RU2785510C2 - Flatness adjustment with optimizer - Google Patents

Abstract

FIELD: rolling production.

SUBSTANCE: invention relates to the field of rolling production; it can be used for flatness adjustment of a metal band rolled in a rolling cage. The method includes determination, using a control device, of a set of controlling values for flatness adjustment elements and control of these elements, while, by means of the control device, the first optimizer is implemented, using which, according to the set ratio, actual adjustment values are set, and a combination of flatness values is determined, and then, according to the set minimization algorithm, actual adjustment values are determined, based on which change vales are determined for controlling values for flatness adjustment elements, using a flatness adjuster implemented in the control device.

EFFECT: use of the invention allows for an increase in the quality of flatness adjustment and, respectively, the quality of a rolled band.

11 cl, 8 dwg

Description

FIELD OF TECHNOLOGY

The present invention originates from a method for operating a rolling stand in which a metal strip is rolled,

wherein the control device for the rolling stand determines with an operating cycle in each case a series of control variables for the corresponding number of flatness control elements of the rolling stand and controls the flatness control elements in accordance with the detected control variables.

The present invention further includes a computer program including a machine code that can be executed by a control device for a rolling stand for rolling a metal strip, wherein the execution of the machine code by the control device causes the control device to operate the rolling stand according to such an operation method.

The present invention also relates to a control device for a metal strip rolling stand, the control device being in the form of a software programmable control device and programmed with a similar computer program so that it operates the rolling stand according to a similar operation method.

The present invention further relates to a rolling stand for rolling a metal strip,

moreover, the rolling stand has several flatness adjustment elements, with the help of which the flatness of the metal strip emerging from the rolling stand can be influenced,

wherein the rolling stand has a similar control device by means of which the flatness adjusting elements of the rolling stand are controlled according to a similar operating method.

BACKGROUND OF THE INVENTION

The above objects are known, for example, from US 2016/0 052 032 A1. In US 2016/0 052 032 A1, a singular value decomposition (singular value decomposition) of the so-called cage matrix is carried out. Based on the decomposition of the singular value, combinations of control variables for the flatness control elements are determined in the rolling stand eigenvalue system, from which the control variables for the flatness control elements are then determined.

From WO 2006/132 585 A1 a method for optimizing the flatness control during strip rolling is known, wherein the control of any number of actuating elements is optimized. The flatness error is determined - that is, the deviation between the actual and calculated. This flatness error is subject to singular value dissection. As part of the dismemberment of the singular value, limitations are taken into account, namely the limitations of the final elements. The scope of the disclosure of WO 2006/132 585 A1 thus corresponds essentially to the scope of the said US application.

From WO 95/19591 A1 a process is known, in particular for rolling a strip and adjusting its flatness. In WO 95/19591 A1, it is essentially a matter of determining the effectiveness of the individual control elements and, if necessary, monitoring them. Further, this document discloses the usual adjustment based on a comparison of calculated and actual.

From US 6 098 060 A a process control method is known in which control variables are determined for a plurality of actuating elements that influence the process. Flatness control for a metal strip is given as an example of such a process. US 6 098 060 A should call for optimal process control.

The flatness of the metal strip is usually referred to as the local or global difference in the lengths of the metal strip as a function of the strip width. As part of the rolling process, the design flatness is set, which the metal strip should have, if possible, after rolling.

The flatness of the metal strip is influenced by various effects. Some of these effects occur during the rolling of the metal strip and cannot be influenced. An example of such an effect is the change in the rolling force during the rolling of the metal strip. Other of these effects can be influenced by appropriate control of the roll stand adjustment elements - hereinafter referred to as the flatness adjustment elements. For example, both the reverse bending of the rolls and - subject to appropriate wear of the rolls - the axial displacement of the rolls have an effect on the flatness. The same applies to the location-dependent effect on the temperature of the work rolls across the width of the metal strip or the work rolls, for example cooling.

The effect of such flatness adjustment elements on flatness can be considered within the scope of the flatness adjustment to be substantially proportional to the control of each flatness adjustment element. Thus, the change in flatness can be described by changing the control variable of the corresponding flatness adjusting element and by the effectiveness of the corresponding flatness adjusting element.

The purpose of each flatness adjustment is to keep the flatness of the metal strip as constant as possible along the entire length of the strip and thus bring it as close as possible to the design flatness. To this end, within the framework of the flatness adjustment, the attempt is made to compensate for the measured deviation from the calculated flatness by means of the flatness adjustment elements. To this end, the settings of the flatness adjusting elements are assigned in such a way that the sum of the resulting flatness changes counteracts the deviation of the controlled variable and compensates for it as best as possible. In this case, one speaks of the optimal operating point.

In the prior art, the assignment of the efficiencies of the flatness adjustment elements is often done manually and in advance. In particular, during the active rolling of the metal strip, a slight additional control of the corresponding flatness adjusting element takes place, and the resulting action is recorded. The resulting efficiency is retained in the control device for the rolling stand. This procedure is error-prone, time-consuming, and imprecise. In addition, changes in efficiency may be recorded - if at all - only with a significant time delay. The time delay is often in the range of months.

