State Detection of Rotary Actuators Using Wavelet Transform and Neural Networks

Abstract

Rotary actuators are among the most commonly used machines in the industry and the algorithm for detecting the level of wear they are subjected to can prevent significant amount of unnecessary maintenance expenses. This paper proposes a new algorithm which can detect the state of the rotating machine using acoustic signals recorded in its vicinity. The algorithm uses a combination of wavelet transform and neural networks and is computationally inexpensive, so it can be implemented on a simple microcontroller. The testing has been done on real acoustic signals recorded in thermal power plant Kostolac in Serbia.