Controlling a BALE press in a pulp baling system to compress a current BALE to a desired height

Abstract

A controller (100) controls a bale press (30) in a pulp baling system (10) to compress a current bale (BC) to have a desired compressed bale height (HPdes) via a closed-loop controller (110): determining a desired compressed bale height (HPdes) based on a set of input parameters (Sinput) specifying a desired resulting bale height (HRdes) for the current bale (BC), a desired stack height (HSdes) of a current stack (SM) that the current bale (BC) will be part of, properties of the current bale and possibly properties of preceding bale(s); and producing predicted control signal(s) (Cpred) to be fed to the bale press (30) for controlling the bale press (30) to compress a current bale (BC). The closed-loop controller (110) contains a trained artificial neural network. The invention also relates to a pulp baling system (10) comprising the controller (100), and a corresponding computer-implemented method, computer program (653), and non- volatile data carrier (655) containing the computer program (653).

Description

[0001] CONTROLLING A BALE PRESS IN A PULP BALING SYSTEM TO COMPRESS A CURRENT BALE TO A DESIRED HEIGHT

[0002] TECHNICAL FIELD

[0003] The present invention relates to controlling a bale press in a pulp baling system to compress a current bale positioned in the pulp bale press to have a desired compressed height when it exits the bale press.

[0004] The invention includes a controller, a pulp baling system comprising such a controller, and a computer implemented method.

[0005] BACKGROUND

[0006] The present patent disclosure concerns a pulp bale press, a method of pulp bale pressing and a pulp bale production line or system.

[0007] Pulp is a material prepared by chemical or mechanical means from various materials (such as wood or rags) for use in, among others, making paper and cellulose products. Pulp comprises lignocellulosic fibers and can be used for producing paper. When pulp is produced in a pulp mill, it can either be directly used in paper production on-site, or pulp can be packaged for shipment to produce paper elsewhere.

[0008] A pulp mill typically comprises a forming section, into which stock to be dried and formed is delivered by means of a headbox. After the forming section there is typically a press section, a dryer section and a slab press or cutting section with a cutter-layboy for forming received pulp into pulp bales. A pulp bale may hereinafter also be referred to as a bale. In a slab press pulp is pressed into bales, while in a cutter-layboy the web is slit in its machine direction into a number of partial webs which in turn are cut in cross direction into sheets and piled into bales. Often, a pulp bale, when seen from the top, has a rectangular shape. The formed bales commonly have a weight of about 200-250 kg. After the bale forming section, the bales are transported on conveyors or the like between different baling process stages in a baling line, e.g., pressing, stacking, wrapping and binding, before they are offloaded and transported to the operator. A conveyor may convey more than one pulp bale at a time.

[0009] In the pressing stage, the formed bales are compressed in a pulp bale press to reduce the volume and increase the density of the bales. As pulp production is typically a continuous process, the pulp bale press, or bale press, is part of the pulp production line, specifically it is part of the baling line in the pulp production line. In a production line, the number of bales to be compressed by the press may be, for instance, as high as 300 bales per hour. A bale press may comprise a base having a conveyor or other arrangement for moving or transporting the bales, which positions the incoming bale to be compressed below a movable pressing element or ram. Once the bale to be compressed is in position, the movable pressing element is pushed downwards towards the base to compress the pulp bale with a certain force, for instance 15,000 kN. This force is, for instance, suitable for a pulp bale weighing 250 kg. Of course, instead of defining the compression force, a corresponding compression pressure may be set or a corresponding pressure depth to which the pulp bale press is controlled to press may be defined, as these are all interchangeable parameters for achieving a desired compression of the bale. The downwards force or pressure of the movable pressing element is usually applied, or the defined pressure depth is usually achieved, using a hydraulic cylinder arranged above the movable pressing element. Often a relatively large hydraulic cylinder is used in order to apply the above noted force. An example of such a pulp bale press is the Valmet RoboPress PR- 15. Once the movable pressing element is at a point where the required force is applied, it is lifted back upwards. Thereafter, the compressed bale is output of the pulp bale press and a new, not yet compressed and thus higher, pulp bale to be compressed is input into the pulp bale press. These necessary steps set a boundary for the number of pulp bales to be compressed by these pulp bale presses.

[0010] In the subsequent stacking, wrapping and binding stages the bales are stacked and packaged for shipment, typically either as separate stacks or as units of two or more stacks bound together. Several stacks or units of stacks will, in storage as well as during transportation, be placed on top of each other in piles. As each stack or unit may weigh a ton or more, it is vitally important that the stacks or units that leave the baling line are as uniform as possibly in height, to reduce the risk of instability of the piles, as such instability is a great safety hazard for any person working in the vicinity of the piles.

[0011] There are some suggested improvements within this area, but further developments are needed to ensure an improved uniformity of the height of stacks produced in the baling line, to improve both the space efficiency and the safety of personnel during stacking, piling, storage and transport, and further to reduce waste due to stacks that do not meet the height requirements within acceptable deviations being discarded.

[0012] SUMMARY

[0013] The object of the present invention is to eliminate or at least to minimize the problems discussed above. It is thereby the object of the present invention to further develop solutions for producing pulp bales with improved production quality performance.

[0014] This is achieved by a controller for controlling a bale press in a pulp baling system to compress a current bale positioned in the pulp bale press to have a desired compressed bale height, a pulp baling system comprising such a controller and a computer implemented method for controlling a bale press in a pulp baling system to compress a current bale positioned in the pulp bale press to have a desired compressed bale height according to the appended independent claims. Further advantageous developments of the present invention are defined in the dependent claims.

[0015] In a first aspect the present invention relates to a controller for controlling a bale press in a pulp baling system to compress a current bale to a desired compressed bale height. The controller according to the invention comprises a closed-loop controller configured to obtain a set of input parameters comprising a desired resulting bale height or a modified desired resulting bale height of the current bale, a desired stack height of a current stack of N bales that the current bale will be part of, a first set of bale properties of the current bale and, if the current bale is not the first bale in the current stack and at least one preceding bale in the current stack is positioned in a stacker of the pulp baling system, a second set of bale properties of each of the at least one preceding bale in the current stack that is positioned in the stacker. The closed-loop controller is further configured to, based on the obtained set of input parameters, determine a desired compressed bale height for the current bale and to produce at least one predicted control signal that is adapted to be fed to the bale press for controlling the bale press to compress the current bale to the desired compressed bale height. The closed-loop controller comprises a trained artificial neural network, ANN, which in turn comprises an input layer configured to obtain the set of input parameters, an output layer configured to provide the at least one predicted control signal, and at least one hidden layer interconnecting the input layer and the output layer. Each of the input, output and at least one hidden layer comprises a respective set of nodes connected to nodes to the respective of neighboring layers via a respective weight.

[0016] Variations of pulp and bale properties in the bales upstream of the bale press and varying residence times in the baling line leads to variations in bale swelling, and thus variations in resulting height of the bales that reach the stacker. At the same time, the stack heights must be within certain tolerances. To maintain the stack heights within these tolerances, the inventors have realized that the bale press needs to compensate for upstream variations in an improved manner. Suitably, the closed-loop controller of the controller described herein thereby takes into account a large number of registered pulp and bale properties of the current bale, i.e. the bale that is next in line to enter the bale press, registered upstream of the bale press. The closed-loop controller further takes into account properties both upstream and downstream of the bale press of any other relevant preceding bales. Furthermore, the closed-loop controller has the knowledge of the trained ANN. If the current bale is not the first bale in the current stack and at least one preceding bale in the current stack is positioned in the stacker of the pulp baling system, i.e. has reached the end of the baling line, properties of each of the at least one preceding bale in the current stack that is positioned in the stacker are also fed into the closed-loop controller and used in the determination of the desired compressed bale height and prediction of the control signal. By this feedback of information on the resulting respective height of any bale already in the stacker and the respective properties of the bales including one or more heights of each of said bale measured along the baling line indicating the amount of swelling at different stages and times, the determination of the desired height of the current bale can be made also taking this additional information into account, in real-time or close to real-time. Thereby, an even further improved prediction of how the current bale should be compressed by the bale press is achieved, based on its individual properties, to reach the desired resulting bale height after swelling. This means that an improved control of the bale press is enabled, leading to an improved control of the resulting stack height, meaning that the desired stack height can be reached with higher accuracy. Based on all this knowledge, the closed-loop controller is configured to predict to which height each current bale needs to be compressed, in order for it to swell to the desired resulting height at the stacker at the end of the baling line. Using the knowledge obtained through training of the trained ANN and knowledge of the characteristics, parameters, and properties of the current bale, the ANN thus determines to a high degree of accuracy the desired compressed bale height such that, after the swelling along the baling line that is predicted based on the properties of the bale and historic knowledge of swelling of bales with known properties, the current bale will reach its desired resulting bale height. Based on this, the ANN suitably predicts the control signal to control the bale press to compress the current bale accordingly. This controller is advantageous because it provides substantially enhanced control of the bale press and the resulting stack height in relation to traditionally controlled bale presses. In fact, for stacks comprising four bales and with a desired resulting stack height of 2000 mm, the inventors have shown that the allowed deviation in resulting stack heights of stacks produced by the baling line, using embodiments of the invention described herein, can be decreased from about 60 to 80 mm, which is the lowest deviation that can be guaranteed by previous solutions, to less than 20 mm, and even as low as 10 mm or less for a stack of four bales. This is a very significant improvement. Thereby, space efficiency and safety of personnel during stacking, piling, storage and transport is ensured. The improved accuracy of course also leads to greatly reduced waste and increase in yield from the baling line. The inventive controller is especially advantageous if more than one bale former is used to form pulp bales for the same baling line and one bale former is already up and running while the other one is being initiated. Traditionally, using prior solutions without the inventive closed-loop controller, there will in this scenario initially be great differences in resulting bale height of bales originating from the first bale former compared to bales originating from the second (possibly third, fourth, etc.) bale former before the regulator has been tuned in. However, using embodiments of the invention a much faster feedback regulation and knowledge of the trained ANN leads to the tune-in process being much faster and the resulting bale height becoming uniform at a much earlier stage.