Further, when using several flatness adjustment elements, finding the optimal operating point is difficult. And although the referenced document US 2016/0 052 032 A1 represents a certain step forward, nevertheless, it can still be improved. In particular, within the framework of this method, only all the flatness adjusting elements can be optimized at the same time, and not individually or sequentially one after the other. Further, with this approach, the boundary conditions to which the flatness adjusting elements are subject cannot be taken into account in the definition of the control variables. However, subsequent accounting leads, as a rule, to the fact that the optimum is no longer realized. Operating the rolling stand outside the optimum operating point often leads to increased wear and increased abrasion of the flatness adjusting elements or other elements of the rolling stand.

Further, the flatness adjusting elements often cannot be controlled independently of each other. In particular, it may be that the control of one flatness adjustment element and the control of another flatness adjustment element mutually degrade each other, cancel each other out or interfere with each other. Finding the optimal operating point under these conditions is very difficult.

As a rule, the measured flatness values are then recorded by means of a suitable measuring device located behind the rolling stand. The measuring device can be made, for example, in the form of a segmented measuring roller. Regulation of violations is carried out, as a rule, with the help of PI controllers. They are laboriously parameterized during commissioning of the rolling stand. Due to the distance to the measuring device and the relatively slow response of the flatness adjustment elements, the adjustment is very inert. As a result, losses in quality must be tolerated, since violations can only be controlled slowly.

From US 2006/0 282 177 A1 it is known to control a technical system using the interior point method. In this case, the corresponding optimizer solves the optimization problem, which includes the necessary states of the system, control variables, and linear boundary conditions. US 2006/0 282 177 A1 refers to a gas turbine as an example of use for a technical system.

From the industry article “Real-time Dynamic Optimization of Nonlinear Systems: A Flatness-based Approach” by M. Guay, Proceedings of the 44th IEEE Conference on Decision and Control, and the European Control Conference 2005, pages 5842-5847, it is known to use the internal points for real-time control. Examples of use are generally batch processes and specifically the bioreactor.

From the text of the lectures “C21 Model Predictive Control” by Mark Cannon, University auf Oxford, 2016, a model predictive controller is known. As an example of use for the regulator, hot rolling is mentioned in particular, with residual stress being mentioned in particular as target values. As a possible solution, the method of the interior point is indicated.

CN 104 698 842 A indicates an interior point method for solving a non-linear optimization problem. It is noted that, in particular, the upper and lower limits of the control variables and the change in the control variables are taken into account as boundary conditions. The cost function seems to include only the target values and the change in the control values.

SUMMARY OF THE INVENTION

The object of the present invention is to provide possibilities by which excellent flatness control is realized.

The problem is solved using the method of operation with the features of paragraph 1 of the claims. Preferred embodiments of the method of operation are the subject of dependent claims 2-14 of the claims.

According to the invention, the method of operating a rolling stand of the type indicated at the beginning is carried out due to the fact that

- that the controller implements the first optimizer, which first temporarily sets the actual correction values and according to the relationship:

или

or

defines a set of flatness values, where

-- f is a collection of flatness values,

-- s is the set of actual correction values,

-- f0 are the initial flatness values,

-- W is the performance matrix, and

-- s` is a set of corrective values determined in the previous working cycle,

and then, by changing the actual correction values, minimizes the ratio

(1),

(one),

where

-- f* is the set of calculated flatness values,

-- s0 is the set of target values for the correction values, and

-- α and β are weight coefficients,

- that the first optimizer takes into account linear, preferably exclusively linear, constraints when determining the actual correction values,

- that the linear constraints have the form

or view

и

,

and

,

where C is a matrix, B is a constrained vector with the actual correction values observed, and c is a constrained vector respected by the difference of the actual correction values relative to the correction values of the previous operating cycle, and

- that the control device determines the control values for the flatness adjustment elements, taking into account the changed actual correction values.

Flatness adjustment elements can be assigned as needed. For example, in a normal rolling stand (particularly a quarto stand or a sixth stand), the flatness control elements may be roll offset, roll backbend, and segmented temperature effects. If necessary, an inclined position of the rolls, in particular the work rolls, can also be additionally implemented. With a cluster rolling stand, such as a 12-high rolling stand or a 20-high rolling stand, other flatness adjustment elements and more flatness adjustment elements can also be implemented.

In relation 1, the first term means the error of the actual and calculated, that is, the deviation of the calculated flatness from the calculated flatness. The second term causes the minimum deviation of the modulation of the flatness control elements from the target values. Target values can be assigned, for example, in such a way that the corresponding flatness adjusting elements are loaded as low as possible. The third term causes a minimal change in the modulation of the flatness adjustment elements. The weight coefficients are non-negative. In most cases, at least one of the weights is greater than 0. Often, both weights are greater than 0. A value of 0 for both weights may be appropriate in a particular case, for example, if only one single flatness adjustment element is available. In this case, the norm for the first term, ie for the actual and calculated error, can also be implemented, for example by simply generating an absolute value.

The rate used can be assigned to each member individually. As a rule, in each case we are talking about the “normal” Euclidean norm, that is, the root of the sum of the squares of the individual terms.

The set of initial flatness values corresponds to the initial flatness vector. Similarly, the set of calculated flatness values corresponds to the calculated flatness vector. Spatial bandwidth resolution can be as needed. It is usually found at more than 10 anchor points, often above 50 anchor points. In some cases, up to 100 reference points are realized, in rare cases even more.

