EP1711283B1 - Control method and control device for a roll stand - Google Patents

Abstract

According to the invention, the thickness, traction forces and flatness is controlled by means of a single control device as part of an integrated, model-predictive thickness, traction and flatness control operation. The integrated control device takes into account the influence the adjustment of the control variables has on the thickness, the traction force and the flatness of the rolled strip and optimises the modifications of the control variable in such a way that a selected quality of the thickness and the flatness control is obtained. The quality of the thickness and the quality of the flatness control, which are weighted differently, can effect the quality function of the control device. The use of the integrated control device improves the performance and the stability of the control device in relation to the control devices which are arranged in a separate manner from each other.

Description

The invention relates to a control method and a controller for a rolling mill, or for the rolling stands of a rolling mill.

When rolling tapes, in particular cold rolling, separate control systems are provided to control the flatness and thickness of the tape at the outlet of the stand and to control the tension acting on the tape.

A control system for the band flatness is, for example, in EP 1 181 992 A2 described. There, the planarity of the tape is detected with a measuring system and compared with the result of an explicit, linear or non-linear on-line profile and planarity model, which takes into account all major factors involved in the rolling process (bending, pivoting, sliding, thermal crowning). The determined flatness error is decomposed into orthogonal components for simplified optimization in the control system. This generation of the flatness error to be taken into account in the subsequent control is designed as an event-triggered scanning system for the consideration of flatness measuring systems with variable sampling time. The determined flatness error is fed to a multi-variable controller. The known control system also has a prediction of the controlled variable, which is included in the dynamic optimization and goes beyond the dead time on. Further, feedforward control takes into account the properties of the incoming strip, the variation of rolling force and thermal crowning. However, it has been shown that this control system does not meet the high quality requirements of the rolled product.

An integrated thickness and flatness control for 20-roller stands will be available in Pu H., Nern H.-J., Roemer R., Nour Eldin HA, Kern P., Jelali M .: State observer design and verification towards integrated flatness control system for the 20 roll sendzimir cluster mill, Proc. Intern. Conf. on Steel Rolling (Steel Rolling '98), 1998, The Iron and Steel Institute of Japan, Chiba, p. 124-129 , and Pu H., Nern H.-J., Nour Eldin HA, Jelali M., Totz O., Kern P .: The Hardware-in-Loop simulations and online tests of an integrated thickness and flatness control system for the 20 rolls of Sendzimir cold rolling mill, Proc. Intern. Conf. on Modeling of Metal Rolling Processes, 1999, London, p. 208-216 and Pu H., Mikhailov L., Nern H.-J., Kern P., Nour Eldin HA: Optimal control of thickness and flatness for the 20 rolls Sendzimir cold rolling mill, Proc. IFAC World Congress, 1999, Beijing, China, p. 481-486 described. The control methods described there can not meet the high quality requirements for the rolled product.

Against this background, the invention has the object to provide an improved control method for a rolling stand.

This object is solved by the subject matter of the independent claims. Advantageous embodiments are specified in the subclaims.

The invention is based on the basic idea to carry out the regulation of the thickness, the train and the flatness with a single controller in the context of an integrated, model-predictive thickness, draft and planarity control. The integrated control thereby takes into account the influence that the adjustment of manipulated variables has on both the thickness, the tension and the flatness of the rolled strip and can optimize the manipulated variable change so that a selected quality of the thickness control and the flatness control is achieved. The quality of the thickness control and the quality of the flatness control can be weighted differently in the quality function of the controller. The use of an integrated control leads to a performance improvement and stability improvement in the regulation compared to separately designed regulations.

From Patent Abstract of Japan to JP 60 018 213 A For example, a control method for a rolling stand with measuring systems for detecting the sheet flatness and the sheet profile is known. Furthermore, the rolling stand described there has a control that from the measurement results of the thickness sensor and the flatness sensor generates a compensation amount which is applied to the output of a model predictive roll bending controller.

From Patent Abstract of Japan to JP 58 065 510 A is a control method with a thickness measurement and a rolling force measurement known. There, two separate regulators are used, which together control the control valve of a hydraulic cylinder, which generates the rolling force in the rolling stand. The measurement result of the rolling load sensor is fed to the one controller. The measurement result of the layer thickness measurement is fed to the second controller. A contemplation of the strip or band flatness does not take place here.

