METHOD FOR MONITORING THE QUALITY PROPERTIES OF A CELLULAR BALL IN A BALL LINE

Description

[0001] The invention relates to a method for monitoring the geometric properties, such as the shape, of a pulp bale in a baling line of a pulp machine, wherein cut pulp sheets are stacked to form a pulp bale. The invention further relates to a baling line of a pulp machine with a camera system and a computer program product.

[0002] CN101369549A relates to a device and method for determining whether semiconductor chips have been stacked on top of one another using automatic image processing. A target plane must be illuminated with a visible light source.

[0003] DE102005009835A1 describes a method for the congruent placement of press packages on the pressing surface of a short-cycle press in the production of decorative panels, whereby suitable, possibly colored, alignment markings are required on the edges of the press packages and similar alignment markings are required on the pressing surfaces.

[0004] JP2014038013A relates to a device and method for detecting the quality of sheet stacking. Laser light is projected linearly across the stack height.

[0005] WO2017071683A1 relates to a device and a method for the congruent laying of workpiece layers to be pressed together in a press, with a camera system for detecting the position of the individual workpiece layers and the individual workpiece stacks.

[0006] DE19909518A1 relates to a method for detecting the position of a stacked flat material, in particular a paper stack at the feeder of a printing press, wherein at least one light beam is directed onto the stack side, reflected from there and evaluated as a measure of the position.

[0007] In pulp machines, the pulp produced in web form is processed into pulp sheets in longitudinal and transverse cutters. In the baling line of the pulp machine, the cut pulp sheets are stacked into pulp bales, the pulp bale is compressed in a baler, and the compressed bale is packaged, tied, and printed. To ensure efficient operation of the baling line, especially the baler, the pulp bales must have sufficient quality characteristics, particularly with regard to geometric properties such as shape. Misshapen pulp bales, as well as uneven stacking of the pulp sheets into pulp bales, cause problems in the subsequent process steps, and especially in the baler. Pulp bales that do not meet quality requirements are preferentially removed from the baling line early on.Typically, the quality of pulp bales is monitored by operating personnel during walk-throughs, resulting in subjective, discontinuous monitoring. Trends indicating a progressive deterioration in the quality of the bales may be identified this way, but not individual pulp bales of inadequate quality. The efficient functioning of the baling line is essential for both the plant operator and the plant manufacturer, as the baling line is at the end of the process chain of a production line.

[0008] Pulp machine significantly determines the efficiency of the entire plant.

[0009] The aim of the invention is a method that allows an increase in the efficiency and availability of the bale line.