Due to the way the optimization problem is posed, the first optimizer thus solves a nonlinear optimization problem with linear constraints. In particular, due to this circumstance, the first optimizer can work in real time. In addition, due to the simultaneous consideration of constraints during optimization, the optimal solution can be determined in a simple manner.

In the first basic embodiment of the present invention,

— that the measured flatness values are recorded with a spatial resolution over the width of the metal strip by means of a measuring device,

— that the measured flatness values and the corresponding calculated flatness values are fed to the control device, and

- that the measured flatness values are fed to the control device as initial flatness values.

The set of measured flatness values corresponds, similarly to the above, to the actual flatness vector.

In the first basic embodiment, the control device implements, as a rule, not only the first optimizer, but also the flatness controller located behind the first optimizer, to which the actual correction values detected by the first optimizer are applied, and which of the actual correction values determines the change values for the control variables for the elements flatness adjustment. In the preferred embodiment of the present invention, the flatness controller is designed as an observer-based controller. The flatness regulator is designed, therefore, in such a way that it

- forms the sum of the actual correction values weighted by the amplification factor and the output signal of the model of the section of the rolling stand,

- determines on the basis of the amount formed in this way a preliminary signal,

- determines, by differentiation of the pre-signal, the change values for the control variables for the flatness adjusting elements, and

- gives a preliminary signal to the model of the section of the rolling stand as an input signal.

The model of the section of the rolling stand can in this case have different partial models for different flatness control elements. In the various partial models, in particular, the respective dynamics of each flatness control element can be taken into account individually.

In the simplest case, the preliminary signal is identical to the formed sum. Preferably, however, the flatness controller determines the pre-signal by filtering the generated sum in the filter. In an embodiment of this procedure, it is further possible that the controller parameterizes the filter dynamically.

In some cases, it is expedient for the controller to implement - in addition to the first optimizer - a second optimizer, which is similar to the first optimizer, that is, it determines the actual correction values in the same way as the first optimizer. In this case, however, it is

- that the set of flatness values for the second optimizer is determined on the basis of the set of nominal flatness values and the actual correction values valid for the second optimizer,

- that the nominal values of flatness correspond to the nominal change in rolling force,

- that the weights for the second optimizer are 0, and

- that the control device determines the control values for the flatness adjustment elements, additionally taking into account the actual change in the rolling force, the nominal change in the rolling force and the actual correction values determined by the second optimizer.

The second optimizer thus determines its actual correction values independently of the first optimizer. The result determined by the second optimizer - in particular the quotient of the corresponding current correction value and the nominal change in rolling force - corresponds to the effectiveness of the corresponding flatness control element. When multiplied by the change in rolling force, it is thus possible to directly determine the corresponding control variable. This procedure has the advantage that it works very quickly, since it is not necessary to wait until in each case the rolled section of the rolled material reaches the measuring device.

In a particular case, it is also possible that although there is only the first optimizer, the first optimizer nevertheless works in the way that has just been explained for the second optimizer. And although this course of action represents pure management. However, it has the advantage that it reacts quickly and can be used in any rolling stand, whether or not a measurement device for recording the measured flatness values is located downstream of the corresponding rolling stand.

Preferably, the first optimizer changes the actual correction values in several iterations. As a result, the optimum of the actual correction values can be determined best.

Preferably, the first optimizer finishes changing the actual correction values as soon as

- it has completed the specified number of iterations, and/or

- he changed the actual correction values within a given time, and/or

- the actual correction values change only slightly from iteration to iteration, and/or

- ratio

changes only slightly, and/or

- another termination criterion is met.

As a result, the possibility of working online and thus the possibility of working in real time can be maintained in particular also if a change in the current correction values only leads to a slow convergence (convergence).

It is possible that the first optimizer takes into account the restrictions not at each iteration, but only at the final result. Preferably, however, the first optimizer considers the constraints at each iteration.

Preferably, the first optimizer determines the actual correction values according to the interior points method. Such methods are reliable, converge very quickly, and require only a relatively small area of memory in the first place. As a result, it is possible to carry out online optimization within the framework of the flatness control.

In a preferred embodiment, the control device automatically determines the efficiency matrix based on the rolling stand models. As a result, the effectiveness of the flatness adjusting elements can be determined quickly and reliably. In particular, tests should not be carried out in which the metal strip is already being rolled. Examples of suitable rolling stand models are a rolling force model, a bending model, a collapse model, a roll gap model, a model for modeling thermal and wear-induced roll barreling, and many others.

The efficiency matrix is determined at least at the time of commissioning the rolling stand. Preferably, however, the control device dynamically determines the efficiency matrix anew in each case just before the start of rolling of the respective metal strip. That is, preferably, for each rolled metal strip, the models are again called and evaluated before rolling the corresponding metal strip, and based on the evaluation, an efficiency matrix is determined.

It is even better if the control device dynamically monitors the efficiency matrix even during the rolling of the respective metal strip. For example, the control device can call up and evaluate the models again even during the rolling of the respective metal strip. It is also possible that the performance matrix is tracked by evaluating the expected flatness values identified by the models and the measured flatness values. Such an assessment can be carried out, for example, using a neural network.

