Abstract-On-board calibration of bidimensional aperture synthesis radiometers with a large number of antennas by the standard correlated noise injection method is technologically very critical because of the stringent requirements on mass, volume, and phase equalization of the noise distribution network. A novel approach, which makes use of a set of uncorrelated noise sources uniformly distributed in the array, is proposed in this paper. Each noise source drives correlated noise only to a small set of adjacent antennas. These sets of antennas are overlapped in order to maintain phase and modulus track along the array. This approach reduces drastically mass and volume of the noise distribution network. Moreover, its phase matching requirement is strongly relaxed because it is only necessary within small sets of adjacent antennas. Power stability of the uncorrelated noise sources is also not a stringent requirement. This procedure allows independent phase and modulus calibration by making use of a reduced number of redundant correlations.

The traditional deep learning-based bearing fault diagnosis approaches assume that the training and test data follow the same distribution. This assumption, however, is not always true for the bearing data collected in practical scenarios, leading to significant decline to fault diagnosis performance. In order to satisfy this assumption, the transfer learning concept is introduced in deep learning by transferring the knowledge learned from other data or models. Due to the excellent capability of feature learning and domain transfer, deep transfer learning methods have gained widespread attention in bearing fault diagnosis in recent years. This article presents a comprehensive review of the development of deep transfer learning-based bearing fault diagnosis approaches since 2016. In this review, a novel taxonomy of deep transfer learning-based bearing fault diagnosis methods is proposed from the perspective of target domain data properties divided by labels, machines, and faults. By covering the whole life cycle of deep transfer learning-based fault diagnosis and discussing the research challenges and opportunities, this review provides a systematic guideline for researchers and practitioners to efficiently identify suitable deep transfer learning models based on the actual problems encountered in bearing fault diagnosis.

No abstract

The Kalman filter is one of the most important and common optimal recursive data processing algorithm in many applications characterized by linear dynamical behavior and affected by random zero-mean white Gaussian noise. However, when measurement processes are considered, inaccuracy is not only due to noise, but also to several contributions to uncertainty that can be due to both random and uncompensated systematic effects. Therefore, when the Kalman filter is used on experimental data, all uncertainty contributions should be considered.While several proposals are available in the literature to modify the Kalman filter in order to consider different probability distributions and systematic effects, represented both in probability and possibility domains, they are not fully compliant with the uncertainty concept adopted in metrology. The aim of this paper is hence to reformulate the Kalman filter theory within the possibility domain in compliance with the measurement uncertainty concept, in order to be able to consider both the random and systematic contributions to uncertainty (regardless of their distribution) that may affect the measurement process.An experimental set-up is considered and the results obtained under different assumptions are reported.

This paper discusses the possibility to use the current ripple introduced by switch mode power converters for low-cost monitoring of polymer electrolyte membrane (PEM) fuel cell (FC) state-of-health, suitable for commercial applications that cannot afford dedicated instrumentation. In more details, an estimate of the ohmic resistance, which is a good indicator of the membrane water content, can be obtained from the high-frequency ripple response by data processing in the frequency domain, while lower frequency ripple at 100/120 Hz (when present) is in the typical frequency range of activation processes. All the available impedance estimates, together with the dc voltage measurement, can be used to promptly detect FC drying and flooding, that are the two opposite failure modes as far as water balance is concerned. The proposed diagnostic approach is tested on a single PEM FC in drying and flooding conditions, by emulating three-phase and single-phase inverter ripples by means of an electronic load