[0010] This is achieved according to the invention in that images of the pulp bale are taken using a camera system, the images are evaluated using a computer-implemented computational model for image analysis, wherein the computational model identifies geometric elements of the pulp bale in the images, selected from a group consisting of corners, edges, surfaces, and volume of the pulp bale, and assesses the geometric properties of the pulp bale from the identified geometric elements of the pulp bale, wherein the computational model was trained using a training data set and the training data set includes training images of pulp bales as well as the identified geometric elements of the pulp bale and the assessment of the geometric properties of the pulp bale for the individual training images.According to the invention, a camera system is used in the baling line for data acquisition, whereby images of the pulp bales are taken. Advantageously, the camera system is arranged relative to the pulp bale in such a way that three surfaces of the pulp bale, e.g. two adjacent side surfaces and the top surface, are depicted in the images. A computer-implemented calculation model, e.g. a neural network, is used for evaluation. The calculation model allows the geometric elements of the pulp bale, such as corners, edges, surfaces, volume, shape of the pulp bale, to be identified in the images and the geometric properties of the pulp bale to be assessed. The assessment of the geometric properties is carried out on the basis of the identified geometric elements of the pulp bale.The camera system comprises at least one camera, but can also comprise several cameras at different locations in the baling line. For example, a camera arranged in the process before a possible compaction step, e.g. by a baling press, can be used to take pictures of unpacked pulp bales. Another camera arranged in the process after the compaction step can, for example, take pictures of the packed pulp bales. The arrangement of the camera before the compaction step to monitor the geometric properties of the unpacked pulp bale is advantageous because the compaction step has a strong influence on individual geometric properties of the pulp bale, in particular the shape. In particular, one aim is to identify pulp bales of insufficient quality before a compaction step, e.g.in a baling press, since pulp bales of insufficient quality can be problematic, especially in the compaction step. The quality characteristics of the unpacked pulp bales include in particular the cutting quality of the pulp sheets, as well as, with regard to the pulp bale, its shape, squareness or stacking accuracy. The cutting quality depends on the condition of the cutting edges used in the slitters and cross cutters. Local damage to the cutting edges, e.g. due to a notch, causes corresponding damage to the cut pulp sheet. When the pulp sheets are stacked to form a pulp bale, the damage becomes apparent, for example, as a line on one side surface of the pulp bale or as local bulging of the bale, i.e. as a deviation of an edge of the pulp bale from a linear shape.Likewise, wear on the cutting edges can cause a distortion of the pulp bale's shape, causing it to increasingly deviate from the ideal cube or cuboid. The invention thus surprisingly allows conclusions to be drawn about the condition of the cross- or longitudinal cutters. Stacking accuracy, in turn, describes how precisely the pulp sheets were stacked on top of each other during the formation of the pulp bale. The more precisely the pulp sheets are arranged, the sharper the edges, or the smoother the surfaces of the pulp bale. Similarly, the side surfaces of an inaccurately stacked pulp bale exhibit increased shading or increased image noise in areas of inaccurate stacking, which can be used by the computational model to assess stacking accuracy.

[0011] In general, the shape and squareness of the pulp bale are important. The greater the deviations from an ideal cubic or cuboid pulp bale, the greater the potential for distortion in subsequent processing units, such as the baling press.

[0012] The quality characteristics of the pulp bales packed in a package include, in particular, the condition, folding, lacing, or printing of the packaging. After the compaction step, e.g., in a baling press, the pulp bale is wrapped. The condition of the packaging indicates whether the latter is intact or damaged. The wrapper sheets used for packaging are folded around the pulp bale, with an even fold representing a quality fold. To secure the fold, the wrapped pulp bales are usually tied, e.g., with a wire cord; an even lacing of the pulp bale is desired. The wrapped pulp bale is often also printed. In The assessment of print quality includes, for example, how completely the print motif was printed on the packaging of the pulp bales. According to the invention, the computer-implemented calculation model was trained using a training data set. The training data set includes training images of pulp bales. In addition to the individual training images, the training data set also includes the identified geometric elements of the pulp bale and the assessment of the geometric properties of the pulp bale.

[0013] It is possible to create the training dataset explicitly in the bale line of a pulp machine. To do this, at least one camera of the camera system is positioned in a defined manner along the bale line. The camera always takes images of the pulp bales when the pulp bale and camera are positioned in the defined manner. This ensures that the arrangement of the pulp bales in the images is similar. This has the advantage that a small training dataset is sufficient to train the computational model specifically for the bale line.

[0014] It is also conceivable that the training model is trained with a large number of training data sets, using images from different bale lines with different relative positions of pulp bales to the camera. This minimizes the influence of the relative positions of pulp bales to the camera on the prediction of the computational model. This requires extensive training of the computational model with training data sets from different bale lines, which is a long-term process. However, conversely, it allows for the immediate use of the trained computational model in a previously unknown bale line.

[0015] The training data set can, for example, also include the assessment of whether a pulp bale depicted in a training image should remain in the baling line or be removed from the baling line. A computer model trained in this way allows for the automated removal of pulp bales from the baling line based on the assessment by the computer model.

[0016] The method according to the invention thus enables continuous, automated, reproducible, and objective classification of the pulp bale and its quality characteristics. In particular, the evaluation of specific quality characteristics is possible using the trained computational model.

[0017] An advantageous embodiment of the method involves moving the pulp bale to a defined position relative to the camera system, and the camera system taking images of the pulp bale in the defined position. This minimizes any potential influence of the positioning of the pulp bale relative to the camera on the prediction of the computational model, as the arrangement of the pulp bale in the images is similar. If a computational model is then trained explicitly for a bale line of a pulp machine, a small training data set allows for sufficient and rapid training of the computational model.

