Industrial process monitoring by multi-channel acoustic signal analysis

Abstract

Machinery monitoring at the shop floor bears relevance in preventive maintenance applications and for manufacturing process optimization. As the installation of monitoring hardware directly on the machinery may be hazardous and expensive due to installation costs, the use of contactless sensors is preferable. In this paper we propose a solution for machinery monitoring based on multi-channel acoustic information analysis. We apply large aperture microphone arrays, perform machine noise source localization using the SRP-PHAT method and classify machine acoustical patterns by means of fuzzy rule-based classification. The results of experiments, performed in an industrial setting, indicate the feasibility of our solution in real conditions.