Kurzschlusserkennung in der Kupferelektrolyse durch neuronale Netzwerke

Translated title of the contribution: Short-circuit detection in the copper tankhouse through neural networks

Matthias Lindthaler

Research output: ThesisMaster's Thesis

549 Downloads (Pure)

Abstract

This thesis deals with the advantages and disadvantages of digitisation and in particular with the problems of introducing these concepts in the non-ferrous industry. In order to show different optimization potentials, some existing concepts of other industries are presented. Based on some examples, the introduction of a digital twin, the use of soft sensors, the endpoint determination in smelting furnaces, the prediction of steel properties and the prediction of the current yield in copper refining electrolysis, which is used for the production of high-purity copper, are discussed. The optimisation of the current yield is discussed in more detail here. The most important parameter for determining efficiency here is the current yield, which, however, is considerably negatively influenced by the formation of short circuits. For this reason, existing possibilities to determine this parameter are presented here and a new concept is introduced which is based on the evaluation of thermal images by means of neural networks. For this purpose the development of this and the theoretical basics for classification and object recognition in images were presented. The difficulty of the so far used evaluations of various IR images in electrolysis was the very narrow distribution of the electrodes and was also strong dependent on the environmental conditions. This could be largely improved here. With the help of automated detection, short circuits can be detected on average twelve hours in advance, which increases the current yield of the electrolysis.
Translated title of the contributionShort-circuit detection in the copper tankhouse through neural networks
Original languageGerman
QualificationDipl.-Ing.
Awarding Institution
  • Montanuniversität
Supervisors/Advisors
  • Andreas Filzwieser, Dr., Supervisor (external), External person
  • Antrekowitsch, Helmut, Supervisor (internal)
Award date1 Jul 2020
Publication statusPublished - 2020

Bibliographical note

embargoed until null

Keywords

  • neural networks
  • digitization
  • industry 4.0
  • short circuit detection
  • thermal imaging
  • copper electrolysis
  • increasing current efficiency

Cite this