[0017] The predicted control signal may be indicative of a predicted press depth and a predicted hold time, for controlling the bale press to compress the current bale to the desired compressed bale height by pressing the current bale to the predicted press depth during the desired hold time. Alternatively, the predicted control signal may be indicative of a predicted pressure and a predicted hold time, for controlling the bale press to compress the current bale to the desired compressed bale height by applying the predicted pressure on the current bale during the predicted hold time. Thereby, alternative manners of controlling the bale press to achieve the desired compression result are provided.

[0018] The controller may further comprise a first summation module configured to derive a modified desired resulting bale height for the current bale based on the desired resulting bale height of the current bale and at least one first output signal from the stacker. From the at least one first output signal, the actual resulting height of at least one previous bale in the current stack or in a preceding stack in the same unit that has reached the stacker can be derived. For this purpose, the at least one first output signal may reflect a registered resulting bale height of at least one preceding bale in the current stack or in the preceding stack in the same unit, wherein the at least one registered resulting bale height is registered by at least one height sensing device arranged at the stacker. Alternatively, or additionally, the at least one first output signal may reflect a registered resulting stack height of the current stack or the preceding stack in the same unit, wherein the resulting stack height is registered by at least one height sensing device arranged at the stacker. Since the number of bales in the current stack that have reached the stacker is known at any given time, and the height of the stack is continuously measured, the resulting bale height of each bale that has reached the stack can be derived from the registered height of the stack at a specific time. The summation module thereby suitably derives the modified desired resulting bale height based on the desired resulting bale height of the current bale and information on the actual resulting height of one or more bales of the current stack or the preceding stack in the same unit that have already been stacked. If the combined resulting bale heights of the at least one preceding bale differ from the combined desired bale heights, the first summation module is thus configured to determine the modified desired resulting bale height for the current bale to compensate for this deviation. The closed-loop controller is then configured to set the desired resulting bale height to the modified desired resulting bale height before determining the desired compressed bale height for the current bale and producing the at least one predicted control signal. Advantageously, real-time compensation for any deviation in resulting bale heights within the current stack or unit of stacks that the current stack belongs to is thereby achieved.

[0019] In some embodiments, the stacks produced by the baling system will be formed into units of two or more stacks that are typically bound together in a unit former at the end of the baling line. In these embodiments, the first output signal may further reflect a resulting stack height of at least one preceding stack that together with the current stack will form a unit, assuming that at least one such stack is present at the stacker or in the unit former. If the resulting stack height of the preceding stack differs from the desired stack height more than a first allowed deviation, the closed- loop controller is further configured to set the desired stack height for the current stack to the resulting stack height of the preceding stack before producing the at least one predicted control signal. Thereby, any deviation greater than what is allowed within the predetermined tolerance defined by the first deviation is detected and appropriate measures can be taken. Suitably, real-time compensation for any variations and deviations in bale and stack height is thereby performed, which advantageously results in the stacks of a unit becoming more uniform than if only the preset reference height for stacks was used. This contributes to improving stability and safety in storage and transport, as well as reducing waste. After the last bale in a unit has been compressed, the desired stack height may suitably be reset to the initial desired stack height, i.e. the reference value for the baling system. In these embodiments, the second set of bale properties may further comprise bale properties of any or all of the bales in the preceding stack in the unit that the current stack will be a part of.

[0020] The closed-loop controller may also be configured to obtain at least one second output signal from the bale press. The at least one second output signal reflects a first registered compressed bale height of the preceding bale, wherein the bale height is registered by a height sensing device connected to and arranged to register the press depth of the bale press. Based on the second output signal, the closed-loop controller is configured to determine if the first registered compressed bale height of the preceding bale differs from the desired compressed bale height for the preceding bale more than a second allowed deviation. Thereby, any deviation greater than what is allowed within the predetermined tolerance defined by the second deviation is detected and appropriate measures can be taken. For this purpose, the controller may further comprise a second summation module configured to derive a modified desired compressed bale height for the current bale based on the desired compressed bale height, the first registered compressed bale height of the preceding bale and at least one third output signal from at least one height sensing device arranged directly downstream of the bale press, wherein the at least one third output signal reflects a second registered compressed bale height of the preceding bale registered by said at least one height sensing device. If it is determined that the first registered compressed bale height of the preceding bale differs from the desired compressed bale height for the preceding bale more than the second allowed deviation, the closed-loop controller is configured to obtain the modified desired compressed bale height from the second summation module and set the desired compressed bale height to the modified desired compressed bale height before producing the at least one predicted control signal. Suitably, real-time compensation for any variations and deviations in bale height is thereby performed, in the form of direct feedback control for compressing the current bale based on the known control signal and compression result of the preceding bale, which has just exited the bale press. Advantageously, the real-time feedback and compensation results in increased accuracy in achieving the desired stack height, and further to stack heights becoming even more uniform. This contributes to improving stability and safety in storage and transport, as well as reducing waste. In some embodiments, if the first registered compressed bale height of the preceding bale differs from the desired compressed bale height for the preceding bale more than the second allowed deviation, the closed-loop controller may further be configured to determine a modified maximum allowed hold time for the bale press, and / or determine that the bale press needs to compress the current bale at least two times, to achieve the desired compressed bale height, and produce the at least one predicted control signal such that it is configured to control the bale press to compress the current bale according to the modified maximum allowed hold time and / or to compress the current bale at least two times. Suitably, it is thereby possible to compensate for inadequate compression of the preceding bale in cases where the problem is not the compression control instructions fed to the bale press, but rather the physical limitations of the bale press itself. After the current bale has been compressed, or after the last bale in the current stack or unit has been compressed, the maximum allowed hold time for the bale press and / or number of compressions for each bale may suitably be reset to the initial values, i.e. the reference value for the baling system.

[0021] The weights of the trained ANN have been assigned through a training process in which the set of input parameters and the at least one first output signal and, in some embodiments, one or more of the at least one second output signal and the at least one third output signal were used as training data and registered control signals configured to control the bale press of the pulp baling system were used as reference data. In one embodiment, the weights of the trained ANN have been determined iteratively via a backpropagation training process that involves comparing training data that express the registered control signals with the at least one predicted control signal produced by an ANN under training. In embodiments described herein, the trained ANN is thus represented by the ANN (i.e. the ANN under training) after this training process has been completed.

[0022] The trained artificial neural network, ANN, may use a Long Short-Term Memory, LSTM, algorithm. Suitably, the trained ANN is thereby able to account for the order dependence of stacks and / or stacked bales. According to further embodiments of this aspect of the invention, the trained ANN is either implemented by means of a computer program that runs on at least one processing unit, or by at least one neuromorphic circuit. In general, the former alternative may offer a higher degree of flexibility, whereas the latter alternative may be more efficient, for example in terms of latency and overall power consumption.

[0023] According to another embodiment of this aspect of the invention, the closed-loop controller is configured to operate according to a PID regulation principle, a linear- quadratic regulation principle or a model predictive control principle. Namely, each of these types of control principles offer specific advantages.

[0024] According to yet another embodiment of this aspect of the invention, the at least one predicted control signal is adapted to control at least one electric motor and / or at least one hydraulic actuator of the bale press, such that the at least one electric motor and / or hydraulic actuator causes the bale press to compress the current bale to the desired compressed bale height. In particular, the at least one predicted control signal may be adapted to cause a respective control current and / or voltage to be produced, which respective control current and / or voltage has such a temporal profile with respect to magnitude and sign and / or is modulated in such a manner that the respective control current and / or voltage operates the at least one electric motor and / or the at least one hydraulic actuator to mechanically control a movable pressing element of the bale press to be pushed downwards towards a base of the bale press to compress a pulp bale positioned between the movable pressing element and the base. In other words, during a period when the at least one predicted control signal controls a particular motor / actuator, the control current and / or voltage may vary over time and / or in terms of modulation such that components of the bale press are controlled to move as intended.

[0025] In a second aspect the present invention also relates to a pulp baling system comprising a controller according to any embodiment of the invention. The pulp baling system further comprises a bale former for forming pulp into a bale, a bale press for pressing the formed bale, a stacker for stacking N bales into a stack, and at least two height sensing devices arranged at different positions along the pulp baling system for registering heights of bales positioned at the different positions in the pulp baling system. The positions of the height sensing devices include at least at the bale press and at the stacker. The pulp baling system may further comprise a wrapping section downstream of the bale press and upstream of the stacker, for wrapping and marking each bale before it is stacked. The pulp baling system may also comprise a unit former downstream of the stacker, for forming a unit comprising at least two stacks.

[0026] In a third aspect the invention includes a computer implemented method for controlling a bale press in a pulp baling system to compress a current bale positioned in the pulp bale press to have a desired compressed bale height when it exits the bale press. The method comprises obtaining, in a closed-loop controller, a set of input parameters comprising a desired resulting bale height or a modified desired resulting bale height of the current bale, a desired stack height of a current stack of N bales that the current bale will be part of, a first set of bale properties of the current bale and, if the current bale is not the first bale in the current stack and at least one preceding bale in the current stack is positioned in a stacker of the pulp baling system, a second set of bale properties of each of the at least one preceding bale in the current stack that is positioned in the stacker. Based on the obtained set of input parameters, the method further comprises determining, by the closed-loop controller, a desired compressed bale height for the current bale and producing, by the closed-loop controller, at least one predicted control signal that is adapted to be fed to the bale press for controlling the bale press to compress the current bale to the desired compressed bale height. The closed-loop controller comprises a trained artificial neural network, ANN, which in turn comprises an input layer configured to obtain the set of input parameters, an output layer configured to provide the at least one predicted control signal, and at least one hidden layer interconnecting the input layer and the output layer. Each of the input, output and at least one hidden layer comprises a respective set of nodes connected to nodes to the respective of neighboring layers via a respective weight. The weights of the ANN may have been assigned through a training process similar to any of the embodiments described in connection with the first aspect of the invention. The method may comprise deriving, in a first summation module, a modified desired resulting bale height for the current bale based on the desired resulting bale height of the current bale and at least one first output signal from the stacker, similar to the at least one first output signal described in connection with the first aspect of the invention, and setting, by the closed-loop controller, the desired resulting bale height to the modified desired resulting bale height before determining the desired compressed bale height for the current bale and producing the at least one predicted control signal. If the first output signal reflects a resulting stack height of a preceding stack that together with the current stack will form a unit of M stacks, the method according to this embodiment further comprises determining, by the closed-loop controller, if the resulting stack height of the preceding stack differs from the desired stack height more than a first allowed deviation, and, if it does, setting, by the closed-loop controller, the desired stack height for the current stack to the resulting stack height of the preceding stack before producing the at least one predicted control signal.

