Prozessanalyse und Prozessoptimierung mit statistischen Verfahren

Abstract

This thesis addresses the question of how data-driven process analysis and optimization can be practically implemented in industrial settings. The increasing use of statistical methods in operational process improvement requires in-depth mathematical and statistical knowledge, which is often not widely available in companies. Against this background, the aim of this work was to develop a methodologically sound and application-oriented model that supports the use of statistical procedures in production environments. The resistance welding process served as an empirical case study for a systematic analysis of its optimization potential. The study is divided into a theoretical and an empirical part. Chapter 3 provides the theoretical foundation for process analysis and optimization, including established approaches such as Lean Management, Six Sigma, and the cause-and-effect principle. Chapter 4 presents selected statistical procedures – including descriptive statistics, regression analysis, analysis of variance, and design of experiments – and explains their application depending on the measurement scales of influencing variables (X) and target variables (Y). Chapter 5 contains the practical case study, which follows a clearly defined research process. It includes problem modeling, selection of relevant influencing factors, hypothesis formulation, and the evaluation of simulated observational data and real experimental data using the statistical software Minitab. The results of the case study show that both welding current and welding time have a significant effect on the torque value as a central output variable. Higher levels of these parameters resulted in stronger welds with lower variability. Clamping force did not show a significant main effect in isolation but revealed relevant interactions when combined with welding time. The best results were achieved with lower clamping force and extended welding time. Based on the statistical design of experiments, the optimal process parameter settings for improving the target variable were successfully identified.

Date of Award	2025
Original language	German (Austria)
Supervisor	Harald Dobernig (Supervisor)