Sample page The facts are real. The words are stand-ins — yours go in their place.
Yiğit Özcan
A question that would not go away, a method that had to be built twice, and the hour in which all of it is defended in public.
“Digital Shadow for Process-Parallel Part Geometry Estimation in CNC Milling”
Three questions
Why this topic?
It began with something that bothered me and would not leave me alone. The field had an answer already, and I did not believe it.
What surprised you?
How often the interesting result was the one I was not looking for, and how long it took me to stop treating it as noise.
What happens after the defense?
Sleep, first. Then finding out whether what I learned survives outside the university.
In their words“At the start I thought I knew the answer. By the third year I knew I had not understood the question.”
About the thesis
In CNC milling, parts are often checked only after machining by using high-precision measurement equipment. This can take a lot of time, especially for complex and accurate components. If an error is found, the part may need to be reworked or even rejected. This PhD research explores how a CNC machine can be made more “aware” of the shape of the part while it is being machined. The thesis develops a Digital Shadow of the machining process. This means that data from the real CNC machine are used to update a virtual representation of the workpiece during machining. The system reads machine positions and other controller data, combines them with models of cutting forces and tool deflection, and updates a virtual workpiece using fast GPU-based material removal simulation. The goal is not to replace certified coordinate measuring machine inspection, but to give operators and engineers earlier insight into possible geometry deviations. By estimating the evolving part geometry during the process, the method can help reduce inspection effort, improve process transparency, and support better decisions in advanced manufacturing. The research shows that this type of in-process geometry estimation is feasible using industrial machine data, analytical models, and high-performance computing.
Selected work
Listed from the KU Leuven agenda.