METHOD AND APPARATUS FOR MEASURING PROPERTIES OF PAPER WEB
The invention relates to a method for measuring properties of a paper web, in which method at least one property of the paper web is measured at least at two locations in the cross direction of the paper web simultaneously at least on two adjacent measuring channels such that the adjacent measuring channels measure at least part of the time a common band, data on common band measurements are collected, statistical calculation methods are applied to the collected data and the readings of two adjacent measuring channels are equalized. The invention further relates to an apparatus for measuring properties of a paper web, the apparatus comprising at least one measuring means having means for transmitting a measuring beam to a measuring point at least on one measuring channel, whereby the apparatus is arranged to measure at least one property of the paper web by measuring at least on two adjacent measuring channels in the cross direction of the paper web simultaneously such that the adjacent measuring channels are at least part of the time arranged to measure a common band, and which apparatus comprises means that are arranged to collect data on the measurements of the common band and to apply the statistical calculation methods to the collected data, means for comparing the readings of the common band of two adjacent measuring channels with one another and means for equalizing the readings of the adjacent measuring channels.
It is known to measure properties of a moving paper web with a measuring device such that a measuring point of a measuring sensor trav- erses in the cross direction of the paper web. The measuring sensor is generally secured to a measuring bar positioned across the paper web. It is also known to use so-called optic traversing in measuring the properties of the paper web as disclosed in US patent 5,073,712. In this method, a measuring sensor is fixedly mounted above the web and a measuring beam to be trans- mitted from the sensor traverses the web in the cross direction. Calibration of these measuring devices is carried out in such a way, for instance, that a reference sample is placed e.g. at the edge of the paper web, outside the web, and the measuring device measures the properties of said reference sample at suitable intervals, and on the basis thereof, calibrates the measuring means in a manner known per se. However, in the solution concerned, the measuring device measures the paper web diagonally, whereby measuring results will not
be obtained from adjacent locations, for instance. The measuring method is also relatively slow.
To speed up a measurement and to obtain adjacent measuring results, it is known to use solutions, in which the paper web properties are measured simultaneously at adjacent measuring points. Solution of this kind are disclosed, for instance, in US patents 4,565,444 and 4,801 ,809. Further, the publication by Pertti Puumalainen "Paperikoneen CD-mittausten tule- vaisuudennakymat, Paperirataa on-line mittaavat laitteet ja niihin liittyvat saadόt, 24.-25.2.1998, Lappeenranta" (The prospects of paper machine CD measurements, Devices measuring the paper web online and adjustments related thereto, 24 - 25 February 1998, Lappeenranta) sets forth a solution in which a plurality of measuring devices are adjacently positioned and each measuring device traverses for a portion of the paper web in the cross direction. Thus each sensor analyzes a small portion of the paper web width. How- ever, calibration is very difficult in these solutions. In the above-mentioned publication by Puumalainen, a reference sample is placed above each measuring device, and for calibration, the measuring bar, onto which the measuring devices are placed, is turned upside down such that each measuring device then measures the values of the reference sample locating in front of the measuring device concerned. However, a problem with this solution is that various reference samples are originally different or they may become different due to aging or various outside influences, such as fouling, and consequently the measuring devices are calibrated onto different levels, i.e. their readings become different. The structure of the solution in question is also very compli- cated and hence cumbersome and expensive.
The object of the present invention is to provide a method and an apparatus in which at least some of the above-mentioned drawbacks can be avoided. A further object is to provide a method and an apparatus by means of which measuring of the properties of a moving paper web is fast and the measuring results are accurate and reliable.
The method of the invention is characterized by employing one or more of the following statistical calculation methods listed: correlation method, principal component regression, partial least squares regression, canonical correlation analysis. Further, the apparatus of the invention is characterized in that the means which are arranged to apply the statistical calculation methods to the
collected data are arranged to employ one or more of the following methods listed: correlation method, principal component regression, partial least squares regression, canonical correlation analysis.
