Search

EP-3994461-B1 - METHOD, SYSTEM AND PAPERBOARD PRODUCTION MACHINE FOR ESTIMATING PAPERBOARD QUALITY PARAMETERS

EP3994461B1EP 3994461 B1EP3994461 B1EP 3994461B1EP-3994461-B1

Inventors

  • Alpire, Adam
  • Eisenstecken, Thomas
  • HILDEBRAND, MICHAEL
  • Kisslinger, Ferdinand

Dates

Publication Date
20260513
Application Date
20200819

Claims (13)

  1. Computer-implemented method for estimating at least one quality parameter of paperboard produced in a paperboard production subprocess (PP) of a paperboard processing pipeline (P) during the production by a data-driven module (DD) comprising a preprocessing module (PM) and a machine-learning module (ML), wherein - sensor data (D1 to D5, D1' to D5') is acquired from at least one processing step (P1 to P5) of the processing pipeline (P) and transferred to a data repository (R), - at least one historical feature (F') is determined by a preprocessing module (PM) by at least partially evaluating historical sensor data (D1' to D5') that was acquired during at least one previously produced batch of paperboard and retrieved from the data repository (R), - a machine-learning module (ML) is trained to reproduce from the at least one historical feature (F') a target value of at least one quality parameter, wherein the target value is determined for a previously produced batch of paperboard corresponding to the historical sensor data (D1' to D5') from which the historical feature (F') was determined, - at least one real-time feature (F) is determined by the preprocessing module (PM) from a stream of current sensor data (D1 to D5) being acquired from a currently produced batch of paperboard and retrieved from the data repository (R) and - an estimate for at least one quality parameter is determined from the at least one real-time feature (F) by the trained machine-learning module (ML) and provided as output value, wherein at least one range for a quality parameter is provided as additional input to the data-driven module (DD), characterised in that for each range a probability value for the respective quality parameter to fall within the respective range is determined as an output value, wherein for one or more of the probability values determined by the data-driven module (DD) for a range of the quality parameter, each probability value is compared with a probability limit assigned to the respective range and the respective quality parameter, resulting in one Boolean comparison value per range and quality parameter, and further characterised in that a predetermined Boolean expression depending on the at least one Boolean comparison value is evaluated, wherein an alarm is triggered, when the predetermined Boolean expression returns logical true.
  2. Computer-implemented method according to claim 1, characterised in that historical sensor data (D1' to D5') is split into a plurality of intervals, preferably 10000 intervals or more, wherein episodes of historical sensor data (D1' to D5') that are associated with error messages or inconsistencies are omitted.
  3. Computer-implemented method according to claim 1 or 2, characterised in that a delay group (T1 to T5) is determined for a processing step (P1 to P5) as the minimum latency of a parameter change in that processing step (P1 to P5) to affect the outcome of the production subprocess (PP), wherein current sensor data (D1 to D5) acquired within the latency of the respective delay group (T1 to T5) backwards from the current production time stamp (t0) is excluded from evaluation by the data-driven module (DD).
  4. Computer-implemented method according to one of the previous claims, characterised in that at least one feature (F, F') comprises at least one time series describing the variation of at least one parameter of the sensor data (D1 to D5, D1' to D5') along discrete time windows (Δ T i ), wherein the machine-learning module (ML) performs a one-dimensional (1D) convolution of the at least one time series along the time axis (T).
  5. Computer-implemented method according to claim 4, characterised in that for a discrete time window (Δ T i ) at least one statistical parameter, preferably a mean value and/or a standard deviation value, is determined as sample value of the at least one time series.
  6. Computer-implemented method according to one of the previous claims, characterised in that validity and/or consistency of current sensor data (D1 to D5) is evaluated and optionally logged in the data repository (R) by the preprocessing module (PM).
  7. Computer-implemented method according to one of the previous claims, characterised in that one quality parameter is a Z-strength of the paperboard.
  8. Computer-implemented method according to one of the previous claims, characterised in that one quality parameter is a Scott bond of the paperboard.
  9. Computer-implemented method according to claim 1, characterised in that for at least one quality parameter a first range is a set of invalid values lower than a lower valid limit, a second range is a set of valid values between the lower valid limit and an upper valid limit and a third range is a set of invalid values higher than the upper valid limit, wherein an alarm is triggered, when the probability value associated with the first range exceeds a predetermined first probability limit or when the probability value associated with the second range falls below a predetermined second probability limit or when the probability value associated with the third range exceeds a predetermined third probability limit.
  10. System comprising at least one sensor, a data repository (R) and a computing means, - wherein the at least one sensor is designed to acquire sensor data (D1 to D5, D1' to D5') in a processing step (P1 to P5) of a processing pipeline (P) comprising a paperboard production subprocess (PP), - wherein the data repository (R) is designed to receive and persistently store sensor data (D1 to D5, D1' to D5') from the at least one sensor and to transfer sensor data (D1 to D5, D1' to D5') towards the computing means and - wherein the computing means comprises a computer-implemented method according to one of the previous claims.
  11. System according claim 10, characterised in that the data repository (R) is formed as a cloud storage that can be accessed by an internet protocol (IP) based communication protocol stack.
  12. System according to claim 10 or 11 further comprising a signalling means, characterised in that the signalling means is designed to be triggered by the computing means and to indicate the need of manual interaction with the paperboard production subprocess (PP) when being triggered.
  13. Paperboard production machine designed to produce paperboard or a semi-product of paperboard comprising a system according to one of claims 10 to 12.