[0027] The method may comprise obtaining, in the closed-loop controller, at least one second output signal from the bale press that similarly to the at least one second output signal described in connection with the first aspect of the invention reflects a first registered compressed bale height of the preceding bale, comparing the first registered compressed bale height of the preceding bale to a previously determined desired compressed bale height for the preceding bale, and determining if the first registered compressed bale height of the preceding bale differs from the desired compressed bale height for the preceding bale more than a second allowed deviation. If it does, the method may according to this embodiment further comprise obtaining, in a second summation module, at least one third output signal that in likeness with the at least one third output signal described in connection with the first aspect of the invention reflects a second registered compressed bale height of the preceding bale registered by at least one height sensing device arranged directly downstream of the bale press, deriving, in the second summation module, a modified desired compressed bale height for the current bale based on the desired compressed bale height, the first registered compressed bale height of the preceding bale and the at least one third output signal, and setting, by the closed-loop controller, the desired compressed bale height to the modified desired compressed bale height before producing the at least one predicted control signal. Alternatively, or additionally, if the first registered compressed bale height of the preceding bale differs from the desired compressed bale height for the preceding bale more than a second allowed deviation, the method may in this embodiment further comprise determining, by the closed-loop controller, a modified maximum allowed hold time for the bale press, and / or determining that the bale press needs to compress the current bale at least two times, to achieve the desired compressed bale height. In this case, the closed-loop controller produces the at least one predicted control signal such that it is configured to control the bale press to compress the current bale according to the modified maximum allowed hold time and / or to compress the current bale at least two times, to achieve the desired compressed bale height.

[0028] The method may further comprise feeding the at least one predicted control signal from the controller to the bale press to control the bale press to compress the current bale to the desired compressed bale height.

[0029] According to a fourth aspect of the invention, the object is achieved by a computer program loadable into a non-volatile data carrier communicatively connected to at least one processing unit. The computer program includes software for executing the above method when the program is run on the at least one processing unit.

[0030] According to a fifth aspect of the invention, the object is achieved by a non-volatile data carrier containing the above computer program.

[0031] Any advantage described in connection with one aspect of the invention, e.g. the controller, is equally applicable to corresponding embodiments of other aspects of the invention, e.g. the pulp baling system, the method, the computer program and the non-volatile data carrier.

[0032] Many additional advantages, beneficial features and applications of the present invention will be apparent from the following description and the dependent claims.

[0033] DRAWINGS The invention will now be described in more detail with reference to the appended drawings, wherein

[0034] Fig. 1 shows a block diagram of a pulp baling system according to an embodiment of the invention;

[0035] Fig. 2 shows a block diagram of a controller according to an embodiment of the invention;

[0036] Fig. 3 shows a block diagram of a controller according to an embodiment of the invention;

[0037] Fig. 4 shows a block diagram of a controller according to an embodiment of the invention;

[0038] Fig. 5 illustrates the general principle according to which the ANN may be trained according to an embodiment of the invention;

[0039] Fig. 6 illustrates how the controller may be implemented in software running on a processing unit according to an embodiment of the invention;

[0040] Fig. 7 illustrates, by means of a flow diagram, the general method for controlling a bale press in a pulp baling system according to the invention;

[0041] Figs. 8-10 show a schematic view of bales moving from position Pl to position Pl l in a pulp baling system according to the invention; and

[0042] Fig. 11-13 illustrate, by means of flow diagrams, further embodiments of the method of Fig. 7.

[0043] All the figures are schematic, not necessarily to scale, and generally only show parts which are necessary in order to elucidate the respective embodiments, whereas other parts may be omitted or merely suggested. Any reference number appearing in multiple drawings refers to the same object or feature throughout the drawings, unless otherwise indicated.

[0044] DETAILED DESCRIPTION First, a controller 100 according to embodiments of the invention, and a pulp baling system 10 comprising such a controller 100 will be described in connection with Figs. 1 to 4, and also in connection with Figs. 8, 9 and 10. The controller 100 is adapted to control a bale press 30 in a pulp baling system 10 to compress a current bale Be to have a desired compressed bale height HPdes. The desired compressed height of the current bale is determined using a controller with a trained artificial neural network, ANN, based on input parameters and a feedback loop, according to embodiments herein.

[0045] Fig. 1 schematically discloses a block diagram of a pulp baling system 10 according to an embodiment of the invention. The pulp baling system 10 comprises a bale former 20 for forming pulp P into a bale B, a bale press 30 for pressing each formed bale B and a stacker 50 for stacking N pressed pulp bales BI:N into a stack S. N is the number N of bales in each stack. N is herein an integer greater than one, preferably in the interval of 3 to 5, but more than 5 bales in a stack may also be possible. More preferably the number of bales in each stack is N=3 or N=4. The pulp baling system 10 further comprises a plurality of sensors each being configured to register one or more property of the bales B. The plurality of sensors include at least two height sensing devices being arranged at different positions along the pulp baling system 10 for registering the height of a bale B in the respective position, as described further in connection with Figs. 8 and 9. The positions where height sensing devices are arranged at least include the bale press 30 and the stacker 50, for registering heights of a respective bale B located in each of said positions. The pulp baling system 10 also comprises a controller 100, according to any embodiment described herein, i.e. in the summary, detailed description, figures and appended claims, for controlling the bale press 30 to compress a current bale Be to have a desired compressed bale height HPdes when it exits the bale press 30.

[0046] The bale former 20 may comprise at least one slab press, and / or at least one cutter-layboy arrangement. If there are more than one slab press and / or cutter- layboy arrangement, the bales may be delivered to a single bale press 30 from the more than one bale formers 20. The pulp baling system 10 may comprise a wrapping section 40 downstream of the bale press 30 and upstream of the stacker 50, for wrapping and marking each bale B before it is stacked, in any suitable manner known in the art. The pulp baling system 10 may comprise a unit former 60 downstream of the stacker 50, for forming a unit U comprising at least two stacks SI, S2. Forming a unit may include binding or in other ways assemble and attach the at least two stacks S 1 , S2 to each other to form the unit U, in any suitable manner known in the art. A unit U herein comprises M stacks, wherein M is an integer and M > 1.

[0047] The pulp bale press 30 is hereinafter also referred to as the bale press 30.

[0048] The pulp baling system 10 may also be referred to as a pulp bale production system, a pulp bale production line or a baling line.

[0049] The invention will be described with reference to various embodiments, and it is in particular to be noted that a feature from one embodiment may freely be introduced into another embodiment except where such a combination is explicitly stated as unsuitable or undesirable. Thus, the embodiments disclosed herein may be combined with each other as desired.

[0050] Fig. 2 shows a block diagram of a controller 100 according to a first embodiment of the invention.

[0051] The controller 100 comprises a closed-loop controller 110. The closed-loop controller 110 is configured to obtain a set of input parameters Sinput comprising a desired resulting bale height HRdes or a modified desired resulting bale height AHRdes of the current bale Be, a desired stack height HSdes of a current stack SM of N bales that the current bale Be will be part of, and a first set BprOpi of bale properties of the current bale Be. If the current bale Be is not the first bale in the current stack SM and at least one preceding bale Be x in the current stack SM is positioned or has reached in a stacker 50 of the pulp baling system 10, a second set BprOp2 of bale properties of each of the at least one preceding bale B(c-x) in the current stack SM that is positioned in the stacker 50 is also included in the set of input parameters Sinput. Based on the obtained set of input parameters Sinput, the closed-loop controller 110 is configured to determine a desired compressed bale height HPdes for the current bale Be and to produce at least one predicted control signal Cpred that is adapted to be fed to the bale press 30 for controlling the bale press 30 to compress the current bale Be to the desired compressed bale height HPdes. The closed-loop controller 110 comprises a trained artificial neural network, ANN, with an input layer configured to obtain the set of input parameters Sinput, an output layer configured to provide the at least one predicted control signal Cpred, and at least one hidden layer interconnecting the input layer and the output layer, each of the input, output and at least one hidden layer comprising a respective set of nodes connected to nodes to the respective of neighboring layers via a respective weight.

[0052] Advantageously, the inventive controller 100 thereby controls the bale press such that the stack heights of stacks produced in the baling line are maintained within set tolerances in a much more efficient manner than previous solutions, by compensating for variations of properties of the bales in the stack in an improved manner.

[0053] The first set Bpropiof bale properties of the current bale Be comprise both preset desired physical properties / characteristics of the bale and the pulp that constitutes the bale, along with assessed corresponding parameters / properties that are registered by sensors upstream the bale press . Said sensors may include the sensors 31, 32, 33, 34, 34’, 34”, 35 and / or 36. Examples of such properties for the bale are bale heights at different positions along the baling line, weight, preliminary position in the stack, pulp properties such as dryness (moisture content), fiber length, bulk, freeness, brightness, pH, etc. The first set Bpropiof bale properties for the current bale includes a selection of any or all of these properties. Timestamps for registered bale height measurements at positions upstream the bale press may also be used, further explaining the swelling of the bale.

[0054] The second set Bprop2 of bale properties of each of the at least one preceding bale B(c- x) in the current stack SM that is positioned in the stacker 50 are properties obtained both upstream and downstream of the bale press 30 and include, in addition to any or all properties listed for the first set Bpropiof bale properties above, also a selection of any or all of height of the bale registered at different positions along the baling line in and downstream of the bale press using sensors that may include any or all of the sensors 37, 38, 41 and 42, the bale’s position in stack (if it has reached the stacker) and resulting bale height in stack (if it has reached the stacker). The second set BprOp2 of bale properties can also be updated to comprise any or all properties of received first, second and / or third output signals SI out, S2Out, S3Out and may comprises bale properties of each of the bales in the at least one preceding stack S(M-IJ in the unit U, in embodiments where more than one stack will form a unit U.

[0055] The number of height sensors arranged along the baling line can be selected to a lower number (preferably always including sensors 31, 37 or 38, and 41 or 42) to keep the cost and computational complexity lower, or a higher number to increase the resolution of the height and swell information to the trained ANN.

[0056] According to embodiments of the invention, the closed-loop controller 110 is configured to operate according to a PID regulation principle, a linear-quadratic regulation principle or a model predictive control principle to produce the at least one predicted control signal Cpred.

