FI89632C - FOERFARANDE FOER BILDNING AV EN SIGNAL - Google Patents

Description

89 632

METHOD OF SIGNAL GENERATION - FÖRFARANDE FÖR BILDNING AV EN SIGNAL

The invention relates to a method for generating a signal characterizing the transverse basis weight of a moving web, in which method the basis weight of the web is measured by a traversing basis weight sensor and a signal characterizing the basis weight of the web is generated. In particular, the invention relates to the measurement of the basis weight of a web made of paper, cardboard-10 or similar pulp-based material.

Today, moving paper webs in paper machines are usually detected by a traverse square meter. By traversing sensor is meant a sensor that continuously measures the basis weight or something proportional to it so that the measuring point continuously moves back and forth from edge to edge of the web. The signal obtained in this case is a combination of machine mouth-1, i.e. longitudinal and transverse 20 variations. These proportions cannot be distinguished from each other by processing the signal alone. The resulting signal is used to adjust both machine direction and transverse variation. As a result, the signal from the traversing basis mass sensor has to be processed, in particular to determine the basis mass variations in the transverse direction of the web.

To solve the problem, it is known to calculate the average of several traversions, i.e. determining the basis weight variation from the plurality of transversions, and calculating the average or average transverse basis weight variation; the value obtained is used as the transverse basis weight variation of the web. Measuring the basis weight variation in this way is time consuming due to the typically 10 to 20 measurements of multiple traver-35 tones required by the method.

It is further known to use, in addition to the traversing basis mass sensor, a stationary basis mass sensor, in which case the signal of the stationary basis mass sensor is subtracted from the signal given by the traversing sensor. This method is expensive due to the number of sensors it requires and not very reliable, 5 because the residual variation interferes with the measurement.

It is an object of the present invention to obviate the above drawbacks. In particular, it is an object of the invention to provide a new method for generating a characteristic signal for the transverse basis mass of a continuous moving web using the output signal of a traversing basis weight sensor.

For aspects of the invention, reference is made to claim 1.

The invention is based on the basic idea that the reference signal corresponding to the variation of the basis weight of the web in the machine direction as an average of the whole web is subtracted from the signal generated by the transverse basis measuring sensor. The reference signal is proportional to the basis weight in the short time period, but the dependence can change very strongly in the long run. The reference signal can be, for example, a signal characterizing the machine-average web transmittance of the web of the troubleshooting system, which can be easily obtained from devices already present in the paper machine. Alternatively, a measuring signal characterizing, for example, the vacuum of the suction roll of a paper machine can be used, which can advantageously be processed to eliminate interference. Alternatively, 30 two reference signals for headbox pressure and consistency can be used.

In connection with signal processing, the so-called the principle of adaptive noise reduction. In this case, the reference signal is preferably fed after an delay of 35 to an adaptive filter, where it is formed into a so-called delay line and linear combiner or so-called. by means of a neural network, a new signal is 3,09632 li, which is subtracted from the traversing signal to form an output signal. The output signal is preferably applied to the adaptive filter as an error value to adapt the coefficient of the linear combiner.

In a preferred embodiment of the invention, it is assumed that the machine and transverse basis weight variations are uncorrelated, as is natural. There is assumed to be a connection between the reference signal and the machine direction signal

OF

mMD (t) = ^ ai Xref (t-Ät-JÄt) i * l 10

In the formula m ^ (t) = machine direction basis weight as a function of time; N is a time constant determined by the correlation time between the travel difference between the basis weight sensor and the reference sensor, the web speed variations and the machine direction basis weight variation; and a are coefficients. xref (t) is the value of the reference sensor signal at time t. At is the delay of the signal xref (t) to be applied, which is determined by the sensor travel difference and the average web speed. 6t is the sampling interval for digital signal processing. The coefficients a ± depend slowly on time, e.g. it can be assumed to be constant for 5-10 traversals. The coefficients aA are adapted for slow variations according to eg LMS or SER algorithm: (ηδ t) '

An - ^ n + l “+ Γ1 y N

ä „(nbt) 25 where μ is the factor that influences the adaptation 4 39632

Xzef (nb t - At - bt)

Xzef {nbt - At - 2δt) κ

Xn =; en = mtz (nb t) - £) ai (nbt) Xre i = 1

Xref (jib t - At - Nbt) for speed and accuracy, and mtr (n6t) is the traversing probe signal at time nöt.

In this case, the received signal en, i.e. the output of the system gives a pure transverse signal.

5 Further, the system adapts to the slow changes in the connection between the reference and the machine direction basis weight signal. Furthermore, with the correct choices of the parameters At and N, the system adapts to the changes in the delay between the reference and the basis weight signal 10 caused by the machine speed.

The dependence between the machine direction basis mass and the reference can also be assumed to be nonlinear

OF

mhd = fiY, xzt) fu - Ht - ibt)), ‘. i = l 15 where f is a monotonic function. In this case, the coefficients are adapted to the so-called neural network backpropagation method.

Thanks to the invention, a reliable value 20 of the basis mass profile of the moving web in question can be formed from the signal generated by the traversing basis mass sensor in a considerably shorter time than with the methods currently used, i. when calculating the average basis weight profile of multiple traverses.

The invention will now be described in detail by means of an exemplary embodiment with reference to Figure 1, 25, which shows the implementation of the method according to the invention in practice.

In Fig. 1, the signal S1 obtained from the traverse square mass sensor 1 is passed to the summing unit 2 and 89632 5 to output 3. The signal from the reference sensor 4, e.g. also the signal applied to the output 3 as an error signal. The signal S2 obtained from the adaptive filter is passed to the summing unit 3 to be subtracted from the signal S1. The output signal corresponds very closely to the transverse ne-10 adhesive profile of the web.

The exemplary embodiment is intended to illustrate the invention only, and embodiments of the invention may vary within the scope of the appended claims.

15

Claims (8)

A method for forming a signal characterizing the transverse surface weight of a moving path, wherein the method's surface weight is measured with a traversing surface weight detector and the signal characterizing the path weight of the path is formed, characterized in that a reference signal is subtracted from the resulting traversing signal. The surface weight of the web in the machine direction varies with variation as an average value for the entire web. 2. A method according to claim 1, characterized in that from the transverse signal a reference signal is characterized, characterized by the average light transmittance of the path in the machine direction. Method according to claim 1 or 2, characterized in that from the transverse signal, a mean transmittance given by a pour detector is subtracted or a reference signal characterizing a negative pressure is subtracted. Method according to any one of claims 1 to 3, characterized in that a reference signal characterizing the surface weight of the web in the machine direction is shown to vary according to the delay measured in an adaptive filter, and is thereby formed by means of a so-called. delay line is a signal which is subtracted from the traversing signal, whereby an output signal is obtained. 30 5. A method according to claim 4, characterized in that the output signal is measured to the adaptive filter as an error value. Method according to any one of claims 1 to 5, characterized in that the variations in the path in the machine and transverse direction are assumed to be non-correlating and between the reference signal and the signal in the machine direction it is assumed to be one in 89632 relations etc. (t) = £ a in xref (t-At-16t) in which the coefficients aA change slowly as a function of time. 7. A method according to any of claims 1 to 10, characterized in that between the surface weight and the reference signal is assumed to be the relation w MMD = ί (Σ a in Xzefit - ΔίΓ “iöt)), i-l 15 where f is a monotonous function.