Description

The invention is related to a method and to a system for estimating paperboard quality parameters. The invention is also related to a paperboard production machine implementing such method for estimating paperboard quality parameters. Paperboard is continuously produced in a processing pipeline comprising a plurality of processing steps, for example mixing incoming pulp, refining pulp, chemically treating pulp and processing pulp by a paperboard machine. The produced paperboard is winded on large reels, wherein one reel may contain paperboard of about 40 kilometres length and about seven metres width. The processing steps may be controlled by settings and/or described by parameters which affect the quality of the produced paperboard. Furthermore, environmental parameters such as temperature or humidity may affect the quality. Changes of processing and environmental parameters may affect the quality in a delayed and non-linear way. Various quality parameters that describe the quality of paperboard are known from the state of the art. As an example, a quality parameter denoted as Z-strength is known that describes the strength of a paperboard along its surface normal. As a further example, a quality parameter denoted as Scott bond is known that describes the in-plane strength of a paperboard. Valid ranges may be specified for one or more of such quality parameters. Paperboard with quality parameters outside this specification is either to be downgraded or to be dumped. In order to determine these quality parameters, according to the state of the art samples are taken manually from a paperboard reel across its width. Such samples are prepared and analysed manually or semi-manually in a laboratory. Due to the high speed of the paperboard machine, samples cannot be taken while paperboard is winded on the reel. Thus, samples can usually be taken about every 45 minutes. The manual or semi-manual analysis takes about further 20 minutes. Therefore, the status and/or the trend of quality parameters are unknown for a longer period, which poses the risk of a substantial scrap rate in paperboard production. Skoglund et al. in 'Comparison between linear and nonlinear prediction models for monitoring of a paperboard machine', Chemical engineering and technology, vol. 25, no. 2, XP055664634 describes determining a grade change based on a thickness of a produced paperboard using neural networks. There is thus a need for an improved estimation of paperboard quality. It is an object of the present invention to provide an improved method for estimating at least one quality parameter of paperboard produced in a paperboard production subprocess of a processing pipeline. This object is achieved by a computer-implemented method according to claim 1. It is a further object of the present invention to provide an improved system for estimating at least one quality parameter of paperboard produced in a paperboard production subprocess of a processing pipeline. This object is achieved by a system according to claim 10. It is yet a further object of the present invention to provide a paperboard production machine implementing an improved method for estimating at least one quality parameter of the paperboard produced. This object is achieved by a paperboard production machine according to claim 13 Exemplary embodiments of the invention are given in the dependent claims. According to a first aspect of the invention, a computer-implemented method is designed for estimating at least one quality parameter of paperboard that is produced in a paperboard production subprocess which is a subprocess of a paperboard processing pipeline. The paperboard processing pipeline is implemented by a data-driven module that comprises a preprocessing module and a machine learning module and comprises at least one processing step implemented by a processing unit or machine. The computer-implemented method comprises the following steps: The computer-implemented method comprises a step, wherein sensor data is acquired from at least one processing step of the processing pipeline and transferred to a data repository. The data repository is designed to store sensor data persistently. Preferably, the data repository is accessible via an internet protocol (IP). Most preferably, the data repository is formed as a cloud storage. The computer-implemented method comprises a further step, wherein at least one historical feature is determined by the preprocessing module by at least partially evaluating historical sensor data. Such historical sensor data was acquired during at least one previously produced batch of paperboard and is retrieved from the data repository. The computer-implemented method comprises a further step, wherein a machine-learning module is trained to reproduce from the at least one historical feature a target value of at least one quality parameter. This target value is determined for a previously produced batch of paperboard corresponding to the hist