[0057] The at least one predicted control signal Cpred may typically be adapted to control at least one electric motor and / or at least one hydraulic actuator of the bale press 30, such that the at least one electric motor and / or hydraulic actuator causes the bale press 30 to compress the current bale to the desired compressed bale height. In particular, the at least one predicted control signal Cpred may be adapted to cause a respective control current and / or voltage to be produced, which respective control current and / or voltage has such a temporal profile with respect to magnitude and sign and / or is modulated in such a manner that the respective control current and / or voltage operates the at least one electric motor and / or the at least one hydraulic actuator to mechanically control a movable pressing element of the bale press 30 to be pushed downwards towards an opposing base of the bale press 30 to compress a pulp bale B positioned between the movable pressing element and the base. In other words, during a period when the at least one predicted control signal CPred controls a particular motor / actuator, the control current and / or voltage may vary over time and / or in terms of modulation such that components of the bale press 30 are controlled to move as intended. The predicted control signal Cpred may be indicative of a predicted press depth, Dp, and a predicted hold time, TH, for controlling the bale press 30 to compress the current bale Be to the desired compressed bale height HPdes by pressing the current bale Be to the predicted press depth Dp during the desired hold time, TH.

[0058] Alternatively the predicted control signal Cpred may be indicative of a predicted pressure, P, and a predicted hold time, TH, for controlling the bale press 30 to compress the current bale Be to the desired compressed bale height HPdes by applying the predicted pressure, P, on the current bale Be during the predicted hold time, TH, since press depth and applied pressure are interchangeable parameters in this context. When applied by the bale press 30, the predicted press depth Dp or predicted pressure P in combination with the predicted hold time TH are suitably adapted to compress the current bale Be to the desired compressed bale height HPdes-

[0059] Fig. 3 shows a block diagram of a controller 100 according to a second embodiment of the invention that differs from the first embodiment in that the controller 100 further comprises a first summation module 120 configured to derive a modified desired resulting bale height AHRdes for the current bale Be based on the desired resulting bale height HRdes of the current bale Be and at least one first output signal S lout from the stacker 50. The at least one first output signal Slout reflects a registered resulting bale height HR(c-x) of at least one preceding bale B(c-x) in the current stack SM or in a preceding stack S(M-IJ in the same unit U, depending on how many bales there are in the baling line between the bale press 30 and the stacker 50 and how many bales in the current stack have reached the stacker 50, because the feedback from the stacker 50 can always be used to adjust the control of the compression of the current bale Be, i.e. the bale next in line to enter the bale press 30. If there are many bales at different positions of the baling line at any given time, this will increase the production speed. However, if there are few bales positioned along the baling line at any given time, the feedback control can instead be improved, as the feedback will be more direct and possibly enabling controlling the compression and thus resulting height of bales based on registered resulting heights of more than one bale in the same unit or even the same stack, which is of course very valuable information for minimizing the deviation of the resulting stack height from the desired stack height.

[0060] The at least one registered resulting bale height HR(c-x) is registered by at least one height sensing device 41, 42 arranged at the stacker 50, and / or a registered resulting stack height HRSM of the current stack SM, wherein the resulting stack height HRSM is registered by at least one height sensing device 41, 42 arranged at the stacker 50. Since the number of bales in each stack, N, and the number of bales in the current stack that have reached the stacker 50 are known at any given time, and the height of the stack is continuously or discretely measured over time at the stacker 50 by any of the at least one height sensing device 41, 42 , the resulting bale height of each bale that has reached the stacker 50 and been placed in the current stack SM can be derived from the at least one first output signal SI out, at any given time.

[0061] If the combined resulting bale heights of the at least one preceding bale of the current stack differ from the combined desired bale heights, in other words indicating that the current stack SM risks having a resulting height that deviates to much from the desired stack height HSdes if bales continue to be produced according to the present settings, the first summation module 120 is thus suitably configured to determine the modified desired resulting bale height AHRdes for the current bale Be such that it will compensate for the detected deviation.

[0062] The closed-loop controller 110 is then in this second embodiment configured to set the desired resulting bale height HRdes to the modified desired resulting bale height AHRdes before determining the desired compressed bale height HPdes for the current bale Be and producing the at least one predicted control signal Cpred.

[0063] Advantageously, real-time compensation for any deviation in resulting bale heights of the bales within the current stack SM is thereby achieved, meaning that the overall aim of producing a stack with a height that is as close as possible to the set desired stack height HSdes can be attained.

[0064] The first output signal SI out may further reflect a resulting stack height HRS(M-I) of a preceding stack S(M-IJ that together with the current stack Sc will form a unit U of M stacks. If the resulting stack height HRS(M-I) of the preceding stack S(M-IJ differs from the desired stack height HSdes more than a first allowed deviation dl, the closed- loop controller 110 is in this case further configured to set the desired stack height HSdes for the current stack Sc to the resulting stack height HRS(M-I) of the preceding stack S(M-i) before producing the at least one predicted control signal Cpred. Thereby, any deviation greater than what is allowed within the predetermined tolerance defined by the preset first deviation d 1 is detected and appropriate measures can be taken. Suitably, real-time compensation for any variations and deviations in bale and stack height is thereby performed, which advantageously results in the stacks of a unit becoming more uniform than if only the preset reference height for stacks was used. This contributes to improving stability and safety in storage and transport, as well as reducing waste. After the last bale in a unit has been compressed, the desired stack height may suitably be reset to the initial desired stack height, i.e. the reference value for the baling system.

[0065] In this case, the second set BprOp2 of bale properties suitably further comprises bale properties of each of the bales in the at least one preceding stack S(M-X) in the unit U.

[0066] The closed-loop controller 110 may be configured to determine if the resulting stack height HRS(M-I) of the preceding stack S(M-IJ differs from the desired stack height HSdes more than the first allowed deviation d 1 by subtracting of one of the desired and resulting stack heights from each other and comparing of the subtraction result to the value of the first allowed deviation d 1.

[0067] Fig. 4 shows a block diagram of a controller 100 according to a third embodiment of the invention that differs from the first or second embodiment in that the closed- loop controller 110 is configured to obtain at least one second output signal S2out from the bale press 30, wherein the at least one second output signal S2out reflects a first registered compressed bale height HP(c-i) of the preceding bale B(c-i) registered by a height sensing device 37 connected to and arranged to register the press depth of the bale press 30, and that the closed-loop controller 110 is further configured to determine if the first registered compressed bale height HP(c-i) of the preceding bale B(c-i) differs from the desired compressed bale height HPdes(c-i) for the preceding bale B(c-i) more than a second allowed deviation d2.

[0068] The closed-loop controller 110 may be configured to determine if the first registered compressed bale height HP(c-ij of the preceding bale B(c-i) differs from the desired compressed bale height HPdes(c-i) for the preceding bale B(c-i) more than a second allowed deviation d2 by subtracting one of the desired and first registered compressed bale heights from each other and comparing the subtraction result to the preset value of the second allowed deviation d2.

[0069] Suitably, it can thereby be determined in runtime if the compression performed by the bale press 30, based on the predicted control signal Cpred, did not achieve a satisfactory compression result, i.e. did not reach the desired compressed bale height HPdes for the compressed bale, within the allowed tolerances defined by the second allowed deviation d2. If it is determined that a satisfactory compression result was not achieved, measures can thereby advantageously be taken in close to real time to compensate for the height deviation, by adjusting the compressed height of one or more subsequent bale in the same stack. For this purpose, the controller 100 may in this third embodiment further comprise a second summation module 130 configured to derive a modified desired compressed bale height AHPdes for the current bale Be based on the desired compressed bale height HPdes, the first registered compressed bale height HP(c-i) of the preceding bale B(c-i) and at least one third output signal S3out from at least one height sensing device 38 arranged directly downstream of the bale press 30, for measuring the height of a bale directly after it exits the bale press 30. The at least one third output signal S3out reflects a second registered compressed bale height HP’(c-ij of the preceding bale B(c-i) registered by said at least one height sensing device 38. If the first registered compressed bale height HP(c-ij of the preceding bale B(c-i) differs from the desired compressed bale height HPdes(c-i) for the preceding bale B(c-i) more than the second allowed deviation d2, the closed-loop controller 110 is here configured to obtain the modified desired compressed bale height AHPdes from the second summation module 130 and set the desired compressed bale height HPdes to the modified desired compressed bale height AHPdes before producing the at least one predicted control signal Cpred. Suitably, real-time compensation for any variations and deviations in bale height is thereby performed, in the form of direct feedback control for compressing the current bale based on the known control signal and compression result of the preceding bale, which has just exited the bale press. Advantageously, the real-time feedback and compensation results in increased accuracy in achieving the desired stack height, and further to stack heights becoming even more uniform. This contributes to improving stability and safety in storage and transport, as well as reducing waste.

[0070] If it is determined that the first registered compressed bale height HP(c-i) of the preceding bale B(c-i) differs from the desired compressed bale height HPdes(c-i) for the preceding bale B(c-i) more than a second allowed deviation d2, this means that the bale press 30 did not succeed in achieving the desired compressed height, within the allowed tolerance, for the preceding bale. The most likely reasons that this happens is that the physical limitations of the bale press 30 did not allow it to apply the required pressure or reach the required press depth indicated by the predicted control signal Cpred, or that the desired hold time, TH indicated by the predicted control signal Cpred exceeded a maximum allowed hold time of the bale press 30. The maximum allowed hold time of the bale press 30 may typically have been set to ensure efficiency of the baling line. If the problem with reaching the compressed height for the preceding bale was due to physical limitations of the bale press 30 and / or the maximum allowed hold time of the bale press being reached and exceeded, then the controller 100 may be configured to adjust one or both of these parameters before the current bale Be is compressed in the bale press 30. Therefore, if the first registered compressed bale height HP(c-i) of the preceding bale B(c-i) differs from the desired compressed bale height HPdes(c-i) for the preceding bale B(c-i) more than the second allowed deviation d2, the closed-loop controller 110 may further be configured to determine a modified maximum allowed hold time TH for the bale press 30, and / or determine that the bale press 30 needs to compress the current bale Be at least two times, to achieve the desired compressed bale height HPdes, and produce the at least one predicted control signal Cpred such that it is configured to control the bale press 30 to compress the current bale Be according to the modified maximum allowed hold time TH and / or to compress the current bale Be at least two times. The desired compressed bale height HPdes for the current bale Be, and any subsequent bales in the current stack, can thus suitably be modified to compensate for the height deviation of the preceding bale B(c-i). Direct compensation, within the current stack SM, to reach the desired stack height for the current stack SM, is thereby advantageously enabled.

