EP2384830A1 - Method for determining the parameters of a model for a roller assembly - Google Patents

Abstract

The method involves providing initial values to a parameter to be determined to a model (13), and feeding the initial values to the model by measuring input parameters that are determined within the roller system. A model-initial parameter is computed by the model based on the input parameters and the initial values of the parameter. The model-initial parameter is compared with an associated initial parameter that is determined by measurement within the roller system. Additional values for the parameter of the model are determined based on the comparison. Independent claims are also included for the following: (1) a method for operating a roller system (2) a computing device i.e. notebook, for performing a method for determining parameters of models.

Description

The invention relates to a method for determining parameters of a model, with which at least one area of a rolling mill can be simulated mathematically. The invention also relates to a method for operating a rolling mill, wherein at least one area of the rolling mill is simulated by calculation with the model having a parameter.

From the DE 10 2006 025 026 A1 a method for operating a rolling mill is known, in which a measured actual value of a Operating size of the rolling mill compared with an associated setpoint and depending on a manipulated variable is determined. It is provided a so-called process observer, which includes a model, in particular a real-time simulation model, with which at least part of the rolling mill is mathematically modeled. Depending on the operating variables of the rolling plant, for example as a function of the desired value, a calculated actual value associated with the measured actual value is determined from this model. Then the measured actual value and the calculated actual value are compared with each other. From the result of the comparison, it is possible to deduce an error in the rolling mill or in the rolling process.

As mentioned, with the model at least part of the rolling mill is simulated in real time by calculation. For this purpose, the model has a plurality of equations with which output variables of the rolling mill are calculated as a function of input variables of the rolling mill. To weight the input variables, multiplicative and / or additive parameters are provided. These parameters must be determined and adjusted with regard to the actual rolling mill to be replicated or the model must be parameterized.

The object of the invention is to provide a method for determining the said parameters.

The invention solves this problem by the method according to claims 1 and 2. In the inventive method In doing so, the following steps are carried out: initial values for the parameters to be determined are given to the model, input variables determined by measurement within the rolling mill are fed to the model, the model calculates at least one model output variable as a function of the input variables and the initial values of the parameters, the model output is compared with an associated output determined by measurement within the rolling mill, and depending on the comparison, first new values are determined for the parameters of the model.

In the invention, therefore, the output variables are calculated by the model. This represents the simulation of the actual rolling mill by the model. Then the calculated model output variables are compared with the output variables present in the rolling mill. If there is a big difference here, this means that the rolling mill is not yet modeled correctly by the model. To improve the replica, the parameters of the model are then reset.

This procedure can be repeated several times. Thus, the model output variables can always be determined from the model using the respectively newly calculated parameters and then compared with the output variables present in the rolling plant. Depending on this comparison, the parameters of the model can then be changed over and over again so that the replica of the rolling mill can be improved by the model.

It is particularly advantageous if the input variables and the output variables form a data record and are assigned to a specific rolling process. Thus, the model can be parameterized by means of single or multiple use of the data set with regard to the associated rolling process.

In an advantageous development, several data sets are available, which are assigned to different rolling passes. Thus, the model can be parameterized or validated by using these different data sets on different rolling passes.

In a particularly advantageous embodiment of the invention, the data set or the data records are stored on a database which is assigned to the model. For example, the data sets can be determined with the aid of a computing device in the actual rolling mill and stored on a storage medium. Thereafter, the storage medium, for example, be taken over in a notebook on which the model is stored as a program. The inventive method can thus be performed on the notebook completely independent of the rolling mill and the local computing device. The parameterization of the model can thus be carried out "off-line", ie temporally and locally independent of the actual rolling mill.

In another embodiment, the record or records will be transferred to the model. In this case too, the parameterization of the model can be carried out, for example, on a notebook independent of the rolling mill. The data sets are not stored on the notebook, but are transferred from the rolling mill to the notebook. In the rolling mill, the data sets can be currently determined or measured or stored in a database associated with the rolling mill.

Other features, applications and advantages of the invention will become apparent from the following description of embodiments of the invention, which are illustrated in the figures of the drawing. All described or illustrated features, alone or in any combination form the subject matter of the invention, regardless of their summary in the claims or their dependency and regardless of their formulation or representation in the description or in the drawing.