From the article " Danieli launches new control and automation system "in Aluminum TODAY, October / November 2000, pages 33 to 34 a control method is known, which consists of several levels, wherein in a first level, a thickness control, a flatness control and a control for the coolant and the flatness of the rolled product are provided. Furthermore, a second level is provided to support the management of the rolling mill.

Out US Pat. No. 5,586,221 is a control method for the thickness control in a rolling stand known.

It has been found that, in particular due to the transverse coupling between thickness and flatness, an interaction of the two control circuits can be determined. In particular, with thin tapes conventionally the prescribed thickness tolerance and the desired flatness could not be achieved simultaneously. In fact, in general, flatness control has been slower to avoid the risk of instability. As a result, however, they can not respond to rapidly changing flatness error, so that the achievable flatness quality is severely limited. By taking into account both the thickness control quality and the flatness control quality, a common controller can adjust both the prescribed thickness tolerance and the desired flatness.

In addition, the final thickness, especially at very small band thicknesses, such as occur during cold rolling, depends on the applied trains. The rolling forces also depend on the band moves. This results in a strong coupling between tension, thickness and flatness control. These are inventively taken into account by the model used by the controller in the one controller. As a result, as usual, the thickness and flatness control no longer be set slower, so that even rapid thickness and flatness fluctuations can be well balanced.

For this purpose, in the control method according to the invention, input variables for the controller are generated as a function of the measured values of the measuring systems. These input variables are used by the controller to generate at least one control signal for at least one manipulated variable of the roll stand due to an integrated, model-predictive thickness, draft and planarity control. Preferably, separate measuring systems are provided for the thickness, flatness and tension of the belt. However, in the context of this invention, measuring systems can also be used which determine several variables, for example the thickness and the flatness, at the same time.

The manipulated variables of the rolling process are in particular the roll bending, the roll pivoting, the roll shifting, the roll cooling, in particular a selective multi-zone cooling and also the change of the support roll form.

The controller uses a prediction model to predict (predictively) the future system behavior. The controller is preferably a Model Predictive Control (MPC) controller embedded in an Internal Model Control (IMC) structure. MPC controllers are off, for example Camacho EF, Bordons C: Model Predictive Control, Springer, 1999 , and Maciejowski JM: Predictive Control with Constraints, Prentice Hall, 2002 , well known, which is why for the description of MPC controllers and their interpretation in full terms to these publications reference is made. IMC structures are made in particular Garcia CE, Morari M .: Internal model control. 1. A unifying review and some new results, Ind. Eng. Chem. Process Des. Dev. 21 (1982), p. 308-323 , well known, therefore, for the description of IMC structures and their interpretation, reference is made to this publication in its entirety.

Preferably, the controller further includes an integrated thickness profile control. The manipulated variables used to control the thickness profile are often the same as those used to control thickness and flatness.

In a preferred embodiment, the controller is constructed as a multi-variable controller with a dynamic optimization, in which the thickness and the flatness are received differently weighted. A dynamic optimization algorithm is preferably used in order to determine optimum actuator positions taking into account predetermined manipulated variable limitations of the manipulated and controlled variables. For example, in the controller, the quality of the controller can be represented by the following cost function: $$J\left(G_{\mathrm{thickness}},G_{\mathrm{flatness}}\right)=G_{\mathrm{thickness}}{J}_{\mathrm{thickness}}+G_{\mathrm{flatness}}{J}_{\mathrm{flatness}}$$

In this quality function the power functions for the thickness J _thickness and flatness J _flatness by the weighting factors g g _thickness and _flatness go a weighted differently. For production reasons, compliance with the thickness tolerance, for example, a greater weight, ie a g _thickness > g _flatness , be assigned. Deviations from the target thickness are therefore punished much more severely than deviations from the desired flatness.

