feedback control and not for learning control. If the shape is flat and the output value of the shape detector 4 is not zero but a specific value, the same result can be obtained by performing feedback control so that the output value of the shape detector 4 becomes the specific value. It is clear that the following effects can be obtained.

Next, the setup calculation performed by the control device 5,
The contents of DDC and learning control will be explained (Fig. 2,
reference). The setup calculation is performed by the setup calculation section 8 after receiving predicted values of rolling conditions (rolling load for each stand, plate thickness, etc.) from another control device. In the setup calculation section 8, a calculation process based on a model formula for setup calculation is performed under conditions where a flat rolled material can be obtained by using the results of learning control described later in addition to the rolling condition predicted values. The model formula is the outlet shape λ of the rolled material for each stand,
It is expressed as a function of rolling load p, work roll bending force F, rolling reduction r, plate width b, etc., and when the stand number is written as n, λ _o = f (p _o , F _o , r _o , b _o , λ _o-1 , ...) ...(5) There is a relationship.

Furthermore, the entry side shape λ _o-1 in equation (5) is determined in the 30th Joint Society of Plasticity Working Conference "Research on Crown and Shape Control Method of Hot Strips (5th Report)" for hot rolled materials. Around 10% as reported
It affects the exit shape λ _o only by a certain amount (∂λ _o /∂λ _o-1 0.1). Therefore, in a multiple tandem mill as shown in this embodiment, the influence of the shape λ ₀ of the rolling base material on the shape λ ₇ of the exit side of the final stand S ₇ is as follows: ∂λ ₇ /∂λ ₀ = ∂λ ₇ /∂λ ₆・∂λ ₆ /∂λ ₅ ...∂
λ ₁ /∂λ ₀ 0.1 ⁷ , which is a negligible value. Therefore, in this example, which targets hot tandem rolling, setup can be performed without considering the shape λ ₀ of the rolled base material, so learning control, which will be described later, can be performed between bars. .

Optimal working roll bending force for each stand
F _o is the predicted value of rolling load p _o , rolling reduction ratio r _o , plate width b _o, etc. (p _o ^* , r _o ^* , b _o ^* ...) and the shape correction value of the rolled material which is the result of learning control. By λ, F ₇ = f ^-1 (P ₇ *, λ _ref +, r ₇ *, b ₇ *, λ ₆ *…) F ⁶ = f ^-1 (P ₆ *, λ ₆ , r ₆ *, b ₆ *, λ ₅ *, ...) 〓 F ₁ = f ^-1 (P ₁ *, λ ₁ , r ₁ *, b ₁ *, λ ₀ *, ...) |λn * | <3 Fmin≦Fn≦Fmax ( (Determined by machine specifications) ...(6) The optimal work roll bending force is determined by solving the equation. Here, F _nio is the minimum value of the work roll bending force, F _nax is the maximum value of the work roll bending force, and λ ₀ =0. Further, the cross mark attached to each value represents a predicted value (setup).

After the above setup is completed, rolling starts.
DDC processing is performed from the time when the rolled material 3 passes directly under the shape detector 4 until it passes through the final stand.

Next, DDC processing is performed by operating the working roll bending force F ₇ of the final stand S ₇ ,
Feedback control is performed so that the shape signal from the shape detector 4 becomes zero. The roll bending force is controlled by expanding and contracting hydraulic jacks 14 and the like provided between the respective bearing boxes of the upper and lower work rolls 2. However, if a one-sided elongation phenomenon occurs due to the condition of the rolled material 3, the feedback control is stopped. The one-side elongation phenomenon refers to a case where the difference (a/l-c/l) in the degree of waviness between both ends of the rolled material is large. If you perform feedback control when one side extension phenomenon occurs,
As shown in Figure 6, (unilateral elongation) → (compound elongation) →
A hunting phenomenon called (one-sided elongation) occurs. The difference in the degree of waviness that causes this hunting phenomenon is a value that is determined economically based on the conditions of the rolled material (material quality, sheet width, sheet thickness) and the characteristics of the shape detector.

In FIG. 2, the difference in the degree of waviness that is determined to be a one-sided elongation phenomenon is indicated as λ _E. If no one-sided stretching phenomenon has occurred, feedback control is performed to manipulate the work roll bending force _F7 of the final stand _S7 .

In this feedback control, the steepness λ is calculated from the undulation obtained from the shape detector, and the result is recognized in a saturation type as shown in FIG. 4 to perform the feedback control. The value of h in FIG. 4 is determined by taking into consideration the noise width e output by the shape detector 4 and the stability of the feedback control system under the condition that the rolled material 3 is flat and no tension is applied. It is. Here, the noise width e refers to the range of values indicated by the detector output when the actual shape is zero in FIG. The value of d is determined in consideration of the stability of the feedback control system, and in this embodiment, it is set to the maximum value within the range in which no overshoot occurs.

The above feedback control system can be expressed by the block diagram shown in FIG. The part surrounded by the dotted line in Fig. 5 is the DDC processing section 6 in Fig. 2.
This is the content that will be executed. However, the target value λ _ref is fixed to zero for the purpose of flattening the rolled material. Block 9 is a diagram showing a method for recognizing the output value of the shape detector shown in FIG.
indicates an integrator. (S is a Laplace exchange operator.) Block 11 is a transfer function that represents the response of the work roll bending force of the final stand, and block 12 is a transfer function that represents the response of the work roll bending force and the shape of the rolled material. This is the transfer function that represents
Block 13 represents dead time τ until the shape of the rolled material directly below the work roll of the final stand is input to the control device 5.

In order to make this feedback control system stable, it is necessary to satisfy the conditions expressed by the following equation.