The problem is further solved using a computer program with the features of paragraph 15 of the claims. According to the invention, the execution of the computer program causes the control device to operate the rolling stand according to the operating method according to the invention.

The problem is further solved by means of a control device for a rolling stand for rolling a metal strip with the features of claim 16. According to the invention, the control device is programmed with the computer program according to the invention, so that it operates the rolling stand according to the operating method according to the invention.

The problem is further solved by using a rolling stand for rolling a metal strip with the characteristics of paragraph 17 of the claims. According to the invention, the rolling stand has a control device according to the invention, by means of which the flatness adjustment elements of the rolling stand are controlled according to the operating method according to the invention.

BRIEF DESCRIPTION OF THE DRAWING

The above described features, features and advantages of this invention, as well as the way to achieve them, become clearer and more understandable in connection with the following description of exemplary embodiments, which are explained in more detail in conjunction with the drawing. In this case, the drawing on the schematic images shows:

fig. 1 - rolling stand for rolling a metal strip on the side;

fig. 2 - rolling stand from Fig. 1 top;

fig. 3 - rolling stand from FIG. 1 when viewed in the direction of transport of the metal strip;

fig. 4 - measuring device from below;

fig. 5 - internal structure of the control device;

fig. 6 is a block diagram of the method;

fig. 7 is a modification of the internal structure of FIG. 5; and

fig. 8 is an alternative embodiment of the internal structure of FIG. 5.

DESCRIPTION OF EMBODIMENTS

According to FIG. 1 to 3, the rolling stand has several rolls 2, 3 for rolling the metal strip 1. As a rule, the rolls 2, 3 include, in addition to the work rolls 2, back-up rolls 3. Often there are no further rolls. In this case, we are talking about a rolling stand, a quarto stand. In some cases there are also further rolls, for example, at the sixth stand, between the two work rolls 2 and both back-up rolls 3, an intermediate roll is provided in each case. Other designs are also known, for example in the form of a 12-roll rolling stand or a 20-roll rolling stand.

The rolling stand is controlled by a control device 4 . The control device 4 is generally designed as a software-programmable control device. This is indicated in FIG. 1 marked “µP” (for “microprocessor”) inside the control device 4. The control device 4 is programmed with a computer program 5. The computer program 5 includes a machine code 6 which can be executed by the control device 4. The execution of the machine code 6 by the control device 4 causes the control device 4 to operate the rolling stand according to the operating method according to the invention.

Next, first with additional reference to FIG. 4 and 5 explain the first main embodiment of the present invention in more detail.

Within the framework of the first basic embodiment, behind the rolling stand, according to FIG. 1 is a measurement device 7 with which measured flatness values fM are recorded during operation of the rolling stand. The measured flatness values fM are recorded with a spatial resolution along the width b of the metal strip 1. For example, the measuring device 7 can be made according to the image in FIG. 4 in the form of a segmented measuring roller, which is located on the exit side of the rolling stand. Such segmented measuring rollers are generally known to those skilled in the art. By recording the measured flatness values fM with a spatial resolution across the width b of the metal strip 1, the measured flatness values fM are not a scalar but a vector. fM thus designates a set of measured flatness values fM, and not just one single value recorded at a specific location when viewed across the width of the metal strip 1 . Spatial resolution can be as needed. The number of reference points for which one single measured flatness value fM is recorded in each case is often in the upper two-digit range.

The measured flatness values fM are supplied according to FIG. 5 to the control device 4 as initial flatness values f0. In addition, the corresponding calculated flatness values f* are supplied to the control device 4 . Also, when talking about design values f* of flatness, it is therefore not about one single design value, but about the totality of design values f* of flatness, that is, about the corresponding vector.

The control device 4 determines with an operating cycle T in each case a series of control variables S for the corresponding number of flatness adjusting elements 8 and controls them in accordance with the detected control variables S. With each operating cycle T, the control variables S are thus determined anew. They then remain valid until the next determination of the control variables S. The operating cycle T is in most cases in the range of less than 1 second, for example between 0.2 seconds and 0.5 seconds.

By means of the flatness adjustment elements 8, the flatness of the metal strip 1 emerging from the rolling stand is influenced. intermediate rolls) are displaced in the axial direction according to the axial displacement A. Alternatively or additionally, a segmented influence on the temperature can be carried out using a suitable device. For example, a corresponding local cooling of the K work rolls 2 can be carried out by means of a cooling device. Other flatness adjusting elements 8 are also possible.

Fig. 5 shows, for the first main embodiment of the present invention, the internal structure of the control device 4. However, shown in FIG. 5 blocks are available, as a rule, not in the form of hardware, but in the form of software modules. They are implemented in this way by executing the machine code 6 of the computer program 5.

According to FIG. 5, the controller 4 implements the optimizer 9. The optimizer 9 is hereinafter referred to as the first optimizer. The first optimizer 9 is within the execution of FIG. 5 single optimizer.

The optimizer according to the present invention is - generally speaking - a computing unit to which some input values are applied. The computing unit then determines the objective function, which includes the input values and the output values set by the computing unit. The calculation unit then changes the output values in order to optimize the objective function. Usually, the computing unit performs several iterations for this, and in each iteration, in each case, based on the input values and the output values \u200b\u200bset at the end, in each case determines the objective function and, based on the identified objective function, changes the output values \u200b\u200bin order to optimize the target functions. Such optimizers are generally known to those skilled in the art. Purely as an example, reference is made to optimizers that work according to the following methods:

- Continuous optimization methods such as simplex method, interior point method, confidence region method, cubic regularization method, SLP method, and Gauss-Newton type method such as SQP method. These methods may be linear or non-linear as appropriate.

