NL2028386B1

NL2028386B1 - Method of Measuring a Lateral Position of an Endless Belt

Info

Publication number: NL2028386B1
Application number: NL2028386A
Authority: NL
Inventors: H Koekebbakker Sjirk
Original assignee: Canon Production Printing Holding Bv
Priority date: 2021-06-04
Filing date: 2021-06-04
Publication date: 2022-12-15
Also published as: DE102022113240A1

Abstract

A method of measuring a lateral position of and endless belt (12), comprising: - measuring lateral positions of features of the belt (12) that are distributed along the length of the belt, and - deriving the lateral position of the belt from the measured positions of the features and from correction values stored in a correction table (24), characterized in that, at least in an initial phase of operation of the belt (12), the correction table (24) is continuously updated by subjecting the measured positions of each feature to a filter procedure that causes the correction values to converge towards a limit that reflects a true position of each feature.

Description

Method of Measuring a Lateral Position of an Endless Belt The invention relates to a method of measuring a lateral position of an endless belt comprising: - measuring lateral positions of features of the belt that are distributed along the length of the belt, and - deriving the lateral position of the belt from the measured positions of the features and from correction values stored in a correction table.

When a belt conveyor is used for example for conveying media sheets past a print head in an ink jet printer, it is generally desired to be able to measure the lateral position of the endless conveyor belt with high accuracy in order to be able to feedback-control the lateral position of the belt and thereby to ensure exact registry between the media sheets and the print head.

WO 2018097717 A2 and US 2007144871 A1 disclose methods of the type indicated above, wherein said features of the belt may be constituted by segments of one or both lateral edges of the belt or by markings that have been formed on the belt. The correction table is used for correcting position errors of the features.

It is possible to define a position of an ideal centre line of the belt, e.g. by integrating the mean value of the positions of the left and right edges of the belt over the entire length of the belt. However, it is difficult to measure the actual positions of the features relative to this center line, because the lateral position of the belt {its center line) is not accurately known at the time when the positions of the features are measured nor at the time when the features are formed on the surface of the belt.

It is therefore an object of the invention to provide a method by which the lateral position of the belt can be measured with improved accuracy.

In order to achieve this object, the method according to the invention is characterized by that, at least in an initial phase of operation of the belt, the correction table is continuously updated by subjecting the measured positions of each feature to a filter procedure that causes the correction values to converge towards a limit that reflects the true position of each feature.

If the correction table is updated in this way during several revolutions of the belt, the fluctuations in the true lateral position of the belt will averaged-out, so that, even though the true position of the belt is not known at the time of each measurement, the sequence of correction values will gradually approximate the true positions of the features with ever increasing accuracy.

More specific optional features of the invention are indicated in the dependent claims.

The filter procedure that is applied to the measured positions of the features may be a sliding-average filter process of the type di,k = a0*Ti + a1*di,k-1 + a2*di,k-2 + … + an*di,k-n wherein Ti is the latest measurement result for the position of the i-th feature, di,k is the k-th update of the correction value di for the i-th feature, di,k-1 is the {(k-1)-st correction value, and so on, and a0, ...,an are weight factors that sum up to 1. Thus, in the simplest case (n = 2), the filter will be given by the formula: di,new = a*Ti + (1-a)di,old If the features are constituted by segments of a lateral edge of the belt, then the lateral positions may be sampled by detecting the position of the edge in regular time intervals, starting at a point of time when a specific reference mark is detected on the belt.

In this way, the sampling of the position values will be synchronized with the revolutions of the endless belt.

For redundancy, it is convenient to measure the lateral positions of both edges of the belt and to take the average of the measurement results as the lateral position of the “feature”. A known feedback control mechanism may be employed for stabilizing the lateral position of the belt, and the results obtained by the method according to the invention may be utilized as input or feedback signal in this control mechanism even in the initial phase in which the correction table is still being updated.

After a sufficient number of revolutions of the belt, the sequence of correction values may have stabilized to such an extent that the correction table may then be “frozen”. Optionally, the update of the correction table may be continued ad infinitum, possibly with weight factors (a) that provide for a slower convergence than in the initial phase. In this way, it is possible to cope with secular changes of the positions of the features which may be due, for example, to thermal expansion or plastic deformation (stretching) of the belt. The invention also relates to a belt conveyor wherein the method described above is employed for controlling the position of the belt. An embodiment example will now be described in conjunction with the drawings, wherein: Fig. 1 is a schematic top plan view of a belt conveyor; Fig. 2 is a diagram illustrating principles of the invention; Fig. 3 is a flow diagram showing essential steps of the method according to the invention; and Fig. 4 illustrates the result of a simulation of a method according to the invention.

