Midterm colloquium Casper van Peijpe

Supervisor: Dr. Peyman Mohajerin Esfahani

Moderator: Amin Sheikhi

Canon Production Printing develops, manufactures and sells high-end industrial printers. The company's research and development department is dedicated to improving the quality of printing. As part of the R&D program, collaborations with academic partners are cultivated under specific projects, such as the Digital Twin Project 4. This project aims to develop methods for proactive and automated health management of manufacturing processes and high-tech machines, such as fault detection and isolation for printers, which is researched in this MSc thesis project. The fluid dynamics in the ink channel of a printer can be captured in a model. Using the output signals of a printer, correlated to the volume flow of the ink, the printer model can be identified. The model is used to construct a fault detection filter. This filter, whose input is the known output signal of the printer, generates a residual signal, which differentiates whether the system is healthy or influenced by a fault, by looking at whether the energy of this residual signal exceeds a certain threshold. A fault isolation filter is designed using the residuals of known faulty data. Linear regression determines what fault is occurring using the residual of the signal that must be isolated by generating a probability vector, indicating the chance of each fault class being present. Results show that both filters work sufficiently for both simulated and real data.