- Discrete optimization methods such as cutting plane method, branch and boundary method, network optimization method, etc.

- Methods of mixed-integer optimization, for example in the form of a combination of continuous and discrete methods.

- Heuristic and metaheuristic optimization methods, such as genetic methods, evolutionary methods, ant optimization methods, swarm optimization methods, simulated annealing methods, and taboo search methods.

- Genetic ways of optimization.

If necessary, the above optimization methods can be combined with execution in a neural network.

The first optimizer 9 is supplied with initial flatness values f0 and calculated flatness values f*. The first optimizer 9 determines the correction values s, namely for each flatness adjustment element 8 in each case one eigenvalue. The correction values s are valid for the current operating cycle T and are therefore referred to in the following as the current correction values s. Like the initial flatness values f0 and the calculated flatness values f*, the reference numeral s thus also denotes the set of actual correction values. Also here we are talking in this way - at least as a rule - not about a scalar, but about a vector. However, in some cases it is possible that there is only one single flatness adjusting element 8 . In this case, the vector s degenerates into a scalar. The meaning of the actual correction values s will become clear from the following explanations.

The current correction values s are fed back to the optimizer 9. However, they are previously delayed in the delay block 10 by one (1) operating cycle T. The correction values s' applied to the first optimizer 9 in a certain operating cycle T thus correspond to the correction values of the previous operating cycle T. Therefore, they are referred to in the following as delayed correction values and are provided with the reference position s`.

The first optimizer 9 determines the actual correction values s by minimizing the ratio:

(5)

(5)

The first optimizer 9 thus modifies the actual correction values s until it minimizes this ratio. In other words: the first optimizer 9 first assigns the current correction values s as provisional values. Using the previously assigned values for the correction values s, the first optimizer 9 then minimizes the higher ratio by changing the actual correction values s. The corrective values s valid for the respective operating cycle T are in this case the last detected or last changed current corrective values s.

In the above ratio, f is the set of flatness values, that is, again a vector. The flatness values f are determined by the first optimizer 9 on the basis of the initial flatness values f0 and the actual correction values s. For example, the first optimizer 9 may determine the flatness f values according to the relationship:

(6)

(6)

W is the performance matrix. It individually indicates which effect a certain individual correction value s has on which of the flatness values f.

s0 is the set of target values for the correction values s. The target values s0 can be assigned, for example, in such a way that the respective flatness adjusting elements 8 are loaded as little as possible, for example controlled as little as possible. The target values s0 can be hard-coded for the control device 4 . Alternatively, they can be given to the control device 4 as variables or parameters.

α and β are weight coefficients. In any case, they have a non-negative value. They are usually greater than 0. They may be hard-coded into the first optimizer 9, or they may be parameterized.

The first optimizer 9 takes the restrictions into account in determining the actual correction values s. Constraints include linear constraints. Preferably, the constraints include only linear constraints at all.

In particular, the first optimizer 9 takes into account in any case linear constraints in the form:

(7)

(7)

Here C is a matrix. B is a vector with the actual correction values s respected by the constraints. Additionally, the first optimizer 9 can take into account further linear constraints in the form:

(8)

(eight)

Wherein c is a vector with constraints respected by the difference between the actual correction values s relative to the delayed correction values s`.

Suitable optimizers are known per se to those skilled in the art. The implementation of the first optimizer 9 can be therefore as needed. Preferably, the first optimizer 9 determines the correction values s according to the interior point method.

The actual correction values s detected by the first optimizer 9 - i.e. the actual correction values s after changing the actual correction values s - are the basis from which the control device 4 determines the control values S for the flatness adjusting elements 8 .

Typically, the first optimizer 9 changes the actual correction values s in several iterations. He thus tries to gradually determine all the best actual correction values s. The first optimizer 9 in this case terminates the change of the actual correction values as soon as at least one of the following termination criteria is fulfilled:

- The first optimizer 9 completed the specified number of iterations.

- The first optimizer 9 changed the actual correction values s within a given time.

- The current correction values s have changed only slightly from the previous iteration. To this end, a member

can be compared with a given threshold value. If the specified term is below this threshold value, then the first optimizer 9 detects only a slight change.

- Ratio

changed in general relative to the previous iteration only slightly. To this end, said ratio can be compared with a predetermined threshold value. If the specified term is below this threshold value, then the first optimizer 9 detects only a slight change.

Alternatively or additionally, it is also possible that the first optimizer 9 checks whether another termination criterion is met. The decisive thing is that, thanks to the resulting termination criterion, it is ensured that only a finite number of iterations are performed.

The most recently identified actual correction values s must comply with the constraints according to inequality (7) or according to inequalities (7) and (8). For correction values s that are determined at an intermediate time and changed again thereafter, this is not necessary. Preferably, however, the first optimizer 9 considers the constraints at each iteration. This is the case in particular if the first optimizer 9 operates according to the continuous optimization method, in particular according to the interior point method.

