Technological background The invention relates to the calendering of a paper web by means of a calender comprising a profile-adjustable device for cross-web directional control of the paper properties, such as the thickness.

In calendering, the paper is driven through a nip formed of rolls rotating against each other. The thickest points of the paper are compressed in the nip. This yields a smoother paper surface, which also is better and more homogeneously glazed.

Current calendering techniques also use so-called profile-adjustable multi-zone rolls, which compensate for the roll bend in the cross-web direction. Such a roll comprises loading shoes, which are pressed radially against the roll mantle within the roll, each of the loading shoes having individually adjustable pressure. This compensates for the profile deformation caused by the roll bend, so that the nip opening becomes equal over the entire web width.

General description of the invention A control method as defined in claim 1 has now been invented. A number of preferred embodiments of the invention are described in the other claims.

The invention allows high-precision overall control of a variable cross-directional (CD) web property, such as the thickness, in any production situation. The model- based control method of the invention is suitable both for machine, soft and multi- roll calenders. A multi-roll calender may be a calender with separately loaded rolls (Valmet OptiLoad).

The paper quality variable to be controlled may be for instance the thickness, glaze or smoothness. More than one variable can naturally be controlled at the same time.

The profiling device may be especially a variable calender crown roll in which the bend is compensated for. It includes one or more profiling means, particularly radial pistons, which serve to compensate for the roll bend. However, the profiling device may also be a temperature control means, such as a heater. The heater may be for

instance an induction heater, possibly connected with a thermo-roll, or a heat blower.

Drawings The accompanying drawings relate to the detailed description of the invention. In the drawings -figure 1 shows a paper compression curve and a cross-directional, thickness profile by zones, -figure 2 shows on-line identification of the parameters of the thickness variation model, -figure 3 is a block diagram of a cross-directional controller.

Detailed description of the invention In accordance with the invention, a profiling device is used in calendering, which allows cross-directional control of a desired paper quality quantity. The calender may comprise one or more nips.

The profiling device may be a profile-adjustable roll. It comprises at least one zone, where the roll includes a profiling means enabling a radially adjustable force to be exerted on the roll mantle from the inside towards the outside. Usually a plurality of such profiling means is provided, for instance at about 10-20 cm intervals. The profiling means are individually adjustable. Usually the profiling means are loading shoes, which are pressed against the roll mantle surface. They are usually hydraulically actuated means.

The invention comprises in calendering the control of a variable quantity, for which a specific value is sought and which can be affected with the profiling device. The quantity may be especially the paper thickness, glaze, smoothness or moisture. A model predetermined is describing the effect of the force of the profiling means on the quantity to be monitored. After calendering, this quantity is measured in the profile-adjustable zone of the web, and if the quantity differs from the set value, the change required in the force of the control means for compensating for the difference is computed using a model. With a separate control of each profiling means, the quantity will be as constant as possible also in the cross-web direction.

Usually the quantity is measured also before the calender, and then the quantity

change in the calender is computed with the model. The variable to be measured is most preferably the paper thickness.

The model may be a mathematical model or a functional presentation taking account of the process conditions. Under normal production conditions, when no changes occur in the machine speed and any other process changes are dynamically slow, feedback regulation can be used.

The mathematical model is preferably such that seeks to take account of the effects of all significant variables. The unknown parameters included in the model are determined by test results, e. g. with a recursive algorithm, such as a Kalman filter or the minimum square sum method. The parameters can be updated during the production conditions. The updating can be performed according to paper grades or production sites. The updating can be done with an on-line algorithm or on the basis of laboratory measurements. The parameters can be stored in tables by paper grades. Paper grade data can also be used for determining the basic constant and the response model for the roll model of the profile-adjustable roll.

In a multi-nip calender, a mathematical model can be used for determining the change occurring in the paper in each nip. By combining the equation models of the different nips a catenated model is obtained for the total system. In this conjunction, the initial value of the parameter change computed in the preceding nip is thus used as input data in the consecutive nip. In a calender comprising several controllable rolls, the model can thus be used in a concatenate manner to describe the total process when the set values for linear load profiles of adjustable rolls are computed for each nip.

The model may comprise parameters depending exclusively on the properties of the paper to be calendered, or parameters depending on the calender. The model coefficients can be determined for instance experimentally for each paper grade, after which the equation model is applicable to any calender. The calender-specific coefficients are determined separately for each calender.

The model can be formulated with a desired shape, e. g. linear, relative to unknown parameters. Then various off-line or on-line identifying methods can be directly implemented for determining the unknown model parameters. The minimum square sum method or the Kalman filter are for instance suitable for this purpose.

One model describing relative paper compression e in a calender nip is the "Crotogino equation" (cf. e. g. Crotogino et al., TAPPI Proceedings, 1988 Finishing and Converting Conference, 49-57) e=(Bj-Bt)/Bj=A+ juBj.

F is a so-called nip intensity factor, more precisely ß = ao + aLlogL-aslogS + aRlogR-ae6 + aMM and consequently Bi is the initial bulk (cm3/g) Bf is the final bulk (cm3/g) L is the nip load (kN/m) S is the machine speed (m/s) R is the equivalent roll radius (m) 6 is the web temperature in the nip (°C) M is the web moisture (%).

