dc.description.abstract | In this thesis the Czochralski process is introduced as well as the technological setup
used in the process. The three problems this thesis aim to solve are presented: implementation of automatic cold ingot detection, automatic detection of structure loss in the body phase and automatic temperature detection in the neck stabilization phase. After the
introduction, the relevant theory is presented. This thesis references the automatic cold ingot detection algorithm developed in a previous project heavily, so this work is presented
in a separate chapter. Then each of the three main problems are described and solutions
are proposed, tested and discussed for each of them separately. At the end of this thesis
is a chapter with conclusions of the proposed problem solutions. The thesis resulted in
problems being uncovered in the implemented automatic cold ingot detection which needs
to be further analyzed and tuned for a satisfactory performance in the factory. For the
structure loss detection, several approaches were proposed. Most of the approaches can
be discarded, but the mean value approach and the machine learning approach both seem
viable for a robust detection of the structure loss phenomenon. In the case of automatic
temperature detection in the neck stabilization phase, there were problems with the camera
not having a high enough resolution to get precise results. If a new camera is acquired this
thesis proposes two separate methods which can be viable for automatically detecting the
desired temperature in this phase, a machine learning approach and a dynamic programming
approach, where the problem is divided into smaller parts. | en |