$$J_{\mathrm{Planheit}}=\frac{1}{2}{\left({\mathit{\Omega }}_{\mathrm{soll}}-{\mathit{\Omega }}_{\mathrm{ist}}\right)}^T{\mathit{\Omega }}_{\mathrm{Planheit}}\left({\mathit{\Omega }}_{\mathrm{soll}}-{\mathit{\Omega }}_{\mathrm{ist}}\right)$$

The quality functions for the thickness and the flatness can be represented for example as follows: $$J_{\mathrm{thickness}}=\frac{1}{2}{\left({\mathit{H}}_{\mathrm{should}}-{\mathit{H}}_{\mathrm{is}}\left(\sigma \right)\right)}^T{\mathit{Q}}_{\mathrm{thickness}}\left({\mathit{H}}_{\mathrm{should}}-{\mathit{H}}_{\mathrm{is}}\left(\sigma \right)\right)$$

$$J_{\mathrm{flatness}}=\frac{1}{2}{\left({\mathit{\Omega }}_{\mathrm{should}}-{\mathit{\Omega }}_{\mathrm{is}}\right)}^T{\mathit{\Omega }}_{\mathrm{flatness}}\left({\mathit{\Omega }}_{\mathrm{should}}-{\mathit{\Omega }}_{\mathrm{is}}\right)$$

$$\left|h_{\mathrm{ist}}-h_{\mathrm{soll}}\right|\le h_{\mathrm{tolmax}},\left|{\sigma }_{\mathrm{ist}}-{\sigma }_{\mathrm{soll}}\right|\le {\sigma }_{\mathrm{tolmax}}$$

Alternatively, the quality of the thickness control and the quality of the flatness control can be achieved by the following quality function: $$\underset{u}{\min }{J}_{\mathrm{flatness}}\left(u\right)$$

$$\left|H_{\mathrm{is}}-H_{\mathrm{should}}\right|\le H_{\mathrm{tolmax}}.\left|{\sigma }_{\mathrm{is}}-{\sigma }_{\mathrm{should}}\right|\le {\sigma }_{\mathrm{tolmax}}$$

The square flatness error is minimized under the proviso that the thickness deviation must always be smaller than an upper barrier. The solvability of this optimization problem can be guaranteed by means of suitable measures, for example with the aid of a Feasible SQP algorithm Maciejowski JM: Predictive Control with Constraints, Prentice Hall, 2002 , which are referred to in full for the solvability of the optimization problem).

In the mentioned optimization algorithms, the train to be taken into consideration usually covers the thickness, since the thickness depends strongly on the train. However, optimization algorithms can be set up to optimize the three sizes thickness, flatness, and tension separately.

Preferably, the prediction of the controlled variable is included in the dynamic optimization. The prediction preferably proceeds beyond the dead time compensation.

The controller preferably takes into account restrictions, in particular for the actuating signals of the manipulated variables and the systematic softening of the restrictions as a function of their importance for the trouble-free operation of the rolling stand. The restrictions are in particular absolute values and rates of change of the manipulated variables. This preserves the feasibility of the optimization problem. As a result, certain boundary conditions of the roll stand can be maintained when rolling sheets. For example, only certain maximum relative positions are allowed to each other for crown excenter of a Sendzimir rolling mill. By taking into account such boundary conditions, the positioning possibilities of the rolling stands, for example a Sendzimir skeleton can be fully utilized.

The controller preferably utilizes an explicit, linear or non-linear on-line profile and planarity model that takes into account the major rolling process variables and actuators, particularly roll bending, roll pivoting, roll shifting, roll cooling, and / or backup roll shape modification. The design of such models is, for example, Berger B., Mücke G., Neuschütz E., Fleischer H. (1982) Regulation of the flatness and the tensile stress distribution on a 20-roll cold rolling mill, BFI Report No. 893 (Final Report of the ECSC research project No. 7210.EA 109); Jelali M. (2000) Explicit models of thickness profile and tension distribution for process control applications. Steel Research 71, No. 6 + 7, 228-232; Schneider A. (2000) Online Modeling and Optimization for a 20-high Rolling, Mill. Dissertation, University of Wuppertal, which is why reference is made to these publications for the description of such models and their interpretation in their entirety. In a preferred embodiment of the control method, an explicit, online-enabled functional model is also used which calculates setpoint values for the flatness control.

