Long-term prediction of bearing condition by the neo-fuzzy neuron

Rolling element bearings are devices used in almost every electrical machine. Therefore, it is important to monitor and track the degradation of bearings. This paper presents a new approach to predict the degradation of bearings by a time series forecasting model called the neo-fuzzy neuron. The proposed approach uses the root mean square extracted from vibration signals as a health indicator. The root mean square is used here as an input of the neo-fuzzy neuron in order to estimate the evolution of bearing's degradation in time. Experimental degradation data provided by the University of Cincinnati is used to validate the proposed approach. A comparative study between the neo-fuzzy neuron and the adaptive neuro-fuzzy inference system is carried out to appraise their prediction capabilities. The experimental results show that the neo-fuzzy model can track the degradation of bearings.