[0071] The second summation module may in this third embodiment be configured to receive the desired compressed bale height HPdes from the closed-loop controller 110 or to obtain the second output signal S2out from the bale press 30, and further to receive the third output signal S3out from the bale press 30.

[0072] The input layer of the trained ANN may also be configured to receive other input, such as a first, second and third output signal SI out, S2 out, S3 out received from the stacker 50 or the bale press 30, and / or modified input parameters derived by a first or second summation module 120, 130 described below in connection with the second and third embodiments of the controller 100.

[0073] Turning now to Figs. 8, 9 and 10, embodiments of the invention will now be explained further in connection with a schematic pulp baling system 10, or baling line, according to the invention, wherein bales move from position Pl to position Pl 1 along the baling line.

[0074] Herein, for ease of explanation, the bales in the baling line are indexed in relation to the current bale Be, meaning that any bale preceding, i.e. being downstream of, the current bale Be in the baling line is referred to as B(c-x), wherein X is a positive integer indicative of its position in the baling line relative to the current bale Be. Similarly, any bale that is subsequent to, i.e. upstream of, the current bale Be in the baling line is referred to as B(c+xj, wherein X is a positive integer indicative of its position in the baling line relative to the current bale Be.

[0075] During operation of the pulp baling system 10, a plurality of bales B are present along the baling line. In the non-limiting example of Figs. 8 and 9, this is illustrated by eight bales B(c-2):(c+5) positioned in nine respective positions Pl to P9, from the bale former 20 (Pl) to the stacker 50 (P9). In the baling line, each bale B is first formed in one of at least one bale former 20, here illustrated by a single bale former 20 in position Pl. In this position, a first height Hl of the formed bale, here height H1(C+5) of bale B(c+5), is registered during pressing, by a height sensing device 31 that may be configured to register the press depth of the press in the former 20 and to derive the height of the formed bale therefrom. After exiting the bale former 20, each bale is then transported by a conveyor or other suitable transportation means to the bale press 30, at position P7 in Fig. 8. Between the bale former 20 and the bale press 30 the bale passes through a plurality of measurement positions P2- P6 where respective sensors 32, 33, 34, 34’, 34”, 35, 36 are arranged to register at least one property of the bale when it is located in the respective position along the baling line. In the example of Fig. 8, a weight measurement device 32 and / or a dryness measurement device 33 is / are arranged at position P2 for registering a first weight W1 and / or a first dryness DI of a bale at that position, in this example a first weight Wl(c+4j and a first dryness Dl(c+4j of the bale B(c+4). In the example of Fig. 8, there are further height sensing devices 34, 34’, 34” arranged for registering second, third and fourth heights H2, H3, H4, of a respective bale at positions P3, P4 and P5 along the baling line 10. At position P6, directly upstream of the bale press 30, a weight measurement device 35 and / or a height sensing device 36 may suitably be arranged for registering a second weight W2 and a fifth height H5 of a bale located at that position, before it enters the bale press for compressed to its desired compressed bale height HPdes at position P7. For the purpose of producing the at least one predicted control signal Cpred as described for embodiments herein, the current bale Be is the bale next in line to enter the bale press 30, i.e. the bale directly upstream of the bale press 30, as illustrated in Fig. 8. All measurements up to this point performed on the current bale Be, including any or all measurements in the positions P1-P6, may thus be included in the first set BprOpi of bale properties of the current bale Be.

[0076] As further shown in Fig. 8, a first registered compressed bale height HP(c-i) of the preceding bale B(c-i), which is in this example positioned in the bale press 30 in position P7, is measured during pressing by a height sensing device 37 that is configured to register the press depth of the bale press 30 and derive the first registered compressed bale height HP of the compressed bale therefrom. Thereafter, a second registered compressed bale height HP’, registered directly after compression, i.e. after the pressure of the bale press has been released, may be registered at position P8 by a height sensing device 38. In the example of Fig. 8, the bale B(C-2) has reached this position and is being measured. As described in connection with Fig. 4, this is especially advantageous if the compression of the bale that has just been compressed in the bale press 20 in position P7 did not achieve the desired compressed bale height HPdes within the allowed tolerances for that bale, whereby it is desirable to compensate for the faulty compressed height by adjusting the desired compressed bale height HPdes for the next bale in the current stack SM, i.e. the bale in position P6, before it is compressed in the bale press in position P7. Thereby, the overall objective of achieving the desired stack height HSdes of the current stack SM may still be achieved even when the desired compressed bale height HPdes is not achieved for each bale in the stack.

[0077] As previously discussed, and as indicated in Figs. 8 and 9 by double arrows indicating different swelling periods 1-6, each bale will swell after compression, which is why it is advantageous to register the height and weight of the bale at a plurality of positions along the baling line. It is further highly advantageous that the present invention includes inputting these registered heights in combination with all known other known properties of the bale and the material of the bale into the trained ANN, to continuously improve the learning and prediction of the trained ANN. How much the bale swells will, besides the properties of the bale, also depend on the time between the bale is formed in the bale former and the time that is it compressed in the bale press 30. Suitably, this parameter may be included in the set of input parameters Sinput to further improve the prediction result.

[0078] It is noted that the resulting bale height HRdes herein refers to the height of the bale as measured or registered in the stacker 50. Of course, the bale will continue to swell some after stacking, but this swelling will not significantly affect the final height of the bale and even less the uniformity of produced stacks and units, as the most-part of the swelling has already happened before the bales enter the stacker. In a simplified estimation, the rate of swelling after compression decreases close to exponentially after compression. Also, as all the bales will continue to swell after stacking, to a very low degree, the continued swelling will be similar from bale to bale meaning that the difference between them will not increase significantly. The current stack Sc is formed in the stacker 50 by stacking the N bales, including the current bale Be, when they reach the stacker 50. A stacking process with four bales, i.e. N=4, is illustrated in the non-limiting example of Fig. 9. As indicated by the timeline, Fig. 9 shows the same position P9 at different points in time, as bales enter the stacker 50 and are stacked on top of each other to form a current stack SM. AS can be seen from Fig. 9, the stacker 40 comprises lifting equipment 90, such as a forklift, for stacking bales on top of each other to form stacks. One or more height sensing device 41, 42 is arranged to continuously or discretely measure the resulting bale height HR(c-x) of the bale B(c-x) currently being added to the stack and / or the height of the stack at a given time. After the stack has been completed, i.e. all N bales have been stacked, the registered height of the stack is the resulting stack height HRSM of the current stack SM. In the example of Fig. 9, a sensor 41 is connected to the lifting equipment 90 and configured to continuously or discretely determine the height of the forks in relation to the base at any given time and a height sensor 42 is arranged to measure the height of the top of the stack compared to the base that it stands on at least once each time a new bale has been stacked. At least one height sensing device 41 and / or at least one height sensing device 42 , or any other suitably arranged height sensing device, may be used. From this information that is registered over time, the height of the stack at any given moment as well as the resulting height of each thereto stacked bale can be derived.

[0079] As previously mentioned, in the context of this disclosure the current bale Be is always the bale next in line to be compressed by the bale press 30. When this bale has been compressed and the height of it has been registered in the bale press 30 and optionally also at the subsequent measuring position P8 directly after exiting the bale press 30, the next bale in line B(c-i) becomes the current bale Be for which the desired compressed height HPdes is to be determined, and a new predicted control signal Cpred is thus produced for the new current bale Be.

[0080] The preceding bale B(c-i) refers to the bale directly preceding the current bale Be, i.e. the bale directly downstream of the current bale Be in the baling line 10. There may also be preceding bales Be x in the baling line, at the stacker 50 forming part of the current stack SM, or forming part of one or more preceding stacks at the stacker 50 or that have exited the stacker, i.e. been transported by a conveyor or the like to a next processing step such as a unit former 60, to a storage, or to be packed and shipped.

[0081] A schematic example of a unit former 60 is shown in Fig. 10 wherein two stacks SM, S(M-I) pass from position PIO to position Pl l where they are bound together, or in another suitable manner attached to each other, or possibly just placed together, to form a unit U. In the non-limiting example of Fig. 10, M=2, i.e. the unit U comprises 2 stacks with a total of M*N = 8 bales.

[0082] Each height sensing device 31, 34, 34’, 34”, 36, 37, 38, 41, 42 herein may be any suitable type of sensor configured to register height.

[0083] Each weight measurement device 32, 35 herein may be any suitable type of sensor configured to register weight.

[0084] Each dryness measurement device 33 herein may be any suitable type of sensor configured to register dryness.

[0085] The weights of the trained ANN have typically been assigned through a training process in which the set of input parameters Sinput, the at least one first output signal Slout and, in some embodiments, one or both of the at least one second output signal S2out and the at least one third output signal S3out were used as training data and registered control signals configured to control the bale press of the pulp baling system were used as reference data. Referring now to Fig. 5, according to one embodiment of the invention, the weights of the trained ANN have been determined iteratively via a backpropagation training process {P} executed in a training unit 200. The training unit 200 repeatedly obtains updates of the set of input parameters Sinput, the at least one first output signal SI out and / or the at least one second output signal S2out and / or the at least one third output signal S3out, and the associated registered control signals Creg. Based on the set of input parameters Sinput, the at least one first output signal S l out and / or the at least one second output signal S2out and / or the at least one third output signal S3out, the ANN under training 110’ produces at least one predicted control signal Cpred’, which aims at being sufficiently similar to the registered control signals Cregassociated with the set of input parameters Sinput, the at least one first output signal S l out and / or the at least one second output signal S2out and / or the at least one third output signal S3out.