FIG. 1 The drawing shows a schematic block diagram of an embodiment of a method for operating a rolling mill by means of a model and FIG. 2 shows a schematic block diagram of an embodiment of a method according to the invention for determining parameters for the model of FIG. 1 ,

That in the FIG. 1 illustrated method is provided for the operation of a rolling mill. This may be a cold rolling mill or a hot rolling mill. The in the FIG. 1 specified general sizes can be realized by different special operating variables of the rolling mill.

So will in the FIG. 1 a measured actual value IWG and an associated setpoint value SW are compared with one another by means of a subtraction 11. The measured actual value IWG may be, for example, a measured strip thickness actual value or a measured strip tension actual value or another actual value of the rolling mill. Accordingly, the setpoint SW may be an associated strip thickness setpoint or strip setpoint. The measured actual value IWG and the setpoint SW relate to a specific rolling stand.

The difference between the measured actual value IWG and the setpoint value SW is supplied to a controller 12, which, inter alia, generates a manipulated variable SG as a function of these two operating variables. With this manipulated variable SG, for example, the roller position of the associated rolling stand can be influenced, in order in turn to act on the strip thickness or the strip tension. In this way, a control and / or regulation of operating variables of the rolling mill can be realized. It is understood that other influences on the rolling mill are possible.

Regardless of these influences, the setpoint value SW is supplied to a model 13, in particular a simulation model operating in real time, which is furthermore acted on by a plurality of operating variables of the rolling mill. These operating variables can be any actual values, setpoints and / or manipulated variables that occur within the rolling mill. Depending on these input variables, individual models of the rolling mill or the entire rolling mill are modeled by the model 13. Thus, for example, an attempt is made to represent the deformation of the rolled stock during a pass through a roll stand in order to be able to determine an actual value for the same operating variable on which basis the measured actual value IWG is already present. This determined actual value represents an output variable of the model 13 and is referred to below as the calculated actual value IWB. There are thus two actual values, e.g. for the strip thickness or the strip tension, namely the measured actual value IWG and the calculated actual value IWB.

Generalized, therefore, the model 13 is provided to determine the calculated actual value IWB for the same operating size of the rolling mill, for which the measured actual value IWG already exists. It is understood that the model 13 is not only suitable for determining this one calculated actual value IWB, but rather a plurality of such output variables.

It is now possible that the time course of the measured actual value IWG and the calculated actual value IWB largely coincide with a fault-free operation, but that the calculated actual value IWB deviates in its absolute value from the setpoint SW. Thus, it is possible, for example, that an oil film present on the belt or rollers can not be taken into account by the model 13, with the result that the measured actual value IWG and the calculated actual value IWB absolutely deviate from each other due to the oil film, but relatively have essentially the same time course. To compensate for such deviations may be provided an adjustment, which is not mandatory, and therefore not explained below and is not taken into account.

The calculated actual value IWB is compared with the measured actual value IWG. For this purpose, the two variables of a subtraction 15 are supplied. On the output side results in an error signal FS, which is substantially equal to zero, when the calculated actual value IWB is approximately equal to the measured actual value IWG, but deviates from zero, if the calculated actual value IWB and the measured actual value IWG substantially differ.

In error-free operation of the rolling mill, the calculated actual value IWB will largely match the measured actual value IWG. This results from the fact that in this case there is no disturbance and thus that of the model 13 carried out arithmetic investigations essentially correspond to the actual operation of the rolling mill.

However, if there is any malfunction in the operation of the rolling mill, for example, a slippage of the rolls occurs during passage of the rolling stock through a rolling stand, so the actual operation of the rolling mill deviates from the behavior due to this disturbance, which is the mathematical basis for the model 13 , As a result, in the actual value IMB calculated by the model 13, the aforementioned disturbance is not taken into account, so that the calculated actual value IWB deviates with great probability from the actual value IWG actually measured in the rolling plant, at which the disturbance takes effect , This results in an error signal FS, which is not equal to zero.

The error signal FS thus represents the result signal of a process observer, which is substantially equal to zero in error-free operation, but which indicates erroneous operation by a non-zero output signal.

