Detection of failures in experimental workbench of rotary machines using intelligent systems

Abstract

The research project consists of the development of an intelligent system for the analysis of the structural integrity of an unbalanced rotor. The data are obtained through simulations performed on an experimental bench. To perform the simulations of the normal (base-line) and abnormal (structural failure) conditions of the unbalanced rotor, the mass of the structure is added, causing the behavior of the structure to be dynamically altered, characterizing an abnormal situation. For the detection of system failure, an intelligent system will be used, namely the ARTMAP Fuzzy neural network, which is an architecture based on adaptive resonance theory. This neural network is stable and robust and allows the inclusion of the continuous training module, that is, an internal mechanism that allows the acquisition of knowledge without the need to restart the training process with the inclusion of new standards, so the proposed system is will automatically improve, thereby ensuring an advantage over other artificial neural network topologies.