For the final determination of the control variables S, the control device implements the design according to FIG. 5 preferably flatness controller 11. The flatness controller 11 is located behind the first optimizer 9. The actual corrective values s detected by the first optimizer 9 are fed to the flatness controller 11 . It determines from them the control variables S for the flatness adjusting elements 8 .

In principle, the flatness controller 11 can be implemented in different ways, for example as a conventional PI controller. However, as depicted in FIG. 5, the flatness controller 11 is designed as a controller according to the observer principle.

Accordingly, the flatness controller 11 first multiplies the actual correction values s by means of a multiplier 12 by the gain factor KP. The gain KP is always positive. As a rule, it is less than 1. If possible, the gain factor KP should be chosen as large as possible. To the weighted actual correction values determined in this way, the output signal s`` of the model 14 of the rolling stand section is added at the nodal point 13 .

On the basis of the sum thus formed, the flatness controller 11 determines the pre-signal S'. The pre-signal S` - like the control variables S and correction variables s, s` - is a vector. In the simplest case, the preliminary signal S` is identical to the formed sum. However, as a rule, in order to determine the pre-signal S', the resulting sum is filtered in the filter 15. The filter 15 can be implemented in particular in the form of a low-pass filter. It is possible that the filter 15 is adjusted only once during commissioning of the rolling stand. Preferably, however, the control device 4 can also parameterize the filter 15 P at later times again and thereby parameterize the filter 15 dynamically. The flatness controller 11 then differentiates the pre-signal S' in the differentiator 16.

The differentiated signal is then integrated in the integrator 17. The output signal of the integrator 17 corresponds to the control variables S or, if the control variables S are obtained as a sum of several terms, to one of the terms. Control values S are issued to the elements 8 adjusting the flatness. It is possible that the integrator 17 is an integral part of the flatness controller 11 . Alternatively, it can be located outside the flatness controller 11 .

The pre-signal S' is supplied by the flatness controller 11 to the rolling stand section model 14 as an input signal. The plot model 14 models the operation of the flatness adjustment elements 8 from the point of view of the control device 4 . In particular, the section model 14 models with which time step response the measuring device 7 registers the flatness error that has occurred in the gap between the rolls of the rolling stand. The model parameters required for this can usually be determined without problems from the plant geometry. This is known to the expert.

If y is controlled in this way at a certain flatness adjusting element 8 at time x, the area model 14 reproduces what action the control of y shows at what time t when registering the measured value. The plot model 14 takes into account the dynamic behavior of the respective flatness control element 8 . The section model 14 further takes into account any dead time, for example the transport time, which (when considered in the transport direction of the metal strip 1) elapses from the action of the corresponding flatness adjusting element 8 on a specific location of the metal strip 1 until the measured flatness values fM for that location are recorded by the measuring device 7 . Also, the region model 14 takes into account any delay time in registering the measured value.

The control device 4 often has access to the models 18 of the rolling stand. For example, the models 18 can be integrated into the control device 4. Models 18 simulate the behavior of a rolling stand during operation. The models 18 may include, for example, a rolling force model, a bending model, a collapse model, a roll gap model, a model for modeling thermal and wear-induced barreling of rolls 2, 3 of the rolling stand, and many others. Preferably, the control device 4 is addressed according to the representation in FIG. 6 at least within the framework of commissioning the rolling stand in step S1 to the models 18 and automatically determines the efficiency matrix W based on the models 18 in the determination device 19 . Only then is the rolling of the metal strip 1 or, if necessary, several metal strips 1 carried out in step S2.

Preferably, steps S3 and S4 are additionally provided. In this case, the control device 4 checks in step S3 whether a new metal strip 1 is to be rolled. If so, the control device 4 proceeds to step S4. In step S4, the controller refers - as in step S1 - to the models 18 and automatically determines the efficiency matrix W based on the models 18 . By means of steps S3 and S4, it is realized in such a way that the controller 4 dynamically determines the efficiency matrix W in each case just before starting to roll the corresponding metal strip 1 again.

It is even possible that the control device 4 dynamically monitors the efficiency matrix W also during the rolling of the respective metal strip 1. This is also the case in the further explained embodiments according to FIG. 7 and 8. If tracking is to be performed, then in the case of the embodiments of FIG. 5 and 7, this can be implemented, for example, by steps S5 to S7. In this case, the controller determines in step S5 the expected flatness values f1. The expected flatness values f1 can be determined, for example, by models using the section model 14 and/or the rolling stand models 18 . In step S6, the control device 4 evaluates the expected flatness values f1 and the measured flatness values fM. Based on the evaluation of step S6, the controller 4 can then monitor the performance matrix W in step S7. In the case of the embodiment according to FIG. 8 dynamic tracking can be implemented differently.

Further, in combination with FIG. 7 explains a modification of the embodiment of FIG. 5. Within the embodiment of FIG. 7 has a further optimizer 9`. Therefore, the optimizer 9` is referred to hereinafter as the second optimizer to distinguish it from the first optimizer 9. The second optimizer 9' is similar to the first optimizer 9. Further, in combination with FIG. 7 explains the operation of the second optimizer 9`. It should be noted that although the same values are used in the following explanation, the values are nonetheless independent of the values used for the first optimizer 9. Thus, although the same reference numerals are used, different values may be used.

