One Realization of an Industrial Device for Machine State Estimation

Performing maintenance checks and state detection of machines with rotary actuators in industry is a challenging task which requires expensive and complicated equipment. One of the ways to perform this is using a system based on recording and analysis of sounds the machine makes during operation. This paper describes an approach which uses a simple, compact device based on a commercial microcomputer that aims to entirely replace the state detection system. It is able to record acoustic signals in close proximity to the machine, process them and use a neural network to output the state of the rotary actuator. The testing has been done on real acoustic signals recorded in the thermal power plant Kostolac in Serbia.