Preferably, simplified prediction models are partly used in the controller, for example by linearization and simplification of corresponding relationships of the complex model. The design of such models is, for example, from Jelali M. (2000) Explicit models of thickness profile and tension distribution for process control applications. Steel Research 71, No. 6 + 7, 228-232 is well known, so for the description of such models and their interpretation is fully incorporated by reference to these publications.

The control method preferably has at least one adaptation method for the online-enabled profile and flatness model, the online-enabled prediction model and / or a set-up model. The adaptation method preferably adapts selected parameters of the model components. These adaptation methods are preferably designed to be robust against model structure errors. By adapting the models, it is possible to take into account changes in the dynamic behavior of the roll stand, as resulting, for example, from wear or the replacement of components. Thus, the adaptation method can adapt the setup models from stitch to stitch.

According to an advantageous embodiment of the control method according to the invention, an event-triggered scanning system is used, which takes into account flatness measuring systems with variable sampling times.

In order to keep the number of controlled variables as small as possible and thus to simplify the optimization problem, according to an advantageous embodiment of the control method, the flatness profile determined from the measurements of the flatness measuring system is decomposed into orthogonal components with the aid of orthogonal functional systems. For this purpose, in particular the Chebyshev polynomial (this is in full Press WH, Teukolsky SA, Vetterling WT, Flannery BP: Numerical Recipes in C, Cambridge University Press, 1992 , referenced), the Gram polynomial (this is in its entirety Ralston A., Rabinowitz P .: A First Course in Numerical Analysis, International Series in Pure Applied Mathematics, McGraw-Hill, 1978 ) or other orthogonal polynomials.

Through a feedforward control used in a preferred embodiment, which takes into account the properties of the incoming strip, the variation of the rolling force and / or the variation of the thermal crowning, resulting disturbances can be compensated. The compensation of the disturbances can be done in a separate module or integrated into the model predictive control.

Preferably, the flatness measuring system estimates the flatness profile across the width of the strip on the basis of measurement results obtained at different times and in particular at different points along the width of the strip. In this case, the current control variables are preferably taken into account. Particularly preferably, the flatness profile is determined immediately after the presence of the next, new measured value of a sensor. This produces a more recent estimate of the flatness profile that does not depend on all the measurements being across the width of the band. For calculating the flatness during one revolution of the measuring roller, a switching Kalman filter is preferably used. For the construction of a switching Kalman filter is fully directed to the parallel application 103 06 837.6. By this determination of the flatness course can also with the use of Planheitsmeßrollen, as used for example during cold rolling, the planning process can be determined promptly. Sensors are distributed over the radius and the width on the flatness measuring roll. These provide temporally spaced from each other and with respect to the width position shifted from each other information about existing at the respective location at the respective measurement time flatness of the tape. By estimating the course of the planarity after the determination of each individual measured value, even rapidly changing planarity changes can be taken into account. These result in measuring systems in which first all measured values are determined over the width of the strip and thus a complete flatness distribution can be determined only after a complete revolution of the planarity measuring roll, in particular with rapidly changing flatness distributions to considerable measuring errors.

A controller according to the invention converts the method steps and properties described above individually or in combination.

By integrating the thickness, tension, and flatness control in a controller, or building a coordination component of these controls, performance degradation and stability problems can be avoided. The coupling, which exists between thickness, tension and flatness, is taken into account by a decoupling matrix, which can be calculated from the inverse of the total transmission matrix of the thickness and flatness control system.

The invention will be explained in more detail with reference to a drawing. In this drawing, an embodiment is shown. It shows the structure of the control method according to the invention.

The illustrated control structure, as it can be used, for example, for a cold rolling stand (rolling stand) 1, mainly has a flatness measuring system 2, a multi-variable controller 3 (MPC module thickness and flatness) and a train control 4. A flatness measuring system 2 can be arranged both in front of and behind the framework 1 so that the framework 1 can be used in reversing operation. The rolling mill 1 is used in this embodiment for rolling very thin strips. Here there is a strong coupling between the factors of thickness and tension on the rolling result.

