Knowledge elicitation by merging heterogeneous data sources in a die-casting process

Abstract

In order to establish adaptive control of a manufacturing process knowledge must be acquired about both, the process and its environment. This knowledge can be obtained by mining large amounts of data collected through the monitoring of the manufacturing process. This enables the study of process parameters and the correlations between the process parameters and with the parameters of the environment. Through this, knowledge about the process and its relation to the environment can be established and, in turn, used for adaptive process control. The aim of this thesis is to study real manufacturing data, obtained through monitoring of a die casting process. First, in order to better understand the problem at hand, a literature review of using Big data and merging data from heterogeneous sources is given. Second, using the real data, a robust algorithm to asses the quality rate was developed, due to data being incomplete and noisy. Merging the process and the environment data was done. In this way it is possible to visualize the influences of various parameters on quality rate and make suggestions for improvement of the die casting process.