The basic idea of the invention is that at least one property of a pa- per web is measured at least at two locations in the cross direction of the paper web simultaneously at least on two adjacent measuring channels such that the adjacent measuring channels measure at least part of the time a common band overlappingly. Data on the measurements of the common band of either measuring channel is collected and the statistical calculation methods are ap- plied to the collected data and the measuring channel readings are equalized, whereby the equalization of readings may comprise a transfer of calibration from one adjacent measuring channel to another if the measuring results are within the limits set on the basis of permanent or collected statistics, or it may comprise using the solution for fault diagnostics if the measuring results are outside said limits. Furthermore, the readings of the common band of two adjacent measuring channels are compared with each other. The idea of one preferred embodiment is to compare data on the common band of the adjacent measuring channels with one another and if the measuring results differ, it is possible to detect a fault in either one of the measuring channels and by comparing with the data history collected from the common band it is possible to conclude which one of the measuring channels has the fault. The idea of another preferred embodiment is to employ at least three adjacent measuring channels, whereby measuring is performed on at least two common bands, and the solution is applied to fault diagnostics such that if the measuring result of one measuring channel differs from the result of the other two, it is possible to detect a fault in the differing measuring channel. The idea of a third preferred embodiment is that at least one measuring channel is calibrated and the calibration is transferred to adjacent measuring channels utilizing common bands. The invention has an advantage that fault diagnostics of the apparatus can be implemented effectively and also measuring points of various measuring locations can be effectively calibrated by a simple mechanical solution, when necessary. The solution is extremely reliable and improves considerably the reliability and usability of the measurements. In the present specification, the term 'paper' refers to paper board and tissue paper, in addition to paper.
In the present specification, calibration refers to defining a quantity for a property of paper that is actually measured (temperature, etc.) The means for measuring the property must be properly calibrated so as to indicate the correct value of a stimulus measured. Thus all calibrated measuring means of the same type indicate the same measured value for the same measured stimulus.
However, several gauges are constructed such that a second property that has not been measured directly will be inferred on the basis of a first property by utilizing the correlation between the properties. For instance, sev- eral gauges used in the paper industry direct the stimulus, such as radiant energy or a particle beam, at the paper whose properties are measured and then measure the modulated radiant flux or particle flux emitted by the paper. The mathematical relation describing the correlation between the properties obtained by these measurements is used for calculating the second property on the basis of the first property. The formula and parameters of this relation have to be previously known or predetermined. In some cases, the second property can be inferred on the basis of several measured properties by using a multi- variable relation.
Because the strength or other properties of the source of stimulus may vary in various gauges or even in the same gauge with time, the correlation between the calibrated measurement and the property correlating therewith may vary in different gauges and at different times in the same gauge. Likewise, the correlation between the measured property and the correlating property may vary due to changes in other unmeasured properties. For in- stance, the correlation between microwave backscatter and sample moisture content changes if the sample contains carbon black.
Standardization is used for compensating for differences and changes in the stimulus or correlation. Standardization is also a means of calibration when the differences appearing in the stimulus or correlation are known or they are known to be insignificant.
In standardization, a property of at least one reference sample, whose one other property is known, is measured and a parameter representing the relation between the known property and the measured property is calculated for measuring means. A plurality of reference samples, one other property of each being known, are preferably used. If a plurality of reference samples, whose other known properties have different values, are used, it is
possible to calculate a plurality of parameters for the relation. Thus, statistical methods, such as the method of least squares, can be used for calculating the most suitable parameters. By means of statistical methods, it is also possible to select the relation formula, in addition to other parameters. Calibration is thus applied to properties that are measured directly and in connection wherewith a drift in the calibration of the measuring means is corrected. These properties include e.g. temperature and thickness of paper. Standardization is applied to properties that are measured indirectly and in connection wherewith one or more of the following features are simultane- ously compensated for: i) differences in correlation between measured and inferred properties, ii) differences appearing in the stimulus used, iii) drift in the calibration of the measuring means. These properties include e.g. basis weight, moisture, ash content, colour, etc. For the sake of clarity, in the present specification, the term 'calibration' also refers to standardization, in addi- tion to calibration.
The invention will be described in greater detail in the attached drawings, wherein
Figure 1 is a schematic view of a measuring apparatus in accordance with the invention seen from the machine direction of the paper web, Figure 2 is a schematic view of measuring paths of the measuring apparatus in Figure 1 ,
Figure 3a and 3b are schematic histrograms of the measuring results of two different sensors on their common band,
Figure 4 shows schematically an application example of a method of statistical mathematics,
Figure 5 is a schematic view of an additional calibration solution, and
Figure 6 is a schematic view of another measuring apparatus in accordance with the invention seen from the machine direction of the paper web. Figure 1 shows a measuring bar 1 which is installed across a paper web and onto which measuring means, i.e. sensors 2a to 2d, are secured. The sensors 2a to 2d measure properties of the paper web 3 simultaneously at adjacent locations such that the adjacent sensors 2a to 2d measure the same web property simultaneously. Thus, data on the properties of the paper web 3 is obtained quickly and on a large area. In the case of Figure 1 , the sensors 2a to 2d comprise both a transmitter and a receiver, whereby measuring is ef-
fected as a reflection measurement. If desired, the transmitter and the received can be placed on the opposite sides of the paper web 3, and then measuring is effected as a transmission measurement in a manner known per se. At simplest, each sensor comprises one measuring channel, but the sen- sor may also comprise a plurality of measuring channels, for instance, in such a way that, in the sensor which measures spectrum, each different channel can measure a different wavelength of the same spectrum. Each measuring channel can also measure a specific spectrum. Different measuring channels of one sensor can measure simultaneously or successively by means of time multiplexing, for instance. The edge reference samples can also be located elsewhere most of the time and they will be moved to positions shown in the figures only for the duration of calibration.