[0018] In a further advantageous embodiment of the method, a light source is used to illuminate the pulp bale, with the camera system taking images of the illuminated pulp bale. Illuminating the pulp bale with a light source for illumination when taking the images allows defined lighting conditions to be set. The influence of the lighting conditions on the prediction of the computational model is thus minimized because the lighting conditions are defined or consistent, and the images were created under the defined lighting conditions. This allows the size of the training data set to be kept small while maintaining the predictive value of the computational model.

[0019] In a favorable embodiment of the method, a line is projected onto a surface of the pulp bale using a light source for projecting a line, in particular a coherent light source, e.g., a line laser. From a distortion of the line on the surface of the pulp bale, the computer model can assess the quality properties of the pulp bale, in particular the stacking accuracy or the cutting quality of the individual sheets. Stacking accuracy describes how precisely the pulp sheets are arranged one above the other during the formation of the pulp bale. The less precisely the pulp sheets are arranged one above the other to form a pulp bale, the more uneven the edges of the pulp bale are, or the rougher the surfaces of the pulp bale are, which results in a greater distortion of a line projected onto a surface of the pulp bale. When using a coherent light source, e.g.,A line laser, for example, precisely maps the projections and recessions of the stacked pulp sheets as distortions, or jumps, in the projected line. The computational model evaluates the distortions or jumps in the projected line, thus allowing the stacking accuracy to be determined. The distortion of the line can be quantified, for example, by the size and frequency of the jumps in the line projected onto the pulp bale.

[0020] In an advantageous embodiment of the method, the calculation model was calibrated with a calibration data set, wherein the calibration data set comprises calibration images of pulp bales as well as a quantification of the geometric elements of the pulp bale for the individual calibration images, e.g. the length of the edge, the area or the volume of the pulp bale. After calibration, the calculation model allows quantitative statements about the pulp bale and the calculation model can, for example, output the edge length, the volume or the area of ​​individual sides of the pulp bale. The quantification of the geometric elements on the calibration images can be carried out, for example, by attaching at least one and preferably two scales to a surface of the pulp bale. Particularly advantageously, the background of the pulp bale has a calibration pattern, for example a checkerboard pattern with defined dimensions.

[0021] The invention also relates to a baling line of a pulp machine comprising a camera system and means for carrying out the method steps according to the invention.

[0022] The invention further relates to a computer program product comprising instructions for carrying out the method steps according to the invention in a baling line of a pulp machine according to the invention.

[0023] The invention will now be described by way of example with reference to the drawings. Fig. 1 shows a pulp bale in a bale line Fig. 2 shows a pulp bale with areas identified by the computational model. Fig. 3 shows a pulp bale with edges identified by the computational model.

[0024] Fig. 1 shows an unpackaged pulp bale 1 stacked from cut pulp sheets in a bale line. The human eye can easily identify the geometric elements of the pulp bale 1, for example, individual surfaces 2 or edges 3. According to the relative arrangement of the pulp bale 1 to the imaging camera system, three surfaces of the pulp bale can be recognized. With the inventive computer-implemented computational model for image analysis, the Fig. 1 The computer model identifies geometric elements of the pulp bale 1 in the image, selected from a group consisting of corners, edges 3, surfaces 2, and volume of the pulp bale, and assesses the quality properties of the pulp bale 1, such as the cutting quality of the pulp sheets as well as the shape, perpendicularity, and stacking accuracy, based on the identified geometric elements of the unpacked pulp bale 1. Fig. 1 A line 4 projected onto a surface of the pulp bale 1 can also be seen. In this optional method step, the line 4 is projected onto the surface 2 of the pulp bale 1 using a light source for projecting a line, in particular a coherent light source, e.g., a line laser. This optional method step allows, in particular, an advantageous assessment of the stacking accuracy or the cutting quality of the individual sheets based on the distortion of the line 4 on the surface 2 of the pulp bale 1.

