Predictive maintenance is emerging as a promising technique to overcome the limitations of periodic maintenance. It integrates automatic condition monitoring to evaluate the health status of a system or device, with the estimation of Remaining Useful Life (RUL) of its components, in order to schedule maintenance only when really needed, minimizing the downtime of a plant. In this paper, we report our results in evaluating the health conditions of the rotating ball bearings of a critical air pump which is part of the Clean Room industrial facility operating in the Micro-Technologies Laboratory of Fondazione Bruno Kessler. We instrumented such component with vibration and acoustic sensors with the aim of identifying a model of the evolving degradation and estimating the RUL of the bearings. The current dataset covers a period of one month before bearing replacement and about six months after the replacement. The first models, based on regression and particle filter processing of critical spectral components extracted from the sensors, indicate an estimated RUL of about 12 months that is in agreement with the average lifetime based on scheduled maintenance. Subsequent model evolutions have been observed in conjunction with scheduled greasing and periods of lower stress, which resulted in remarkable deviations from the initial degradation trends and in a consequent notable increase of the estimated RUL. Results achieved so far are promising and could be used to extend the temporal distance among periodic maintenance interventions according to the estimated RUL.

This paper illustrates a data-driven approach adopted to address the PHME2020 Data Challenge competition. The aim of the challenge was to estimate the Remaining Useful Lifetime (RUL) of an experimental filtration device analyzing its clogging status by means of static (e.g. data sheets, fluid type, sensors) and dynamic (e.g. sensing data) information. We address the problem employing different state-of-the-art feature extraction, feature selection, and machine learning techniques. In this paper, we describe the approach followed to assess the problem and to generate robust and adaptable prediction models together with a corresponding performance assessment and robustness evaluation. The performance of the proposed solution is calculated in term of penalty score. The final penalty score 57.24 ranked 2nd in the above-mentioned data challenge competition.