In highly automated factories, log data from tools and services accumulate with rates of several GB per minute. Deriving insights from these data remains a challenge task. Multiple problems need to be addressed. Do my logs contain unusual messages?
How to enable structured log drilldown in distributed and highly dynamic manufacturing environments? How to avoid information overflow when interfacing with factoring staff through automated reporting?
In this internship or thesis project, you would evaluate and implement methods to structure log data based on log content similarities and structural time series modeling. This will enable the detection of abnormal events such as rare tool states, unusual material routes or stalled processes. Suitable methods shall be condensed and implemented into a process to simplify root cause analysis. In addition, abnormal log events and their factory-wide aggregation shall be used to design a global anomaly score intended to streamline and simplify alerting.
You should have programming experience or interest in Java, SQL, Azure & AI. Interest or knowledge in manufacturing processes would be helpful as well.