[0086] An evaluation module 210 is configured to check if a difference A between the registered control signals Cregand the at least one predicted control signal Cpred’ is less than a threshold value eth. If the evaluation module 210 finds that said difference A is equal to or larger than the threshold value eth, the evaluation module 210 is configured to generate a set of adjustment parameters {P}, which causes one or more of the weights in the ANN to be modified to a respective higher or lower value that are expected to lower the difference A. This backpropagation training process continues until a convergence criterion is met. In simplified terms this may be said to occur when the difference A becomes smaller than the threshold value eth. The backpropagation training process comprises a suitable number of epochs that enables reliable convergence and at the same time avoids overtraining. After the training process has been completed, the ANN under training 110’ represents the trained ANN in the closed-loop module 110. The number of hidden layers in the trained ANN is at least one.

[0087] The trained artificial neural network, ANN, may use a Long Short-Term Memory, LSTM, algorithm.

[0088] Fig. 6 illustrates a block diagram of the controller 100 according to one embodiment of the invention. It is generally advantageous if the controller 100 is configured to effect the procedure described herein in an automatic manner by executing a computer program 653 in a processing device 651, which includes at least one processing unit. The processing device 651 is communicatively connected to a memory unit 655, i.e. non-volatile data carrier, storing a computer program 653, which, in turn, contains software for making the processing device 651 execute the actions mentioned in this disclosure when the computer program 653 is run on the at least one processing unit in the processing device 651. According to this embodiment of the invention, the trained ANN in the closed-loop module 110 is preferably implemented by means of a computer program that runs on the processing device 651. According to another embodiment of the invention, the trained ANN is instead implemented in hardware, such as in one or more neuromorphic circuit, i.e. mixed- signal integrated circuit containing both analog circuits and digital circuits, which aims at mimicking biological neural functions.

[0089] To sum up, and with reference to the flow diagrams in Figs. 7, 11, 12 and 13, we will now describe the computer-implemented method according to the invention for controlling a bale press in a pulp baling system to compress a current bale Be to have a desired compressed bale height HPdes.

[0090] Fig. 7 illustrates the general method, including:

[0091] In step 710: checking if a set of input parameters Sinput has been received.

[0092] If so, steps 720 to 740 and optionally 750 follow. Otherwise, the procedure loops back, and stays in step 710.

[0093] In step 720: obtaining, in a closed-loop controller 110, a set of input parameters Sinput*

[0094] The set of input parameters Sinput is the set of input parameters as described in connection with any of the embodiments of the controller 100 herein, and thus comprises at least a desired resulting bale height HRdes or a modified desired resulting bale height AHRdes of the current bale Be, a desired stack height HSdes of a current stack SM of N bales that the current bale Be will be part of, a first set BprOpi of bale properties of the current bale Be and, if the current bale Be is not the first bale in the current stack SM and at least one preceding bale Be x in the current stack SM is positioned in a stacker 50 of the pulp baling system 10, a second set BprOp2 of bale properties of each of the at least one preceding bale Be x in the current stack SM that is positioned in the stacker 50.

[0095] In step 730: determining, by the closed-loop controller 110, a desired compressed bale height HPdes for the current bale Be based on the obtained set of input parameters Sinput.

[0096] In step 740: producing, by the closed-loop controller 110, at least one predicted control signal Cpred that is adapted to be fed to the bale press 30 for controlling the bale press 30 to compress the current bale Be to the desired compressed bale height HPa es*

[0097] The method may further comprise, in a step 750: feeding the at least one predicted control signal Cpred from the controller 100 to the bale press 30 to control the bale press 30 to compress the current bale Be to the desired compressed bale height HPdes.

[0098] Figs. 11-13 illustrate different embodiments of the method of Fig. 7.

[0099] Turning first to Fig. 11, in this embodiment the method comprises, before determining the desired compressed bale height HPdes for the current bale Be and producing the at least one predicted control signal Cpred:

[0100] In step 1110: checking if at least one first output signal Slout has been received.

[0101] If so, steps 1120 to 1140 follow. Otherwise, the procedure loops back, and stays in step 1110.

[0102] In step 1120: obtaining at least one first output signal Slout from the stacker 50.

[0103] The at least one first output signal Slout is the at least one first output signal as described in connection with any of the embodiments of the controller 100 herein, and thus reflects a registered resulting bale height HRc-x of at least one preceding bale Be x in the current stack SM or in a preceding stack SM-I in the same unit U, depending on how many bales there are in the baling line between the bale press 30 and the stacker 50 and how many bales in the current stack have reached the stacker 50, because the feedback from the stacker 50 can always be used to adjust the control of the compression of the current bale Be, i.e. the bale next in line to enter the bale press 30.

[0104] The at least one registered resulting bale height HRc-x is registered by at least one height sensing device 41, 42 arranged at the stacker 50; or a registered resulting stack height HRSM, HRSM-I of the current stack SM or the preceding stack SM-I in the same unit U, wherein the resulting stack height HRSM, HRSM-I is registered by at least one height sensing device 41, 42 arranged at the stacker 50. In step 1130: deriving, in a first summation module 120, a modified desired resulting bale height AHRdes for the current bale Be based on the desired resulting bale height HRdes of the current bale Be and the at least one first output signal SI out.

[0105] This step may comprise obtaining, in the first summation module 120, the desired resulting bale height HRdes of the current bale Be from the closed-loop controller 110.

[0106] In step 1140: setting, by the closed-loop controller 110, the desired resulting bale height HRdes to the modified desired resulting bale height AHRdes.

[0107] Suitably, if information in the obtained at least one first output signal SI out indicates that the current stack SM risks having a resulting height that deviates to much from the desired stack height HSdes if bales continue to be produced according to the present settings, the method according to this embodiment thus suitably determines the modified desired resulting bale height AHRdes for the current bale Be such that it will compensate for the detected deviation and sets the desired resulting bale height AHRdes to this compensated value before the desired compressed bale height HPdes for the current bale Be is determined and the at least one predicted control signal Cpred is produced. Advantageously, real-time compensation for any deviation in resulting bale heights of the bales within the current stack SM is thereby achieved, meaning that the overall aim of producing a stack with a height that is as close as possible to the set desired stack height HSdes can be attained.

[0108] If the first output signal SI out further reflects a resulting stack height HRSM-I of a preceding stack SM-I that together with the current stack Sc will form a unit U of M stacks the method may also comprise the steps of Fig. 12 before determining the desired compressed bale height HPdes for the current bale Be and producing the at least one predicted control signal Cpred. In this embodiment, the method comprises:

[0109] In step 1210: determining, by the closed-loop controller 110, if the resulting stack height HRS(M-I) of the preceding stack S(M-IJ differs from the desired stack height HSdes more than a first allowed deviation d 1. If so, the procedure continues in step 1220. Otherwise, the procedure loops back, and stays in step 1210.

[0110] The determination of step 1210 may comprise subtracting one of the desired and resulting stack heights from each other and comparing the subtraction result to the preset value of the first allowed deviation dl.

[0111] In step 1220: setting, by the closed-loop controller 110, the desired stack height HSdes for the current stack Sc to the resulting stack height HRS(M-I) of the preceding stack S(M-I) before producing the at least one predicted control signal Cpred.

[0112] Suitably, real-time compensation for any variations and deviations in bale and stack height is thereby performed, which advantageously results in the stacks of the unit U becoming more uniform than if only the preset reference height for the stacks was used. This contributes to improving stability and safety in storage and transport, as well as reducing waste. After the last bale in the unit U has been compressed, the method may comprise resetting the desired stack height HSdes to the initial desired stack height, i.e. the reference value for the baling system.

[0113] In the flow diagram of Fig. 13, further embodiments that include determining if the preceding bale B(c-i) was not compressed satisfactorily and directly compensation for this before producing the at least one predicted control signal Cpred for the current bale Be are illustrated. Any embodiment of Fig. 13 may be combined with any embodiment described in connection with Figs. 7, 11 and / or 12.

[0114] As shown in the embodiments of Fig. 13, the method comprises:

[0115] In step 1310: checking if at least one second output signal S2out has been received.

[0116] If so, steps 1320 to 1340 follow. Otherwise, the procedure loops back, and stays in step 1310.

[0117] In step 1320: obtaining in the closed-loop controller 110, at least one second output signal S2out from the bale press 30.

[0118] The at least one second output signal S2out is the at least one second output signal as described in connection with any of the embodiments of the controller 100 herein, and thus reflects a first registered compressed bale height HP(c-i) of the preceding bale B(c-i) registered by a height sensing device 37 connected to and arranged to register the press depth of the bale press 30.

[0119] In step 1330: comparing the first registered compressed bale height HP(c-i) of the preceding bale B(c-i) to a previously determined desired compressed bale height HPdes(c-i) for the preceding bale B(c-i).

[0120] In step 1340: determining if the first registered compressed bale height HP(c-ij of the preceding bale B(c-i) differs from the desired compressed bale height HPdes(c-i) for the preceding bale B(c-i) more than a second allowed deviation d2.

[0121] This may comprise subtracting one of the desired and first registered compressed bale heights from each other and comparing the subtraction result to the preset value of the second allowed deviation d2.

[0122] Suitably, it can thereby be determined in runtime if the compression performed by the bale press 30, based on the predicted control signal Cpred, did not achieve a satisfactory compression result, i.e. did not reach the desired compressed bale height HP for the compressed bale, within the allowed tolerances defined by the second allowed deviation d2.

[0123] If it is determined that a satisfactory compression result was not achieved, measures can thereby advantageously be taken in close to real time to compensate for the height deviation, by adjusting the compressed height of one or more subsequent bale in the same stack. The method then continues with the embodiment described in steps 1350 to 1370, or the embodiment described in steps 1380 to 1390, or both.

[0124] The method may thus comprise:

[0125] In step 1350: obtaining, in a second summation module 130, at least one third output signal S3out from at least one height sensing device 38 arranged directly downstream of the bale press 30, which at least one third output signal S3out reflects a second registered compressed bale height HP’(c-i) of the preceding bale B(c- i) registered by said at least one height sensing device 38. In step 1360: deriving, in the second summation module 130, a modified desired compressed bale height AHPdes for the current bale Be based on the desired compressed bale height HPdes, the first registered compressed bale height HP(c-i) of the preceding bale B(c-i) and the at least one third output signal S3out.

[0126] In step 1370: setting, by the closed-loop controller 110, the desired compressed bale height HPdes to the modified desired compressed bale height AHPdes.