The determination of the error signal takes place in real time, ie substantially simultaneously with the respective current determination of the measured actual value IWG. This makes it possible to intervene immediately in the control and / or regulation of the rolling mill, in particular in the case of a fault signal FS different from zero, ie in particular also in real time.

For carrying out the described method, a device for controlling and / or regulating the rolling mill is provided. This may preferably be a digital computing device which is provided with an electronic memory, on which a program is stored, which can run on the computing device in real time, and which is then suitable for carrying out the method.

As has been explained, the above-described model 13 serves the FIG. 1 To simulate individual areas of the rolling mill or the entire rolling mill by calculation, in particular to simulate in real time. This is done by constructing the model 13 from a plurality of equations in which the input variables supplied to the model 13 are computationally processed. With the aid of the equations, the output quantity (s) of the model 13 is / are then calculated.

The equations can be any equations suitable for mathematically representing the actual technical relationships of the rolling mill. For example, it may be exponential equations of higher order, with which the time course of outputs in dependence on a plurality of input variables can be calculated. For the weighting of the individual input variables, each input variable is usually associated with a multiplicative and / or additive parameter which ultimately determines the influence of the Input size / s to the output size / s. These parameters depend on the actual rolling mill.

Before operating the rolling mill, it is thus necessary to determine the parameters and to adjust the model 13 of the rolling mill accordingly. The model 13 must therefore be parameterized. This will be explained below with reference to FIG. 2 explained.

That in the FIG. 2 The method illustrated is intended to provide the parameters of the equations of model 13 of FIG FIG. 1 to investigate. The procedure of FIG. 2 can be carried out independently of the operation of the rolling mill. In particular, the method can be carried out on a computing device, for example on a notebook, which operates "off-line" with respect to the rolling mill and possibly in real time.

It is now assumed that on the example provided notebook, which is necessary for the implementation of the method FIG. 2 is provided in connection with the FIG. 1 model 13 of the rolling mill is shown in the form of a program. This means that all those equations are programmed on this computing device, which are provided for the mathematical simulation of the rolling mill in real time. The parameters of these equations have an output value that can be specified, for example, when programming the equations. The initial value is yet in no way adapted to the actual rolling mill.

It is further assumed that there is a database 17 on which one or more datasets 18 of rolling passes of the actual rolling mill are stored. Under a Walzgang thereby the complete passage of a rolling stock is understood by the rolling mill. The data associated with a rolling pass record 18 thus refers only to a specific passage of the rolling stock through the rolling mill.

Each of the data sets 18 contained in the database 17 contains a multiplicity of operating variables, which may be actual values, setpoints, manipulated variables and / or other operating variables of the rolling plant. These operating variables are recorded time-dependent on the associated Walzgang, ie on the passage of time of the rolling stock through the rolling mill.

The stored operating variables are at least those operating variables of the rolling plant which are required by the model 13 as input variables and which are calculated by the model 13 as output variables. These farm sizes are in the FIG. 2 in each case referred to in their entirety as input variables EG and output variables AG of the model 13.

The database 17 may be a part of the notebook on which the model 13 of the FIG. 2 located. In this case, the data records 18 must be stored in some way during operation of the rolling mill on a storage medium and then made available to the notebook, in particular in real time. It is also possible that the database 17 is assigned to the actual rolling mill. In this case, the data records 18 can be transmitted via a communication link to the notebook provided as an example, on which the model 13 of the FIG. 2 located. This transmission of the data records 18 can take place in real time, so that the model 13 of the FIG. 2 can take place in real time. But this is not mandatory.

To determine the parameters of the equations of the model 13, the model 13 is now calculated on the basis of the output values of the parameters and the input variables EG of a first data set 17 stored in the database 17 and the associated output variables of the model 13 are determined , These output variables, which are referred to in their entirety as model output variables AG *, are compared with the output variables AG of the same first data set 17 stored in the database 18.

This is in the FIG. 2 shown by way of example on the basis of the actual value IWB calculated by the model 13, which value coincides with the measured actual value IWG stored in the database 17 associated first record 18 is compared. In the FIG. 2 this comparison is shown as block 19.