[0025] Fig. 2 shows an unpacked pulp bale 1 and the hatched areas 2 with spherical corners identified by the computational model. The edges 3, which demarcate adjacent areas from each other, were thus also identified by the computational model. Fig. 3shows the pulp bale 1 and the edges 3 identified by the calculation model and represented by dashed lines, whereby an ideal pulp bale is indicated analogously to a cube or cuboid. The calculation model thus allows the quality properties, in particular the geometric properties, such as the shape or rectangularity, of the pulp bale 1 to be monitored. From the shape of the edges identified by the calculation model, the calculation model can, for example, further assess the cutting quality of the pulp sheets and the stacking accuracy of the pulp bale. The optional use of a light source to project a line onto a side surface of the pulp bale enables a more precise assessment of various quality properties of the pulp bale 1, such as the stacking accuracy or the cutting quality, based on the assessment of the distortion of the projected line.

[0026] The present invention offers numerous advantages. It allows for increased efficiency and availability of the baling line through continuous, automated monitoring of the quality characteristics of the pulp bales. The quality monitoring of the pulp bales is objective and reproducible, allowing pulp bales of inadequate quality, for example, those with inadequate geometric properties, particularly an inadequate shape, to be removed from the baling line at an early stage without causing problems in subsequent process steps. The present invention also allows conclusions to be drawn about the condition of, or monitoring of, the blades used in the longitudinal and transverse cutters. Reference symbol

[0027] (1) Pulp bale (2) Surface (3) Edge (4) Line

Claims

1. Method for monitoring the geometrical properties, such as the shape, of a pulp bale (1) in a baling line of a pulp machine, wherein cut pulp sheets are stacked to form a pulp bale (1) are and images of the pulp bale (1) are taken with a camera system of the pulp bale (1) and the images are evaluated with a computer-implemented computational model for image analysis, characterised in that that the computer model on the images geometric elements of the pulp bale (1) selected from a group consisting of corners, edges (3), surfaces (2) and volume of the pulp bale, identified and from the identified geometric elements of the pulp bale (1) assessing the geometric properties of the pulp bale (1), wherein the computational model has been trained with a training data set and the training data set comprises training images of pulp bales (1) and, for the individual training images, the identified geometric elements of the pulp bale (1) and the assessment of the geometric properties of the pulp bale (1).

2. Method according to claim 1, wherein the pulp bale (1) is moved to a defined position relative to the camera system and the camera system takes images of the pulp bale (1) in the defined position.

3. Method according to one of claims 1 to 2, wherein a light source is used to illuminate the pulp bale (1) and the camera system takes images of the illuminated pulp bale (1).

4. Method according to one of claims 1 to 3, wherein a line (4) is projected onto a surface (2) of the pulp bale (1) using a light source for projecting a line (4), in particular a coherent light source, e.g. a line laser, and the calculation model assesses the stacking accuracy of the pulp bale (1) or the cutting quality of the individual sheets from a distortion of the line (4) on the surface (2) of the pulp bale (1).

5. Method according to one of claims 1 to 4, wherein images of the pulp bale (1) are taken with the camera system, the pulp bale is unpacked and the computational model assesses the cut quality of the pulp sheets or the shape, squareness or stacking accuracy of the pulp bale (1).

6. Method according to one of claims 1 to 5, wherein images of the pulp bale (1) are made with the camera system, the pulp bale (1) is wrapped making a package and the computational model assesses the condition, folding, tying, or printing of the package.

7. Method according to one of claims 1 to 6, wherein the computational model has been calibrated with a calibration data set and the calibration data set comprises calibration images of pulp bales (1) and, for the individual calibration images, a quantification of the geometric elements of the pulp bale (1), e.g. the length of the edge (3), the area (2) or the volume of the pulp bale (1).

8. Baling line of a pulp machine with a camera system and means suitable for carrying out the method according to one of claims 1 to 7.

9. A computer program product comprising instructions for causing the device of claim 8 to perform the method steps of one of claims 1 to 7.