Data-driven prognosis applied to complex vacuum pumping systems

This paper presents a method to address system prognosis. It also details a successful application to complex vacuum pumping systems. The proposed approach relies on an automated data-driven learning process as opposed to hand-built models that are based on human expertise. More precisely, using historical vibratory data, we first model the behavior of a system by extracting a given type of episode rules, namely First Local Maximum episode rules (FLM-rules). A subset of the extracted FLM-rules is then selected in order to further predict pumping system failures in a datastream context. The results that we got for production data are very encouraging as we predict failures with a good time scale precision. We are now deploying our solution for a customer of the semi-conductor market.