The sensors 2a to 2d are arranged to traverse a portion of the width of the paper web 3 as indicated by arrows A. Measuring data are thus ob- tained simultaneously from several adjacent locations of the paper web 3, and moreover, measurements can be carried out alternately at every location across the paper web. This so-called minitraversing has an advantage that not very many adjacent measuring channels are needed, but the paper web 3 can be measured considerably more accurately than with the commonly used one sensor that traverses the entire web. The measuring rate also increases, since in minitraversing one reciprocating movement cycle may be about one second, for instance.
Figure 2 shows measuring paths 11a to 11d of the adjacent measuring channels. Thus there may be one or more measuring channels in each sensor 2a to 2d, and Figure 2 illustrates a case where each sensor 2a to 2d comprises one measuring channel. The adjacent measuring paths are arranged such that they have a common band c, i.e. that their measuring areas partly overlap as shown in Figure 2. History data on the measuring results of the common bands c are collected and methods of statistical mathematics are applied thereto. By means of the measurements of the common channels c it is possible to equalize the calibrations of the measuring channels.
The common band c can be further divided into two or more measuring zones cx and cy, when different measuring channels measure on the common band c on a plurality of common measuring zones cx and cy. Figure 3a shows a histrogram of measuring results of the measuring channel proceeding along the measuring path 11a for the common band cx and, in a cor-
responding manner, Figure 3b shows a histrogram of measuring results of the measuring channel proceeding along the measuring path 11b for the same common channel cx. Averages χ (cx11a) and (CX11„) of the measuring results are defined from the histrogram. Further, dispersion σ(cx11a) of measuring re- suits of the measuring channel proceeding along the measuring path 11a and dispersion σ(cx11b) of the measuring channel proceeding along the measuring path 11 b are also defined. The measuring result averages of the measuring channels can be equalized in a relatively simple manner and the responses of the measuring results can be corrected on the basis of the dispersion. Equali- zation of calibration can be presented simply by means of the following formula x(cx1ia) = αx(cx11b)+β, where x(cx11a) is the measuring result of the measuring channel proceeding along the measuring path 11a, x(cx11b) is the measuring result of the measuring channel proceeding along the measuring path 11b, α is a slope, i.e. response correction, and β is an offset correction. After correction, the average x (cxl1a) of the measuring channel proceeding along the measuring path 11a can be set equal with the average x (cx11b) of the measuring channel proceeding along the measuring path 11b and the dispersion σ(cx11a) of the measuring results of the measuring channel proceeding along the measuring path 11a can be set equal with the dispersion σ(cx11b) of the measuring results of the measuring channel proceeding along the measuring path 11b.
For calibration equalization, limits can be set to the slope α and the offset correction β, which limits they are not allowed to pass in a normal situation. Of the statistical calculation methods, calculating a long-term average, for instance, allows to eliminate errors, if any, caused by short-term noise in the measuring result. Naturally, other signal processing methods known per se can be used as well. The example given above in connection with Figures 3a and 3b is a simple application example of the solution, but naturally the solution can be applied in a corresponding manner in e.g. multivariable applica- tions, such as spectrum measuring, utilizing calculating methods known per se. Suitable statistical methods include e.g. principal component regression
(PCR), partial least squares regression (PLS), canonical correlation analysis and other corresponding methods, as set forth, for instance, in Statistical Factor Analysis and Related Methods, Basilevsky A, published by Wiley, 1994. Yet another example of statistical methods is to use correlation methods. Fig- ure 4 illustrates measuring points 8 between the measuring channels proceeding along the measuring paths 11a and 11b on a common measuring zone cx. In Figure 2, the measuring location of the measuring channel proceeding along the measuring path 11a on the common measuring zone cx is indicated by cx11a and, correspondingly, the measuring location of the measur- ing channel proceeding along the measuring path 11b on the common measuring zone cx is indicated by cx11b. In Figure 4, the measuring points 8 represent the value of the measuring location cx11a when the value of the measuring point cx11b is known, and vice versa. cx11a and cx11b may also be averages of measuring groups. The measuring points 8 allow to define a correlation curve 9, which can be used for equalizing the calibrations.