This equation, for instance, can be used as a model in the control method of the invention.

The model above comprises a number of parameters that depend only and exclusively of the properties of the paper to be calendered. The coefficients of the model can be determined experimentally, for instance, for each paper grade separately. Since the equation model has no term that takes account of properties such as the elasticity of a soft roll, the coefficients are determined separately for each calender.

The most significant control quantities conventionally used to act on the thickness variation are the linear load and the chilled roll temperature. The path speed and the roll radiuses affect the nip period. It is usually desirable to keep the machine speed constant, with the exception of controlled speed variations. For instance, by excluding the roll radius variation, the number of unknown parameters can be reduced under control in the equation model, if desired.

The unknown parameters in the model can be determined with various off-line or on-line identifying methods, such as for instance the recursive minimum square sum method or the Kalman filter.

Instead of a mathematic model, one can use functions that have been obtained experimentally for the paper grade to be calendered and take account of the effect of the process conditions. Such functions are available e. g. from a quality control system or from laboratory test results. These functions can be expressed e. g. as compression graphs.

The compression curve of a given paper grade can be experimentally determined under specific process conditions (linear load, roll surface temperature, path humidity, path temperature, machine speed). The curve can be determined by searching a functional adaptation of measurement points with off-line methods (e. g. the minimum square sum method), or recursive on-line methods can be used for determining unknown parameters in the compression curve. When on-line methods are used, thickness values corresponding to different linear load points and the curve shape are determined according to a separate program sequence, the linear load level being automatically altered.

A model or an experimentally obtained function, as well as any recursive updating of the model, can be computed in an automation system, for instance, or in a separate computing unit directly connected to this.

Figure 1 shows to the right an example of a compression curve describing the paper thickness as a function of the linear load. When the linear load is changed A, kN/m, the paper thickness changes Al Am. Accordingly, when a change Al ßm is desired in the paper thickness, a change Al kN/m in the linear load has to be made on the corresponding linear load level. To the left, figure 1 shows a cross-directional differential thickness profile in various cross-directional zones N. The linear load is changed optimally in each zone with a cross-directional regulator as determined by the compression curve.

From the difference between the determined thickness profile and the target profile, the cross-directional controller generates a control signal, which takes account of the nip/profile-specific compression. Using the models, such set values are computed for each zone of the profile-adjustable roll that the cross-directional target profile is achieved. The unknown parameters of the model or the compression

curves have been determined for the paper grade to be driven, or corresponding data have been retrieved from a database. A zone-related change Ai kN/m required in the linear load corresponding to the change Ai ; jam required in the thickness of the differential profile is computed for each zone i of the profile-adjustable roll.

When the total required thickness change is distributed on several rolls (for instance 50%/50% over two rolls), the control of each roll is performed on the basis of the linear load profile values prevailing in each roll. These take account of the basic linear load (machine-directional load) and the prevailing cross-directional load in each zone.

Figure 2 shows the basic principle of on-line identification when implemented in a double-nip soft calender. The calenders comprise a profile-adjustable soft roll 1 and a heatable chilled roll 2. The thickness profile 4 of the web being driven is monitored on a transverse measuring beam 3. A suitable equation model is used for the control, e. g. the Crotogino equation. The correctness of the identified parameters are checked by computing estimates for the linear load and, e. g., the surface temperature of the chilled roll, using the estimated parameters and the equation model. By comparing e. g. the current value of the linear load and the estimate of the linear load computed with the model system, one can conclude whether the obtained parameter estimates are acceptable. The parameter values acceptable for each paper grade and process production unit are stored in a separate database or in grade formulas in the automation system, from where they can be retrieved whenever this particular grade is produced.

Figure 3 shows a block diagram of the design of a cross-directional regulator. In this cross-directional regulator operating on the feedback principle, a differential profile is formed from the measured thickness profile and the target profile, the differential profile being conducted to the controller/computing block. The cross-directional controller/computing block determines the profile control requests (kN/m) to be transmitted to the optimising computation of the profile-adjustable roll. The cross- directional regulator/computing block takes account of the current process condition by means of the paper compression model. The data obtained from the compression model is connected to the zone-related Ai kN/m controls, either by computation, or by determining the optimal values for the set parameters of the controller (amplification K, integration period Ti and derivation period Td) at different production points (linear load level), using the so-called amplification tabulation method or any other method. The zone-related set parameters (K, Ti, Td) of the

controller are changed so as to minimise the differential profile by one single control, the regulation being notably accelerated. The regulations that minimise the differential profile are thus predicted.

The compression model or the compression curves are determined with a separate on-line or off-line method in an automation system or in a separate computing unit directly connected with this. The data obtained from the model are also used in the optimation of the profile-adjustable roll, so that the basic constant of the roll model is corrected with compression data relating to each production point and paper grade. The basic constant values affect the response model shape either by sharpening or equalising the peak. The basic constant can be corrected either with off-line or on-line methods.