∂λ ₇ /∂F ₇・d・k≦h (1≦k≦3) ...(7) and e≦h By determining the value of h to satisfy this equation (7), the shape detector 4 It is possible to obtain a feedback control system that corrects the rolled material 3 to be flat regardless of its characteristics. However, the most desirable state in shape control is that the initial shape of the rolled material rolled under the optimal work roll bending force Fn determined by setup calculation is flat, and the DDC manipulated variable ΔF ₇ becomes zero. It is something. In order to achieve this state, the following learning control is performed to improve the accuracy of setup calculations.

Learning control is performed by the learning control unit 7, and the rolling material is controlled by taking in the actual rolling value during rolling (same content as the predicted rolling condition value described above) and the DDC operation amount after the rolled material 3 becomes flat. Find the initial shape and correct the setup calculations. Obtaining the initial shape of the rolled material 3 with the shape detector 4 is because, in addition to the influence of mechanical vibration of the rolled material 3, the effects of upward feeding, downward feeding, and meandering of the rolled material are also added, which are phenomena peculiar to the early stage of rolling. It becomes even more difficult. Therefore, since the initial shape cannot be determined from the shape signal of the shape detector 4, the initial shape is determined by the following method.

In other words, the DDC when the rolled material becomes flat
The influence coefficient of the manipulated variable ΔF ₇ and the work roll bending force F ₇ used in the setup calculation on the shape λ ₇ of the rolled material ∂λ ₇ / ∂F ₇ = ∂f (p ₇ , F ₇ , r ₇ , b ₇ ……)／∂F ₇ ……(8
), calculate the initial shape λ _r of the rolled material using the following formula. The amount of change in shape Δλ of the rolled material due to ΔF ₇ is Δλ=∂λ ₇ /∂F ₇ ·ΔF ₇ (9). Since the shape of the rolled material becomes flat as a result of adding the manipulated variable ΔF ₇ when the value of the initial shape is λ _r , λ _r + ∂λ ₇ / ∂F ₇ · ΔF ₇ = 0...(10) The following relationship is obtained. Note that this value of 0 means flatness. Therefore, the initial shape λ _r is obtained from equation (10) as λ _r =−∂λ ₇ /∂F ₇ ·ΔF ₇ (11). By doing this,
The initial shape, which could not be determined from the shape detector 4, can now be easily determined from the manipulated variable ΔF ₇ of the DDC and the influence coefficient ∂λ ₇ /∂F ₇ on the shape of the plate.

On the other hand, since there is always a difference between the predicted rolling conditions used in the setup calculation and the actual rolling value, in order to correct only the setup calculation, the model equation for the exit side of the final stand is calculated using equation (5) using the actual rolling value. Recalculate the initial shape λ _c based on . This calculation result λ _c
is the predicted value of the shape of the rolled material according to the model formula, which corresponds to λ _ref in the conventional method, and λ ₁ = f(p ₁ , F ₁ , r ₁ , b ₁ , λ ₀ , ...) λ ₇ = f(p ₇ , F ₇ −ΔF ₇ , r ₇ , b ₇ , λ ₆ , ...) ...(12) It is given by the calculation.

Here, p _o , F _o , r _o , b _o , . . . are actual rolling values at stand No.n. The difference λ c −λ _r between the initial shape λ _r obtained by equation (11) and the initial shape _{λ c} obtained by recalculating using the model equation from the rolling actual values is exponentially smoothed for each rolled material as follows. By using this value as a correction value for the shape target value in the setup calculation of the rolled material, it became possible to obtain a rolled material with a good initial shape.

〔Effect of the invention〕

As described above, according to the shape control method according to the present invention, the detected output value of the shape detector is used for normal feedback control, and the detected output value of the shape detector is not used for learning control. Learning control is performed by correcting the set-up calculation value using the shape control operation amount at the end of rolling and the shape influence coefficient on the rolled material due to this operation amount, so the learning control is performed as much as possible within the accuracy limit of the shape detector. The shape quality of the rolled material can be improved (specifically, about three times as much as the conventional one).

[Brief explanation of drawings]

Fig. 1 is a block diagram showing a control device for implementing the shape control method of the present invention, Fig. 2 is a flowchart showing each process of the control method of the present invention, and Fig. 3 is a block diagram showing the control device for implementing the shape control method of the present invention. FIG. 4 is an explanatory diagram showing the relationship between the detector output and the recognized value. FIG. 5 is an explanatory diagram showing the feedback loop in the present invention. FIG. FIG. 7 is an explanatory diagram of the shape of the rolled material, FIG. 8 is an explanatory diagram of conventional shape detection signal processing, and FIG. 9 is an explanatory diagram of the principle of the shape detector. 1...Reinforcement roll, 2...Work roll, 3...
Rolled material, 4...Shape detector, 5...Control device, 6
...DDC processing unit, 7...Learning control unit, 8...Setup calculation unit, 9, 10, 11, 12, 13
……Block, F _o ……Working roll bending force of No.n stand.

Claims (1)

[Claims] 1. In rolling equipment having a function of feedback controlling the shape of the rolling mill based on the detection signal of a shape detector installed on the exit side of the rolling mill, Feedback controls the shape of the rolled material during rolling, and calculates the initial shape of the rolled material based on the shape control operation amount at the end of rolling of the rolled material and the shape influence coefficient on the rolled material due to the operation amount. A method for controlling the shape of a rolled material, characterized in that: 2. A method for controlling the shape of a rolled material, characterized in that the initial setting of rolling conditions for the next rolled material is corrected using the shape of the rolled material determined in claim 1.