In this paper, we propose a method for online assessing the performance of a prognostic approach in situations of very poor knowledge on the degradation process. In particular, we deal with cases in which the entire degradation process, from the beginning of the operation until failure, has never been observed and, thus, the traditional offline performance metrics cannot be applied. The proposed method is applied on a prognostic approach based on a particle filter and optimized tuning kernel smoothing (PF-OTKS). Case studies regarding the degradation of turbine blade and aluminum electrolytic capacitor are considered.

Railway wheel wear prediction is essential for reliability and optimal maintenance strategies of railway systems. Indeed, an accurate wear prediction can have both economic and safety implications. In this paper we propose a novel methodology, based on Archard's equation and a local contact model, to forecast the volume of material worn and the corresponding wheel remaining useful life (RUL). A universal kriging estimate of the wear coefficient is embedded in our method. Exploiting the dependence of wear coefficient measurements with similar contact pressure and sliding speed, we construct a continuous wear coefficient map that proves to be more informative than the ones currently available in the literature. Moreover, this approach leads to an uncertainty analysis on the wear coefficient. As a consequence, we are able to construct wear prediction intervals that provide reasonable guidelines in practice

In recent years, transfer learning has been widely applied in fault diagnosis for solving the problem of inconsistent distribution of the original training dataset and the online-collecting testing dataset. In particular, the domain adaptation method can solve the problem of the unlabeled testing dataset in transfer learning. Moreover, Convolutional Neural Network (CNN) is the most widely used network among existing domain adaptation approaches due to its powerful feature extraction capability. However, network designing is too empirical, and there is no network designing principle from the frequency domain. In this paper, we propose a unified convolutional neural network architecture from a frequency domain perspective for a domain adaptation named Frequency-domain Fusing Convolutional Neural Network (FFCNN). The method of FFCNN contains two parts, frequency-domain fusing layer and feature extractor. The frequency-domain fusing layer uses convolution operations to filter signals at different frequency bands and combines them into new input signals. These signals are input to the feature extractor to extract features and make domain adaptation. We apply FFCNN for three domain adaptation methods, and the diagnosis accuracy is improved compared to the typical CNN.