As is shown in Fig. 1, a belt conveyor 10 comprises an endless belt 12 trained around two deflection rollers 14 at least one of which is driven for rotation. A lower run of the belt 12 (not visible in the drawing) is deflected at a steering drum 16 that is tiltable under the control of a controller 18 for correcting a possible error in the lateral position of the belt 12. This lateral position is derived from measurement results of two edge detectors 20 which monitor the lateral positions of the left and right edges of the belt 12. The two edge detectors 20 may comprise sensors configured to detect features provided on the belt 12. Preferably said feature are in the form belt markers are uniquely identifiable, for example by one or more markers having different shapes from other regularly shaped markers, such that each regular marker can be identified and tracked by counting its position with respect to a differently shaped marker. As has been shown exaggeratedly in the drawings, the edges of the belt 12 are not perfectly straight. This makes it difficult to determine the true lateral position of the belt 12 or, more precisely, the true position of an idealized center line 22 of the belt. In order to be able to nevertheless determine the lateral position of the belt with high accuracy, the controller 18 maintains a correction table 24 that stores a finite number N of correction values for correcting lateral positions di as measured by the edge detectors

20. In order to be able to synchronize the correction table 24 with the belt 12, a reference mark 26 has been applied to the belt at one edge, and one of the edge detectors 20 is capable of detecting the passage of this reference mark and to send a corresponding sync signal to the controller 18. The controller 18 will then read the lateral positions measured by the edge detectors 20 in regular time intervals, the length of the time intervals being given by the period that it takes the belt 12 to make a full revolution, divided by the number N of correction values that can be stored in the correction table

24. In this way, the length of the belt 12 is virtually divided into N segments and a correction values is assigned to each of these segments. In principle, a single edge detector 20 would be sufficient. The second edge detector is provided for redundance and for improved accuracy, for the lateral position as well as for when determining the belt's speed. Eventually, for each of the N segments of the belt the controller 18 calculates the average of the lateral positions measured by the two edge detectors 20 and takes this average as a measure for a deviation di of the ideal center line 22 of the belt 12 from its target position. In Fig. 2, a straight vertical line 28 represents the target position of the belt 12 (target position of the center line 22), and a slanting line 30 represents the true actual position of the center line 22. In the example shown here, this true position of the belt gradually changes while the belt is running. In this simplified example, the length of the belt is divided into only four segments represented by short vertical bars 32. More precisely, the bars 32 represent the deviations d1, d2, d3, d4 of the average of the positions detected by the edge detectors 20 from the target position (line 28). The diagram in Fig. 2 represents a time interval that covers more than four full 5 revolutions of the belt 12 and thus includes several deviations d1, d2, … for each segment. A vertical column of numbers represents the actual lateral position D’ of the belt at the time when the position of each segment is measured. In the example shown, this true position shows a linear decrease from the value 8 (in arbitrary units) to the value 6.4 during the time interval in consideration.

Since the edges of the belt are not perfectly straight, the deviations d1, d2, … fluctuate around the true position D’ which is also represented by the slanting line 30. Whenever the edge detectors 20 measure the position of a specific segment of the belt, the result T that will be obtained in this measurement (average of the two edge positions) can be calculated by adding a correction value T-D’ to the true position D’ of the belt. Thus, for example, in the measurement represented by the topmost bar 32 in Fig. 2, the measured deviation d1 is given by: d1 =D" + (T-D’) = 8+1=9. It will be observed that the correction values T-D’ assigned to each segment of the belt are constant in time, i.e. the correction value for the first segment is always 1, the correction value for the second segment is always -1, the correction value for the third segment is always 2 and the correction value for the fourth segment is also -2 in this example.

In a practical scenario however, neither the true position D’ of the belt nor the correction values T-D’ are known a-priori. Thus, the problem is to find a method by which D’ and T-D’ can be inferred from the measurement results obtained from the edge detectors 20. A method that can efficiently fulfil this task will now be described by reference to Fig. 3.

Step S1 in Fig. 3 is a step of initializing a weight factor a and a cycle counter c to a = 0.2 andc=0.

In step S2, it is checked whether the reference mark 26 (Fig. 1) has been detected by the pertinent edge detector 20. If that is not the case (n}, the step S2 is repeated until a detection has occurred (y).

Then, in step S3, a segment counter i is initialized to i = 0, and an estimate D for the true position D’ of the belt is updated in step S4. A reasonable estimate D for the true position of the belt is the average of the deviations d1, d2, d3, d4 that have been measured during the last full revolution of the belt. Thus, in the simple example given in Fig. 2, the estimate D would be given by: D=(d1+d2+d3+d4)/4 The next step S5 is a step of counting a delay time after which the next segment will be measured. This delay time is the period required for a full revolution of the belt 12 divided by the number N of segments. Then, in step S6, the segment counter i is incremented by one, and the correction table 24 is updated in step S7. To that end, a filter procedure is applied to the deviations di that have been obtained for the segment that is designated by the segment counter i. In the example shown, the latest measurement result T that has been obtained for the deviation of this segment is multiplied by the weight factor a, the previous deviation di old that was obtained for that segment is multiplied by the weight factor (1-a), and the sum of the two products is stored as the new deviation di,new. Then, the correction value (di,new — D) is stored in the correction table 24.