Based on the comparison by the block 19, the output values of the parameters of the model 13 are now changed. This is in the FIG. 2 indicated by the influencing variables PG, which act on the model 13 from the block 19, and thus should make clear the influence of the parameters of the model 13 as a function of the comparison carried out. The parameters of the model 13, after the explained first run of the method, no longer have the initial values, but first new values.

Thereafter, the method described above is run again, but not with the output values for the parameters of the model 13, but with the first new values of the parameters determined in the first pass of the method. In this second run of the method, the operating variables from the first data record 18 of the database are used again. On this basis, a new comparison is made by the block 19 and second new values for the parameters of the model 13 are determined.

This repetition of the method can be carried out until the respectively newly determined values for the parameters of the model 13 no longer differ significantly from the previous values.

Thereafter, the entire process can be continued with a second data record 18 from the database 17. In this case, therefore, the operating variables from the second data set 18 as well as the previously determined values for the parameters are used by the model 13 in order then to re-influence the values for the parameters via the block 19. The method can in turn be repeated several times with the second data set 18.

Thereafter, the entire process can be continued with further data records 18 from the database 17, so that the last determined parameters of the model 13 are always replaced by new parameters. This procedure can be repeated until the values of the parameters no longer change significantly.

With the now present on the exemplary provided notebook values of the parameters can then in connection with the FIG. 1 explained device for controlling and / or regulating the rolling mill, so in particular the existing in the actual rolling mill digital computing device are loaded. This is synonymous with the fact that the model 13 of the FIG. 1 based on the determined values for the parameters. After that, the basis of the FIG. 1 explained process observers are performed by the present in the actual rolling mill computing device with the now parameterized model 13.

Claims (9)

Method for determining parameters of a model (13), with which at least one area of a rolling mill can be imitated mathematically, with the following steps: the model (13) initial values for the parameters to be determined are given, the model (13) are determined by measurement within the rolling mill determined input variables (EG), of the model (13) at least one model output variable (AG ^* ) in dependence is calculated from the input variables (EG) and the initial values of the parameters, the model output variable (AG ^* ) is compared with an associated output variable (AG) determined by measurement within the rolling mill, depending on the comparison, first new values for the parameters of the Model (13) determined. Method for operating a rolling mill, in which at least one area of the rolling mill is simulated by means of a model having a parameter, with the following steps: the model (13) is given initial values for the parameters, the model (13) is determined by measurement input variables (EG) determined within the rolling mill are fed from the model (13) to at least one model output variable (AG *) as a function of the input variables (EG) and the Initial values of the parameters calculated, the model output (AG ^* ) is compared with an associated, determined by measurement within the rolling mill output (AG), depending on the comparison first new values for the parameters of the model (13) are determined. Method according to claim 1 or 2, comprising the following steps: the model (13) is supplied again with the input variables (EG), the model (13) is given at least one model output variable (AG ^* ) as a function of the input variables (EG) and the first new values of the parameters are calculated, the model output variable (AG *) is compared with the associated output variable (AG), second new values for the parameters of the model (13) are determined as a function of the comparison. The method of claim 3, wherein the input variables (EG) and the output variables (AG) form a data set (18) and are assigned to a specific Walzgang. The method of claim 4, wherein there are multiple datasets (18) associated with different rolling passes. Method according to one of claims 4 or 5, wherein the data record (s) (18) are stored on a database associated with the model (13). Method according to one of claims 4 or 5, wherein the data set (s) (18) are transmitted to the model (13). Method according to one of the preceding claims, wherein a measured actual value (IWG) of the rolling plant is compared with an associated desired value (SW) (11) and depending on this a manipulated variable (SG) is determined, wherein a calculated actual value (IWB) in dependence on the Setpoint value (SW) and depending on other operating variables of the rolling mill of the model (13) is determined, and wherein the measured actual value (IWG) with the calculated actual value (IWB) compared (15) and the comparison result as an error signal (FS) is used , Computer for carrying out the method according to claim 1, in particular notebook, on which the model (13) is programmed, on which the initial values for the parameters are stored, and on which the input variables (EG) and the output variable (AG) are stored.

Images