[0127] As steps 1350 to 1370 are performed before producing the at least one predicted control signal Cpred, real-time compensation for any variations and deviations in bale height is suitably achieved, in the form of direct feedback control for compressing the current bale based on the known control signal and compression result of the preceding bale, which has just exited the bale press. Advantageously, the real-time feedback and compensation results in increased accuracy in achieving the desired stack height, and further to stack heights becoming even more uniform. This contributes to improving stability and safety in storage and transport, as well as reducing waste.

[0128] Alternatively, or additionally, the method may, as illustrated in Fig. 13, comprise step 1380 and / or step 1390, comprising:

[0129] In step 1380: determining, by the closed-loop controller 110, a modified maximum allowed hold time TH for the bale press 30.

[0130] In step 1390: determining, by the closed-loop controller 110, that the bale press 30 needs to compress the current bale Be at least two times.

[0131] Step 740 of producing, by the closed-loop controller 110, the at least one predicted control signal Cpred then comprises producing the at least one predicted control signal CPred such that it is configured to control the bale press 30 to compress the current bale Be according to the modified maximum allowed hold time TH and / or to compress the current bale Be at least two times, to achieve the desired compressed bale height HPd es*

[0132] Direct compensation, within the current stack SM, to reach the desired stack height for the current stack SM, is thereby advantageously enabled. As described herein, adjusting, typically increasing, the allowed hold time TH and / or allowing for each bale to be compressed more than one time is advantageous if the reason that the compression performed by the bale press 30 on the preceding bale was not satisfactory is due to physical limitations of the bale press 30, such as maximum pressure or press depth to be applied.

[0133] After the current bale has been compressed, or after the last bale in the current stack or unit has been compressed, the maximum allowed hold time for the bale press and / or number of compressions for each bale may suitably be reset to the initial values, i.e. the reference value for the baling system.

[0134] The method according to any embodiment herein may further comprise, in a step preceding the steps of Figs. 7 and 11-13, assigning the weights of the trained ANN through a training process in which the set of input parameters Sinput, the at least one first output signal Slout and optionally also one or both of the at least one second output signal S2out and the at least one third output signal S3out are used as training data, and registered control signals Cregconfigured to control the bale press 30 of the pulp baling system 10 are used as reference data. The method may comprise determining the weights of the trained ANN iteratively via a backpropagation training process {P} comprising comparing training data that express the registered control signals Cregwith the at least one predicted control signal Cpred produced by an ANN under training 110’, which ANN under training 110’ represents the trained ANN after that the training process has been completed.

[0135] Advantages of the method embodiments are apparent from the description of corresponding embodiments of the controller 100 and pulp baling system 10 herein.

[0136] In another aspect of the invention, there is provided a computer program 653 loadable into a non-volatile data carrier 655 communicatively connected to a processing unit 651, the computer program 653 comprising software for executing the method according to any of the embodiments described in connection with Figs. 7, 11, 12 or 13, when the computer program 653 is run on the processing unit 651.

[0137] In a further aspect of the invention, there is provided a non-volatile data carrier 655 containing the computer program 653. Variations to the disclosed embodiments can be understood and effected by those skilled in the art in practicing the claimed invention, from a study of the drawings, the disclosure, and the appended claims. It is to be noted that features from the various embodiments described herein may freely be combined, unless it is explicitly stated that such a combination would be unsuitable.

[0138] The term “comprises / comprising” when used in this specification is taken to specify the presence of stated features, integers, steps or components. The term does not preclude the presence or addition of one or more additional elements, features, integers, steps or components or groups thereof. The indefinite article "a" or "an" does not exclude a plurality. In the claims, the word “or” is not to be interpreted as an exclusive or (sometimes referred to as “XOR”). On the contrary, expressions such as “A or B” covers all the cases “A and not B”, “B and not A” and “A and B”, unless otherwise indicated. The mere fact that certain measures are recited in mutually different dependent claims does not indicate that a combination of these measures cannot be used to advantage. Any reference signs in the claims should not be construed as limiting the scope.