The second optimizer 9` similarly determines the actual correction values s by minimizing the relation:

(9)

(9)

The second optimizer 9' thus changes the actual correction values s valid for it until it minimizes this ratio. The second optimizer 9' preferably takes into account the same constraints as the first optimizer 9. In addition, the second optimizer 9' determines the correction values s also preferably according to the interior point method. However, for the second optimizer 9` the weight coefficients α, β have the value 0. The second optimizer 9` optimizes the following ratio as a result:

(10)

(ten)

For this reason, for the second optimizer 9` it does not matter what value the correction values s` of the previous working cycle T have.

f is - similarly to the first optimizer 9 - a set of flatness values, that is, again a vector. The flatness values f are determined by the optimizer 9 on the basis of the initial flatness values f0 and the actual correction values s valid for the second optimizer 9`. The second optimizer 9` determines - similarly to the first optimizer 9 - the flatness values f according to the relation:

(11)

(eleven)

The initial flatness values f0 correspond in the case of the second optimizer 9', however, not to the measured flatness values fM, but to the set of nominal flatness values fW. These in turn correspond to the nominal change FWN of the rolling force FW. W is - as before - the efficiency matrix. It individually indicates which effect a certain individual correction value s has on which of the flatness values f. According to the image in Fig. 7, in particular, we can talk about the same efficiency matrix W, which is also used for the first optimizer 9.

In the case of the embodiment according to FIG. 7, the control device 4 determines the control values S for the flatness adjusting elements 8, additionally taking into account the actual change δFW of the rolling force FW, the nominal change FWN of the rolling force FW and the actual correction values s detected by the second optimizer 9'. In particular, the control device 4 calibrates the actual correction values s detected by the second optimizer 9' by a factor F, the factor F being the sum of the partial actual change δFW of the rolling force FW and the nominal change FWN of the rolling force FW:

(12)

(12)

If necessary, in addition to this - before or after graduation by the coefficient F - equalization in the filter can be carried out.

As already noted, the individual blocks of the internal structure of the control device 4 are, in the case of a software implementation, software modules. In the case of the embodiment according to FIG. 7, it is therefore possible to use the same program module - namely the optimizer implementation - for both the first and the second optimizer 9, 9`. In particular, for this reason, in FIG. 7 at the same time also block 10` delay. Both software modules need only be individually parameterized for the respective application. For example, weight coefficients α, β can be set for the first optimizer 9 to values other than 0, and for the second optimizer 9, 9` to 0.

Within the embodiment of FIG. 7, there are first and second optimizer 9, 9', i.e. both optimizer 9, whose actual correction values s are fed to the flatness controller 11, and optimizer 9', whose actual correction values s are merely multiplied by a factor F. However, according to the image on fig. 8, which depicts the second main embodiment of the present invention, it is also possible that there is only one single optimizer 9' that operates in the manner explained above in connection with FIG. 7 for the second optimizer 9`. In this case, the optimizer 9', whose actual correction values s are merely multiplied by the factor F, is thus the only and thus the first optimizer 9' according to the present invention.

A separate detailed explanation of the operation of the first optimizer 9' of FIG. 8 is not required in this case, since the first optimizer 9' of FIG. 8 works in the same way as the second optimizer 9' of FIG. 7.

This invention has many advantages. Thanks to the online optimization, the optimal control variables S can be determined in each operating cycle T. By designing the flatness controller 11 as an observer controller, a reaction to disturbances or other changes can be carried out as quickly as possible. The models 14, 18 required for the implementation of the present invention are usually already present in the control device 4 . As a result, there is no additional effort to develop such models 14, 18. The filter 15 makes possible an efficient, plant-specific adaptation of the flatness controller 11 to the disturbance behavior. By dividing into a first optimizer 9 on the one hand and a flatness controller 11 on the other hand, the flatness error analysis (which occurs in the first optimizer 9) is further decoupled from the dynamic adjustment behavior (which occurs in the flatness controller 11). As a result, conditions are created for a simple, modular configuration of the control device 4 for almost any type of stand. This facilitates the parameterization of the control device 4 during the engineering phase and shortens the commissioning time. In addition, installation-dependent adaptations can be carried out in a targeted and simple manner. Decoupling further simplifies implementation so that errors are avoided both in the engineering phase and in the commissioning phase. Optimum adjustment increases the service life of the mechanical components of the rolling stand. By compensating for flatness errors due to changes in the rolling force FW by means of pre-adjustment, the throughput is further increased, since disturbances are reacted very quickly. Operational economy is increased as less scrap is produced. Thanks to the optimization carried out, the position limits of the flatness adjustment elements 8 are further taken into account already within the framework of determining the control variables S.

While the invention has been illustrated and described in detail and in detail by means of the preferred embodiment, the invention is not limited to the disclosed examples, and other variations may be deduced from there by the skilled person without leaving the protection scope of the invention.

