Standard bearing fault detection features are shown to be ineffective for estimating bearings remaining useful life (RUL). Addressing this issue, in this paper we propose an approach for bearing fault prognostics employing Rényi entropy based features describing the statistical properties of the envelope of the generated vibrations and a set of Gaussian process (GP) models. These models are non-parametric black-box models which search for the relationships among measured data rather than trying to approximate the modelled system by fitting the parameters of the selected basis functions. Bearing RUL is estimated as a posterior distribution following the Bayes' rule using GP models' output as likelihood distribution. The proposed approach was evaluated on the data set provided for the IEEE PHM 2012 Prognostic Data Challenge.
COBISS.SI-ID: 27855399
This paper presents an alternative computational method for on-line estimation and tracking of the impedance of PEM fuel cell systems. The method is developed in order to provide the information to diagnostics and health management system. Proper water management remains the main issue influencing the reliability and durability of PEM fuel cell technology and despite the thorough understanding of the underlying processes and extensive experimental work, the hardware implementation of the methods is often not properly addressed. In this scope, we will show how the characteristic values of the fuel cell impedance, required by the diagnostic system, at different frequencies can be computed by robust and computationally efficient algorithms, which are suitable for implementation in embedded systems. The methods under consideration include continuous-time wavelet transform and extended Kalman filter. The results demonstrate that these methods
COBISS.SI-ID: 27802151