An intelligent system for dynamic system state forecasting

Abstract

In this paper, an adaptive neuro-fuzzy (NF) forecasting system is proposed, and its robustness is investigated experimentally. After the NF predictor is initially trained using a data set from the Mackey-Glass differential equation, it is implemented for two applications, an online gear system condition monitoring and a material fatigue testing to forecast future states of a fatigue crack propagation trend in test specimens. From the forecasting tests and simulation analyses, it is found that the developed NF system is a very reliable prognostic scheme; it can capture system dynamic behavior quickly, and track system responses accurately. © Springer-Verlag Berlin Heidelberg 2005.