Output details

A fault detection tool using using analysis from an autoregressive model pole trajectory

Vacuum pumps are important for the production of integrated circuits (ICs) in ‘clean rooms’. A pump failure can ruin a batch of ICs worth £100000+. Most vacuum pumps are based on rotating machines and their health can be monitored by analysing vibration data collected using accelerometers. We created a novel technique (different from vibration signature analysis) based on autoregressive spectral estimation and pole-zero system analysis in which the power of the poles is perfect classifier / fault indicator. The analysis has been validated with data obtained from an actual vacuum pump. This was EPSRC-funded with 3 PhDs and 1 post-doc.