Then, in step S8, an estimate for the true position P of the belt (deviation from the target position) is calculated by subtracting the correction value (di,new - D) from the current measurement result T, and the result is output as the momentaneous measurement result for the position of the belt.

In step S9, it is checked whether the segment counter i has reached the total number N of segments of the belt. If that is not the case, the steps S5 to S8 are repeated for the next segment.

When the correction table 24 has been updated in this way for all segments of the belt, the counter i will reach the value N, and the loop is left with step S10 where the cycle counter c is incremented by one.

Then, in step S11, it is checked whether the cycle counter c has reached a certain value, e.g. 50. If that is not the case, the process loops back to step S2. The result “i = N” means that the belt has made a full revolution, so that the reference mark 26 should again be approaching the edge detector 20 and will be detected again in step S2. Then, the steps S2 to S11 are repeated in a loop, so that the correction table 24 continues to be updated in each revolution cycle of the belt.

When it is detected in step S11 that the cycle counter c has reached the value of 50, this means that the filter procedure in step S7 has been iterated fifty times.

During the time period of fifty revolutions of the belt, the controller 18 has feedback-controlled the position of the belt, so that the belt has been oscillating around its target position (line 28). Consequently, it can be assumed that the fluctuations of the true belt positions and, accordingly, the fluctuations of the estimates D have cancelled out, and the correction values have sufficiently converged to the true deviation of the lateral edge position of each segment from the true position of the belt, and consequently, the correction values have stabilized.

Under this condition, the weight factor "a" may be reduced to zero in step S12, which means that the deviations di,new will not change anymore, i.e. the correction table 24 will is “frozen”. After step S12, the process loops back to step S2, and in the subsequent cycles of the belt the process will loop through the steps S2 to S11. Since c is larger than 50, the step S12 will never be reached again.

In another embodiment, the weight factor "a" may be reduced to a value that is smaller than 0.2 but larger than O in step S12. In that case, the correction table would still change dependent upon the measurement results T, but the amount of change would be smaller, and the convergence would be slower.

This slower convergence assures stable measurement results for the position of the belt but retains the possibility to adapt to minor changes in the shape of the edge of the belt 12, which changes may be due for example to thermal expansion or any other deformation of the belt.

Fig. 4 is diagram analogous to Fig. 2 and illustrates the results of a simulation of the process described above for a belt with the four segments as in Fig. 2. A meandering curve 30 reflects oscillations of the true position of the belt under the control of the controller 18, and the measurement results P obtained by the method according to the invention are represented by horizontal bars, i.e. the length of each bar has been obtained by adding the correction value (di,new - D) to the current result T provided by the edge detectors 20. It can be seen, that, thanks to the correction values, the measurement results are significantly closer to the true position of the belt (curve 30%) than in Fig. 2 and the deviations become smaller as the simulation proceeds (from top to bottom in Fig. 4). It should be noted that the simulation in Fig. 4 covers only twenty full revolutions of the belt 12, i.e. 80 segments. With increasing number of revolutions, the oscillations of the belt will be averaged-out even more effectively, and the accuracy of the measurements will improve further.

It should be observed that the controller 18 may be programmed to employ different control algorithms for controlling the lateral “wobble” of the belt. In the training phase, in which the correction table 24 is continuously updated (weight factor a > 0), the controller may employ a relatively slow control algorithm so that any deviations of the belt from its target position are corrected only slowly. This will help to speed-up the stabilization of the correction values in the correction table. Then, when the correction table has become sufficiently stable, the controller may switch to a faster control algorithm where deviations of the belt from its target position are corrected more rapidly, which results in a higher accuracy but also in a high wobble frequency control. By freezing the correction table in this mode of operation, interference between the high frequency wobble control and the process of measuring the lateral position of the belt can be avoided.

EMBODIMENTS

1. A method of measuring a lateral position (P) of and endless belt (12), comprising: - measuring lateral positions of features of the belt (12) that are distributed along the length of the belt, and - deriving the lateral position (P) of the belt from the measured positions of the features and from correction values stored in a correction table (24), characterized in that, at least in an initial phase of operation of the belt (12), the correction table (24) is continuously updated by subjecting the measured positions of each feature to a filter procedure that causes the correction values to converge towards a limit that reflects a true position of each feature.

