Diploma and Master Theses (authored and supervised):

J. Schmidt:
"Interactive Variability Analysis for Initial Sample Testing of Industrial CT Data";
Supervisor: E. Gröller, L. Fritz; Institut für Computergraphik und Algorithmen, 2011.

English abstract:
Casting is a manufacturing process by which melt is poured into a mould and then allowed to solidify. Due to the pouring process as well as during solidification casting defects like shrinkage, pores or cracks may appear that can influence the cast objects´ mechanical properties and usability. Designing a new type of casting always requires the production of a new pattern first. Before release of serial production a series of initial samples is created. During this evaluation step possible problems are identified that may occur for example due to the mould layout. In the course of this master´s thesis effort, a new approach for first article inspection is introduced. The goal of this approach is to explore a series of industrial CT volumes, with the objective to evaluate the casting design. Using the pipeline investigations of the reliability of the production can be performed. The variability analysis is based on a segmentation approach for defect detection which enables to differentiate discontinuities from the rest of the volume In case regions of high defect concentration should be identified, a clustering approach is applied to all samples within the series. The input data set used for clustering consists of the defects´ centres of gravity. This way the clustering algorithm can be applied to a set of data points in 3D space. In the further course the cluster information is used to analyse the variability of the sample series, since they represent regions of high defect concentration. The analysis of large defects, in contrast, directly operates on the data produced by the defect detection stage. The data of all samples is compared to find possible overlaps. For the variability analysis it is important to know whether regions can be found that contain defects in more than one trial sample. The identification of such regions shows that the casting process tends to produce errors in certain parts of the object.

Electronic version of the publication:

Created from the Publication Database of the Vienna University of Technology.