Claims

CLAIMS1) A controller (100) for controlling a bale press (30) in a pulp baling system (10) to compress a current bale (Be) to have a desired compressed bale height (HPdes), the controller (100) comprising: a closed-loop controller (110) configured to: obtain a set of input parameters (Sinput) comprising a desired resulting bale height (HRdes) or a modified desired resulting bale height (AHRdes) of the current bale (Be), a desired stack height (HSdes) of a current stack (SM) of N bales that the current bale (Be) will be part of, a first set (BprOpi) of bale properties of the current bale (Be) and, if the current bale (Be) is not the first bale in the current stack (SM) and at least one preceding bale (Bc- x) in the current stack (SM) is positioned in a stacker (50) of the pulp baling system (10), a second set (BprOp2) of bale properties of each of the at least one preceding bale (B(c-xj) in the current stack (SM) that is positioned in the stacker (50); and based on the obtained set of input parameters (Sinput) determine a desired compressed bale height (HPdes) for the current bale (Be) and produce at least one predicted control signal (Cpred) that is adapted to be fed to the bale press (30) for controlling the bale press (30) to compress the current bale (Be) to the desired compressed bale height (HPdes) ; wherein the closed-loop controller (110) comprises a trained artificial neural network, ANN, which comprises: an input layer configured to obtain the set of input parameters (Sinput) , an output layer configured to provide the at least one predicted control signal (Cpred) , and at least one hidden layer interconnecting the input layer and the output layer, each of the input, output and at least one hidden layer comprising arespective set of nodes connected to nodes to the respective of neighboring layers via a respective weight.2) The controller (100) of claim 1, wherein the predicted control signal (Cpred) is indicative of a predicted press depth, Dp, and a predicted hold time, TH, for controlling the bale press (30) to compress the current bale (Be) to the desired compressed bale height (HPdes) by pressing the current bale (Be) to the predicted press depth (Dp) during the desired hold time, TH.3) The controller (100) of claim 1, wherein the predicted control signal (Cpred) is indicative of a predicted pressure, P, and a predicted hold time, TH, for controlling the bale press (30) to compress the current bale (Be) to the desired compressed bale height (HPdes) by applying the predicted pressure, P, on the current bale (Be) during the predicted hold time, TH.4) The controller (100) of any one of the preceding claims, further comprising a first summation module (120) configured to derive a modified desired resulting bale height (AHRdes) for the current bale (Be) based on the desired resulting bale height (HRdes) of the current bale (Be) and at least one first output signal (SI out) from the stacker (50), which at least one first output signal (SI out) reflects: a registered resulting bale height (HR(c-xj) of at least one preceding bale (B(c-xj) in the current stack (SM) or in a preceding stack (S(M-IJ) in the same unit (U), wherein the at least one registered resulting bale height (HR(c-xj) is registered by at least one height sensing device (41, 42) arranged at the stacker (50); or a registered resulting stack height (HRSM, HRS(M-IJ) of the current stack (SM) or the preceding stack (S(M-IJ) in the same unit (U), wherein the resulting stack height (HRSM, HRS(M-IJ) is registered by at least one height sensing device (41, 42) arranged at the stacker (50), wherein the closed-loop controller (110) is configured to set the desired resulting bale height (HRdes) to the modified desired resulting bale height (AHRdes) before determining the desired compressed bale height (HPdes) for the current bale (Be) and producing the at least one predicted control signal (Cpred).5) The controller (100) of claim 4, wherein the first output signal (SI out) further reflects a resulting stack height (HRS(M-IJ) of a preceding stack (S(M-IJ) that together with the current stack (Sc) will form a unit (U) of M stacks and wherein, if the resulting stack height (HRS(M-IJ) of the preceding stack (S(M-IJ) differs from the desired stack height (HSdes) more than a first allowed deviation (dl), the closed-loop controller (110) is further configured to set the desired stack height (HSdes) for the current stack (Sc) to the resulting stack height (HRS(M-IJ) of the preceding stack (S(M-IJ) before producing the at least one predicted control signal (Cprcd) •6) The controller (100) of claim 5, wherein the second set (BprOp2) of bale properties further comprises bale properties of each of the bales in the preceding stack (S(M- i)) in the unit (U).7) The controller (100) of any one of the preceding claims, wherein the closed-loop controller (110) is further configured to: obtain at least one second output signal (S2Out) from the bale press (30), wherein the at least one second output signal (S2Out) reflects a first registered compressed bale height (HP(c-ij) of the preceding bale (B(c-ij) registered by a height sensing device (37) connected to and arranged to register the press depth of the bale press (30); and determine if the first registered compressed bale height (HP(c-ij) of the preceding bale (B(c-ij) differs from the desired compressed bale height (HPdes<c- i)) for the preceding bale (B(c-ij) more than a second allowed deviation (d2).8) The controller (100) of claim 7, further comprising a second summation module (130) configured to derive a modified desired compressed bale height (AHPdes) for the current bale (Be) based on the desired compressed bale height (HPdes), the first registered compressed bale height (HP(c-ij) of the preceding bale (B(c-ij) and at least one third output signal (S3out) from at least one height sensing device (38) arranged directly downstream of the bale press (30), which at least one third output signal (S3out) reflects a second registered compressed bale height (HP’(c-ij)of the preceding bale (B(c-ij) registered by said at least one height sensing device (38), wherein, if the first registered compressed bale height (HP(c-ij) of the preceding bale (B(c-ij) differs from the desired compressed bale height (HPdes(c-i)) for the preceding bale (B(c-ij) more than the second allowed deviation (d2), the closed- loop controller (110) is configured to obtain the modified desired compressed bale height (AHPdes) from the second summation module (130) and set the desired compressed bale height (HPdes) to the modified desired compressed bale height (AHPdes) before producing the at least one predicted control signal (Cpred) .9) The controller (100) of claim 7 or 8, wherein, if the first registered compressed bale height (HP(c-ij) of the preceding bale (B(c-ij) differs from the desired compressed bale height (HPdes(c-i)) for the preceding bale (B(c-ij) more than the second allowed deviation (d2), the closed-loop controller (110) is further configured to: determine a modified maximum allowed hold time (TH) for the bale press (30), and / or determine that the bale press (30) needs to compress the current bale (Be) at least two times, to achieve the desired compressed bale height (HPdes) , and produce the at least one predicted control signal (Cpred) such that it is configured to control the bale press (30) to compress the current bale (Be) according to the modified maximum allowed hold time (TH) and / or to compress the current bale (Be) at least two times.10)The controller (100) of any one of the preceding claims, wherein the weights of the trained ANN have been assigned through a training process in which the set of input parameters (Sinput) and the at least one first output signal (Slout) were used as training data and registered control signals (Creg) configured to control the bale press (30) of the pulp baling system (10) were used as reference data.1 l)The controller (100) of any one of the claims 1 to 9, wherein the weights of the trained ANN have been assigned through a training process in which the set of input parameters (Sinput), the at least one first output signal (Slout) and one ormore of the at least one second output signal (S2Out) and the at least one third output signal (S3out) were used as training data and registered control signals (Creg) configured to control the bale press (30) of the pulp baling system (10) were used as reference data.12)The controller (100) of claim 10 or 11, wherein the weights of the trained ANN have been determined iteratively via a backpropagation training process ({P}) comprising: comparing training data that express the registered control signals (Creg) with the at least one predicted control signal (Cpred’) produced by an ANN under training (110’), which ANN under training (110’) represents the trained ANN after that the training process has been completed.13)The controller (100) of any one of the preceding claims, wherein the trained artificial neural network, ANN, uses a Long Short-Term Memory, LSTM, algorithm.14)The controller (100) of any one of the preceding claims, wherein the trained ANN is implemented by means of a computer program run on at least one processing unit (651).15)The controller (100) of any one of the claims 1 to 13, wherein the trained ANN is implemented on at least one neuromorphic circuit.16)The controller (100) according to any one of the preceding claims, wherein the closed-loop controller (110) is configured to operate according to a proportional- integral-derivative, PID, regulation principle, a linear-quadratic regulation principle or a model predictive control principle.17)The controller (100) according to any one of the preceding claims, wherein the at least one predicted control signal (Cpred) is adapted to control at least one electric motor and / or at least one hydraulic actuator of the bale press (30), such that the at least one electric motor and / or hydraulic actuator causes the bale press (30) to compress the current bale (Be) to the desired compressed bale height (HPdes).18) A pulp baling system (10) comprising a bale former (20) for forming pulp (P) into a bale (B), a bale press (30) for pressing the formed bale (B), a stacker (50) for stacking N bales (B) into a stack (S), at least two height sensing devices (31, 34, 34’, 34”, 36, 37, 38, 41, 42) arranged at different positions along the pulp baling system (10), the positions at least including at the bale press (30) and at the stacker (50), for registering heights of bales (B) positioned at the different positions in the pulp baling system (10), and a controller ( 100) according to any one of the claims 1-16 for controlling the bale press (30) to compress a current bale (Be) to have a desired compressed bale height (HPdes) .19)The pulp baling system (10) of claim 18, further comprising a wrapping section (40) downstream of the bale press (30) and upstream of the stacker (50), for wrapping and marking each bale (BN) before it is stacked.20)The pulp baling system (10) of claim 18 or 19, further comprising a unit former (60) downstream of the stacker (50), for forming a unit (U) comprising at least two stacks (SI, S2).21) A computer implemented method for controlling a bale press (30) in a pulp baling system (10) to compress a current bale (Be) to have a desired compressed bale height (HPdes), the method comprising: obtaining, in a closed-loop controller (110), a set of input parameters (Sinput) comprising a desired resulting bale height (HRdes) or a modified desired resulting bale height (AHRdes) of the current bale (Be), a desired stack height (HSdes) of a current stack (SM) of N bales that the current bale (Be) will be part of, a first set (BprOpi) of bale properties of the current bale (Be) and, if the current bale (Be) is not the first bale in the current stack (SM) and at least one preceding bale (B(c-xj) in the current stack (SM) is positioned in a stacker (50) of the pulp baling system (10), a second set (BprOp2) of bale properties of each of the at least one preceding bale (B(c-xj) in the current stack (SM) that is positioned in the stacker (50); based on the obtained set of input parameters (Sinput) determining, by the closed-loop controller (110), a desired compressed bale height (HPdes) for thecurrent bale (Be) and producing, by the closed-loop controller (110), at least one predicted control signal (Cpred) that is adapted to be fed to the bale press (30) for controlling the bale press (30) to compress the current bale (Be) to the desired compressed bale height (HPdes) ; wherein the closed-loop controller (110) comprises a trained artificial neural network, ANN, which comprises: an input layer configured to obtain the set of input parameters (Sinput), an output layer configured to provide the at least one predicted control signal (Cpred) , and at least one hidden layer interconnecting the input layer and the output layer, each of the input, output and at least one hidden layer comprising a respective set of nodes connected to nodes to the respective of neighboring layers via a respective weight. )The method of claim 21, further comprising: deriving, in a first summation module (120), a modified desired resulting bale height (AHRdes) for the current bale (Be) based on the desired resulting bale height (HRdes) of the current bale (Be) and at least one first output signal (SI out) from the stacker (50), which at least one first output signal (SI out) reflects: a registered resulting bale height (HR(c-xj) of at least one preceding bale (B(c-xj) in the current stack (SM) or in a preceding stack (S(M-IJ) in the same unit (U), wherein the at least one registered resulting bale height (HR(c-xj) is registered by at least one height sensing device (41, 42) arranged at the stacker (50); or a registered resulting stack height (HRSM, HRS(M-IJ) of the current stack (SM) or the preceding stack (S(M-IJ) in the same unit (U), wherein the resulting stack height (HRSM, HRS(M-IJ) is registered by at least one height sensing device (41, 42) arranged at the stacker (50); andsetting, by the closed-loop controller (110), the desired resulting bale height (HRdes) to the modified desired resulting bale height (AHRdes) before determining the desired compressed bale height (HPdes) for the current bale (Be) and producing the at least one predicted control signal (Cpred) .23)The method of claim 22, wherein the first output signal (Slout) further reflects a resulting stack height (HRS(M-IJ) of a preceding stack (S(M-IJ) that together with the current stack (Sc) will form a unit (U) of M stacks and wherein the method further comprises: determining, by the closed-loop controller (110), if the resulting stack height (HRS(M-I)) of the preceding stack (S(M-IJ) differs from the desired stack height (HSdes) more than a first allowed deviation (dl), and, if it does, setting, by the closed-loop controller (110), the desired stack height (HSdes) for the current stack (Sc) to the resulting stack height (HRS(M-IJ) of the preceding stack (S(M-IJ) before producing the at least one predicted control signal (Cpred) .24)The method of any one of the claims 21 to 23, further comprising: obtaining, in the closed-loop controller (110), at least one second output signal (S2Out) from the bale press (30), which at least one second output signal (S2Out) reflects a first registered compressed bale height (HP(c-ij) of the preceding bale (B(c-ij) registered by a height sensing device (37) connected to and arranged to register the press depth of the bale press (30); comparing the first registered compressed bale height (HP(c-ij) of the preceding bale (B(c-ij) to a previously determined desired compressed bale height (HPdes(c-i)) for the preceding bale (B(c-ij); and determining if the first registered compressed bale height (HP(c-ij) of the preceding bale (B(c-ij) differs from the desired compressed bale height (HPdes<c- i)) for the preceding bale (B(c-ij) more than a second allowed deviation (d2).25)The method of claim 24, further comprising, if the first registered compressed bale height (HP(c-ij) of the preceding bale (B(c-ij) differs from the desiredcompressed bale height (HPdes(c-i)) for the preceding bale (B(c-ij) more than the second allowed deviation (d2): obtaining, in a second summation module (130), at least one third output signal (S3Out) from at least one height sensing device (38) arranged directly downstream of the bale press (30), which at least one third output signal (S3Out) reflects a second registered compressed bale height (HP’(c-i)) of the preceding bale (B(c-ij) registered by said at least one height sensing device (38); deriving, in the second summation module (130), a modified desired compressed bale height (AHPdes) for the current bale (Be) based on the desired compressed bale height (HPdes), the first registered compressed bale height (HP(c-i)) of the preceding bale (B(c-ij) and the at least one third output signal (S3out); and setting, by the closed-loop controller (110), the desired compressed bale height (HPdes) to the modified desired compressed bale height (AHPdes) before producing the at least one predicted control signal (Cpred) . )The method of claim 24 or 25, further comprising, if the first registered compressed bale height (HP(c-ij) of the preceding bale (B(c-ij) differs from the desired compressed bale height (HPdes(c-i)) for the preceding bale (B(c-ij) more than the second allowed deviation (d2): determining, by the closed-loop controller (110), a modified maximum allowed hold time (TH) for the bale press (30), and / or determining that the bale press (30) needs to compress the current bale (Be) at least two times, to achieve the desired compressed bale height (HPdes), and producing, by the closed-loop controller (110), the at least one predicted control signal (Cpred) such that it is configured to control the bale press (30) to compress the current bale (Be) according to the modified maximum allowed hold time (TH) and / or to compress the current bale (Be) at least two times, to achieve the desired compressed bale height (HPdes) .27)The method of any one of the claims 21 to 26, further comprising feeding the at least one predicted control signal (Cpred) from the controller (100) to the bale press (30) to control the bale press (30) to compress the current bale (Be) to the desired compressed bale height (HPdes).28)The method of any one of the claims 21 to 27, further comprising, in a preceding step, assigning the weights of the trained ANN through a training process in which the set of input parameters (Sinput) and the at least one first output signal (SI out) are used as training data and registered control signals (Creg) configured to control the bale press (30) of the pulp baling system (10) are used as reference data.29)The method of any one of the claims 21 to 27, further comprising, in a preceding step, assigning the weights of the trained ANN through a training process in which the set of input parameters (Sinput), the at least one first output signal (Slout) and one or more of the at least one second output signal (S2Out) and the at least one third output signal (S3out) are used as training data and registered control signals (Creg) configured to control the bale press (30) of the pulp baling system (10) are used as reference data.30)The method of claim 28 or 29, further comprising determining the weights of the trained ANN iteratively via a backpropagation training process ({P}) comprising comparing training data that express the registered control signals (Creg) with the at least one predicted control signal (Cpred) produced by an ANN under training (110’), which ANN under training (110’) represents the trained ANN after that the training process has been completed.31)A computer program (653) loadable into a non-volatile data carrier (655) communicatively connected to a processing unit (651) of a controller (100), the computer program (653) comprising software for executing the method according to any one of the claims 21 to 30 when the computer program (653) is run on the processing unit (651).32)A non-volatile data carrier (655) containing the computer program (653) of claim 31.

Images