As shown, the flatness deviation is determined by means of a flatness measuring system at the outlet of the framework. The flatness measuring system is preferably based on a flatness measuring roller. This measures the belt tension discretely at individual measuring points distributed over the measuring roller width and the measuring wheel circumference. The flatness profile (flatness distribution) is estimated immediately on the basis of the individual measurement results. The estimated flatness profile is decomposed into orthogonal (independent) components. In this case, the type of decomposition used is changed depending on the nature of the occurring flatness error in order to describe the flatness error with as few components as possible and thus to simplify the optimization problem.

The orthogonal components thus determined are compared with values provided by an on-line model of the plant. The resulting difference is used as a control variable and fed to the multi-variable controller 3. There is a comparison with a decomposed into independent components target flatness curve. The multi-variable controller consists of an online-enabled model and a dynamic optimization, including manipulated variable limitations and predicted control variable course.

From the input variable, the controller determines control signals for roll bending, roll pivoting, axial displacement of the rolls, as well as for multi-zone cooling and, if necessary, a change in the back-up roll shape. To take into account the influence of the rolling force, the properties of the incoming strip (for example, a measured in the inlet of the scaffold Flatness, and the thermal crowning) also a feedforward control is made, which compensates for these influences. In addition, starting from the desired flatness curve, a feedforward control is also introduced, which is likewise introduced into the actuating signals determined by the multivariable controller.

The controller selects a primary manipulated variable as a function of the strip thickness of the strip to be rolled over which it preferably influences the rolling process. For strip thicknesses below a specified size, the train is used as the primary control variable. The contact force and the adjustment position of the rollers is then treated as additional secondary control variable.

In order to be able to compensate for changes in the dynamic behavior, which can be caused, for example, by wear, the replacement of components of the roll stand and changes in the material properties of the roll stand, the models are adapted online during the rolling of a single strip (in-bar adaptation). In addition, there will be an adaptation from fret to fret. The in-bar adaptation compensates for relatively rapid changes, for example caused by changes in the tape's temperature, while compensating for wear-related changes due to the band-to-band adaptation.

An event generator allows the use of an event-triggered sampling system to accommodate variable-time flatness measuring systems.

Claims (11)

1. Control process for a roll stand having measuring systems for recording the thickness, tension and flatness and a control unit, characterised in that input variables are created for the control unit dependent on the measured values of the measuring systems, and in that the control unit produces at least one control signal for at least one control variable of the roll stand based on the integrated, model-predictive control of thickness, tension and flatness.
2. Control process according to Claim 1, characterised by a multiple variable control unit which has dynamic optimisation into which the thickness and flatness are received differently weighted.
3. Control process according to Claim 2, characterised in that a prediction of the control variables is included in the dynamic optimisation.
4. Control process according to Claim 2 or 3, characterised in that the restrictions for dynamic optimisation and the systematic softening of restrictions are included.
5. Control process according to any one of Claims 1 to 4, characterised in that the control unit applies an explicit, linear or non-linear, online-compatible profile and flatness model.
6. Control process according to any one of Claims 1 to 5, characterised by at least one simplified prediction model.
7. Control process according to either of Claims 5 and 6, characterised by an adaptation process for the online-compatible profile and flatness model, the online-compatible prediction model and/or a set-up model.
8. Control process according to any one of Claims 1 to 7, characterised by a disturbance variable feed-forward which compensates the disturbances resulting from the properties of the incoming strip, the variation in the rolling force and/or thermal cambers.
9. Control process according to any one of Claims 1 to 8, characterised in that the flatness behaviour determined from a measurement by the flatness measuring system is broken down into orthogonal components by means of orthogonal function systems.
10. Control process according to any one of Claims 1 to 9, having a process to determine flatness, during rotation of a measuring roller, by means of measurements which are carried out transverse to the direction of travel of the strip offset to one another, having the following steps:

    - Determining a measured value at at least one measuring point

    - Calculating the characterising values of a flatness behaviour equation from the determined measured value(s) by means of a switching Kalman filter.
11. Control unit for a roll stand, which implements the control steps of a control process according to any one of Claims 1 to 10.

Images