2. The method according to claim 1, wherein the features of the belt are segments of at least one edge of the belt (12).

3. The method according to claim 2, wherein the step of measuring a lateral position of a feature of the belt comprises measuring the lateral positions of both edges of the belt in the same segment and calculating the average of the two results,

4. The method according to any of the preceding claims, wherein the filter procedure is a sliding-average filter procedure.

5. The method according to claim 4, wherein the filter procedure is defined by the formula di,new = a*Ti + (1-a)*di,old wherein di,new is a new value for the lateral position of a feature designated by an index i, Tiis a result of a latest measurement of the position of that feature, di,old is the previous lateral position of that feature, and a < 1 is a weight factor.

6. The method according to any of the preceding claims, wherein the correction values have the form dinew-D wherein di,new is a value of the lateral position of the feature, which value results from the filter procedure, and D is an estimate for the actual lateral position of the belt (12) as determined on the basis of an average of measured lateral positions of several features.

7. The method according to claim 6, wherein D is the average of the lateral positions of all features of the belt (12) as measured in one revolution of the belt.

8. The method according to any of the preceding claims, wherein a weight factor (a) which determines a speed with which the correction values converge towards the limit is reduced after the initial phase of operation of the belt, preferably wherein the weight factor (a) is reduced to zero.

9. The method according to any of the preceding claims, wherein the lateral position of the belt (12) is feedback-controlled on the basis of the measured lateral position, preferably wherein, in the initial phase of operation of the belt, a feedback- control algorithm for the lateral position of the belt is employed that results in oscillations of the lateral position of the belt with a first frequency, and, after the initial phase of operation, a feedback-control algorithm is employed which results in an oscillation of the lateral position of the belt with a second frequency that is higher than the first frequency.

10. A belt conveyor (10) comprising an endless belt (12), a steering mechanism (16) configured to alter a lateral position of the belt, a detector (20) for detecting features of the belt, and a controller (18) configured to determine the lateral position of the belt on the basis of detection results of the detector (20) and to control the steering mechanism (16), characterized in that the controller is configured to perform the method according to any of the claims 1 to 9.

Claims

CONCLUSIONS

A method for measuring a lateral position (P) of an endless belt (12), comprising: - measuring lateral positions of features of the belt (12) distributed over the length of the belt, and - measuring deriving the lateral position (P) of the tire from the measured positions of the features and from correction values stored in a correction table (24), characterized in that at least in a first phase of the operation of the tire (12 ), the correction table (24) is continuously supplemented by subjecting the measured positions of each feature to a filtering procedure that causes the correction values to converge to a limit corresponding to the actual position of each feature.

The method of claim 1, wherein the tape features are segments of at least one edge of the tape (12).

The method of claim 2, wherein the step of measuring a lateral position of a feature of the tire comprises measuring the lateral position of both edges of the tire (12) in the same segment and calculating the average of the two Results.

A method according to any one of the preceding claims, wherein the filtering procedure is a sliding average filtering procedure.

The method of claim 4, wherein the filtering procedure is determined by the formula di,new = a*Ti + (1-a)*di,old where di,new is the value for the lateral position of a feature with a designation by an index i, Ti is a result of a last measurement of the position of that feature, di,old is the previous lateral position of that feature, and a < 1 is a weighting factor.

A method according to any one of the preceding claims, wherein the correction values take the form of dinew-D where di,new is a value of the lateral position of the feature, which value is the result of the filtering procedure, and D is an estimate for the true lateral position of the tire as determined from an average of the measured lateral positions of multiple features.

The method of claim 6, wherein D is the average of the lateral positions of all features of the tire (12) as measured during one revolution of the tire (12).

A method according to any one of the preceding claims, wherein a weighting factor (a) reducing the rate at which the correction values converge is brought back to the limit after the first phase of belt operation, preferably the weighting factor being returned to zero.

A method according to any one of the preceding claims, wherein the lateral position of the tire (12) is a closed-loop control based on the measured lateral position, preferably wherein, in the first phase of the operation of the tire, a control algorithm closed loop for the lateral position of the tire is applied, resulting in oscillations of the lateral position of the tire at a first frequency, and where, after the first phase of the operation of the tire, a closed loop control algorithm is applied, resulting in in an oscillation of the lateral position of the tire with a second frequency, which is higher than the first frequency.

A belt conveyor {10) comprising an endless belt (12), a control mechanism (16) adapted to change a lateral position of the belt, a detector (20) for detecting a characteristic of the belt, and a control unit (18) arranged for determining the lateral position of the belt on the basis of detection results from the detector (20) and for controlling the steering mechanism (16), characterized in that the control unit is arranged to execute the method according to carry out any of claims 1 to 9.