The invention can also be used for fault diagnostics such that readings of adjacent measuring channels, or if the sensor comprises only one measuring channel, sensors on the common band c are compared with each other. The readings are also compared with the history data of the common band c. If the readings of the adjacent sensors then differ from each other and the reading of one sensor corresponds to the history data reading of the common band but the reading of the other sensor does not, it is possible to conclude that there is a fault in the sensor whose reading does not correspond to the history data of the common band c. It is possible to set specific limits to how much the sensor readings are allowed to deviate without a need for fault diagnostic inspections. The limits can be permanent or they can be dynamically adjusted or changed by means of the history of collected measuring results. For instance, if the basis weight value of paper indicated by the sensor is 10 to 120 g/m2 and the sensor adjacent thereto indicates the value 80 to 110 g/m2 and the history data indicates that the basis weight is 80 to 110 g/m2, it can be stated that the first sensor is out of order.
If there are at least three adjacent sensor such that they have at least two common bands c, the solution can be used for fault diagnostics, for instance, such that if the readings of the first sensor 2a and the second sensor 2b differ from one another, the reading of the second sensor 2b will be com-
pared with the reading of the third sensor 2c on the common band c of the second sensor 2b and the third sensor 2c. If the readings of the second sensor 2b and the third sensor 2c correspond, it can be concluded that the first sensor 2a is faulty. But if the readings of the second sensor 2b and the third sensor 2c do not correspond, it can be concluded that the second sensor 2b is faulty.
In the apparatus, edge reference samples 5 can be arranged at one or either edge of the paper web 3 for calibration. The measuring paths 11a to 11d of the adjacent sensors 2a to 2d are arranged such that they have a common band c, i.e. that their measuring areas partly overlap. In that case, calibration is carried out such that the outermost sensors 2a and 2d are calibrated by means of the edge reference samples 5. Thereafter is measured the reading of the sensor 2a on the common band c and the adjacent sensor 2b measures on the same band, whereby the adjacent sensor is calibrated by comparing the measuring results of the sensors. Said cycle is repeated on a next adjacent sensor as many times as necessary. This kind of calibrating measurement is preferably repeated several times and in succession, whereby it is possible to compensate for errors that result from the adjacent sensors not measuring exactly the same location in the machine direction of the paper web 3 during the paper making process. The outermost sensors 2a and 2d need not necessarily travel over the edge reference samples 5 at other times than in calibration situations. Further, the measurements need not necessarily overlap at other times than in calibration. Furthermore, in measuring the sensors 2a to 2d can be mainly stationary, whereby they would be moved back and forth only in calibration. The edge reference samples 5 can be utilized for the cali- bration of the apparatus during the paper making process both when using reflection measurement and transmission measurement.
The apparatus may also comprise a reference sample 4 which is movable along a path indicated by a broken line B in the cross direction of the paper web 3 across the beams measuring different measuring points. Addi- tional calibration of said apparatus is thus carried out such that the reference sample 4 is moved through the measuring beam of each sensor 2a to 2d, and each sensor 2a to 2d is calibrated at the moment when the reference sample 4 coincides with the measuring beam. Each sensor 2a to 2d is then calibrated by the same reference sample 4, whereby their readings are made equivalent in a simple manner. Calibration is carried out such that the reference sample 4 having a given content or value for a measurable property is measured. If the
reading of the sensor 2a to 2d differs from this value, it is adjusted so that the sensor 2a to 2d shows the correct value. If the reference sample 4 is shifted above the paper web 3 as shown in Figure 1 , i.e. on a different level than the web, the calibration can be carried out any time when needed, also during the paper making process. In that case, the measurements represent the common effect of the web and the reference sample, whereby these measurements, together with the measurements performed on the paper web alone, can be used for calibrating the sensors. Advantageously, when using the reference samples together with the paper web 3, said reference samples cause changes in measurements that exceed the expected changes in the paper web measurements during the measuring. If it is desired that the reference sample 4 travels on the same level with the paper web 3, the calibration has to be performed during a web break, when there is no paper web at all at the measuring location. Likewise, when using transmission measurement, the calibration has to be performed during a web break.