LIST OF REFERENCES

1 metal strip

2 work rolls

3 backup rolls

4 control device

5 computer program

6 machine code

7 measuring device

8 flatness adjustment elements

9, 9` optimizers

10, 10` delay blocks

11 flatness adjuster

12 multiplier

13 nodal point

14 plot model

15 filter

16 differentiator

17 integrator

18 rolling stand models

19 detection device

A axial displacement

B, c vectors

C matrix

b width

F coefficient

f flatness values

f0 initial flatness values

fM measured flatness values

fW flatness ratings

f* calculated flatness values

FR back bending force

FW rolling force

FWN nominal change in rolling force

K local cooling

KP gain

P parameters

S control variables

S` preliminary signal

S1 to S7 steps

s correction values of the current operating cycle

s` correction values of the previous working cycle

s`` rolling stand model output

T working cycle

W efficiency matrix

α, β weight coefficients

δFW actual change in rolling force

Claims (60)

1. A method for controlling the flatness of a metal strip (1) rolled in a rolling stand, including determination by means of the rolling stand control device (4) with an operating cycle (T) of a series of control variables (S) for the corresponding number of rolling stand flatness adjustment elements (8) and control of the flatness adjustment elements (8) in accordance with the determined control variables (S) , characterized in that by means of the control device (4) the first optimizer (9, 9') is implemented, with the help of which the actual corrective values (s) are first preliminarily set and according to the ratio

or

, where f is the set of flatness values (f), s is the set of actual correction values (s), f0 are the initial values (f0) of flatness, W is the efficiency matrix (W) which individually indicates which effect the individual correction value s has on which of the flatness values (f), s' is a set of correction values (s') determined in the previous working cycle (T), determine the set of values (f) flatness, and then by changing the actual correction values (s) minimize the ratio

, where f* is the set of design values (f*) of flatness, s0 is the set of target values (s0) for correction values (s), α and β are non-negative weight coefficients (α, β), to determine the actual correction values s for the corresponding operating cycle T, which are the latest or last modified when minimizing the said ratio by the actual correction values s, moreover, when determining the actual corrective values (s), the first optimizer (9) takes into account linear constraints, wherein by means of the control device (4) the flatness controller (11) located after the first optimizer (9) is implemented, to which the actual correction values (s) determined by the first optimizer (9) are supplied and which, on the basis of the actual correction values (s), determines the change values for the control values (S) for elements (8) of flatness adjustment, moreover, by means of the flatness controller (11) form the sum of the gain-weighted coefficient (KP) of the actual correction values (s) and the output signal (s'') of the model (14) of the section of the rolling stand, a preliminary signal (S') is determined on the basis of the sum formed in this way, by differentiating the pre-signal (S'), the change values for the control variables (S) for the flatness adjustment elements (8) are determined, and submitting a preliminary signal (S') to the model (14) of the section of the rolling stand as an input signal. 2. The method according to p. 1, characterized in that using the measurement device (7), the measured flatness values (f0) are recorded with a spatial resolution along the width (b) of the metal strip (1), wherein the measured flatness values (f0) and the corresponding calculated flatness values (f*) are fed to the control device (4) as initial flatness values (f0). 3. The method according to any one of paragraphs. 1 or 2, characterized in that by means of the control device (4), the second optimizer (9') is implemented, which is similar to the first optimizer (9), wherein the set of flatness values (f) for the second optimizer (9') is determined on the basis of the set of nominal flatness values (fW) and the actual correction values (s) valid for the second optimizer (9'), and the nominal flatness values (fW) correspond to the nominal change (FWN) of the rolling force (FW), where weight coefficients (α, β) for the second optimizer (9') have the value 0, and the control values (S) for the flatness adjustment elements (8) are determined by the control device (4) with additional consideration of the actual change (δFW) of the rolling force (FW), the nominal change (FWN) of the rolling force (FW) and those determined by the second optimizer (9' ) actual correction values (s). 4. The method according to p. 1, characterized in that the initial values (f0) of flatness correspond to the set of nominal values (fW) of flatness, the nominal values (fW) of flatness correspond to the nominal change (FWN) of the force (FW) of rolling, and the weight coefficients (α, β) have the value 0, while the control values (S) for the flatness adjusting elements (8) are determined by the control device (4) taking into account the actual change (δFW) of the rolling force (FW), the nominal change (FWN) of the rolling force (FW) and certain actual correction values (s) . 5. The method according to any one of paragraphs. 1-4, characterized in that the actual correction values (s) are changed by the first optimizer (9, 9') in several iterations. 6. The method according to p. 5, characterized in that changing the actual correction values (s) using the first optimizer (9, 9') is completed under the condition: - the first optimizer (9, 9') has performed the specified number of iterations, and/or - the first optimizer (9, 9') changed the actual correction values (s) within a given time, and/or - the actual correction values (s) change only slightly from iteration to iteration, and/or - ratio

changes only slightly, and/or - another criterion for terminating changes is met. 7. The method according to claim 5 or 6, characterized in that the first optimizer (9, 9') considers linear constraints at each iteration. 8. The method according to any one of paragraphs. 1-7, characterized in that correction values (s) are determined by the first optimizer (9) according to the interior point method. 9. The method according to p. 8, characterized in that the efficiency matrix (W) is automatically determined by the control device (4) based on the models (18) of the rolling stand. 10. The method according to p. 9, characterized in that the efficiency matrix (W) is determined dynamically by means of the control device (4) anew each time just before the start of rolling of the respective metal strip (1). 11. The method according to p. 10, characterized in that the efficiency matrix (W) is dynamically monitored by the control device (4) during the rolling of the respective metal strip (1).

Images