The reference sample 4 can be arranged to be movable in a manner shown in Figure 5. In the case of Figure 5, in a measuring situation, a measuring beam transmitted from a radiation source 10 propagates through a measuring window 6 of the sensor 2a to the paper web 3 as depicted by a broken arrow c. In a calibration situation, it is possible to divert the measuring beam with a means, such as a mirror 7, to control radiation to propagate along the arrow D and to hit the movable reference sample 4. The paper web moves in the direction of the arrow E and the reference sample 4 is shifted in the transverse direction with respect thereto. The mirror 7 and the reference sam- pie 4 are arranged such that the distance proceeded by the measuring beam will not change, i.e. the optical distance between the mirror 7 and the paper web 3 is the same as the optical distance between the mirror 7 and the reference sample 4. The solution of Figure 5 can be applied to measurements utilizing radiation, for instance optical or other electromagnetic measurements. This solution has an advantage that it can also be used during the paper making process, and nevertheless, no distance compensation is needed in the calibration. Figure 5 depicts only the transmitted beam, but for instance in reflection measurement, the measuring beam preferably returns substantially along the same path as the transmitted beam. When the angle between the measuring beam c and the paper web 3 deviates from 90 °, the effect of the mirror reflection on the measuring result can be eliminated. If desired, the ref-
erence sample 4 can also be positioned inside the sensor 2a, i.e. in the same housing with the measuring beam transmitter, and thus the reference sample 4 is protected from outside influences, for instance, fouling. On the other hand, a simple structural solution is to position the reference sample 4, in accor- dance with Figure 5, outside the measuring bar 1 , whereby the beam indicated by the arrow D propagates through a side window 6a. Instead of using a means for controlling the beam, the distance proceeded by the measuring beam can be retained equal in an ordinary measuring situation and in calibration, for instance, by turning the beam-transmitting sensor by turning the measuring bar 1 , for instance.
Figure 6 shows a measuring arrangement operating on a transmission measurement principle. The sensors 2a to 2d transmit the measuring beams towards the paper web, and having passed through the web the modified beams arrive in sensors, in this case in a detector 2e to 2h. The sensors 2e to 2h are secured to the measuring bar 1', substantially at a corresponding location with the sensors 2a to 2d. The sensors 2e to 2h can also be moved a part of the width of the paper web 3 in the cross direction thereof. Thus the sensors 2a to 2d and 2e to 2h move substantially at the same time and at the same location. The reference beam 4 is shifted across the path of the meas- uring beams at the same level where the paper web 3 normally travels. Hence, the reference sample need not be compensated for distance, but calibration is simple to perform.
The reference sample 4 and the edge reference samples 5 are reference material with known properties. Further, the reference samples can consist of a plurality of different reference samples, when they have different reference sample sections for different proportions of the same property, for instance, for basis weights and other properties. Thus, when calibration is performed, from the reference sample is selected the section whose properties are closest to the properties of the paper web 3 to be measured, for instance, from the different sections are selected the one whose basis weight is closest to the basis weight of the paper web 3 to be measured. Absolute calibration can then be implemented. If the reference samples consist of a plurality of different sections, the properties of the different sections vary on an equally wide range as or wider range than the expected variation in a corresponding property of the paper web, whereby calibration for variation sensibility can be implemented. The reference sample may also have different reference sample
sections for the calibration of different paper web properties, such as moisture and ash content measurements. Different sections of the movable reference sample 4 can be moved either with one traversing means or the sections can be divided such that they are moved with a plurality of traversing means. The reference sample can be e.g. a transmission reference, an absorption reference or a reflection reference, when the reflection can be a mirror reflection or a diffusion reflection.
The reference sample 4 is so minuscule that in the calibration situation it covers the path of the measuring beam of only one sensor or some sen- sors 2a to 2d, for instance. Thus, when calibrating by means of the movable reference sample 4, only some of the sensors are excluded from measurement, while the others continue to measure in an ordinary manner. Typically, the measuring devices are currently calibrated once in an hour, for instance. If necessary, the solution of the invention allows more frequent calibration, since the calibration is quick to carry out and it disturbs the ordinary measurements only for a relatively short period of time.
The adjacent measurements according to the invention can be implemented by using either a plurality of adjacent sensors or by measuring a plurality of different measuring points with one sensor, employing for instance one sensor that measures on a plurality of measuring channels simultaneously as described in US patent 4,565,444, for instance. Further, in addition to mechanical traversing, measuring areas of adjacent sensors can be arranged to overlap by means of optical multiplexing, for instance.
The drawing and the related description are only intended to illus- trate the inventive idea, and the details of the invention may vary within the scope of the claims.