Remaining Useful Life estimation for noisy degradation trends

In process safety and supervision, many works are based on the data driven approaches. Among them, one widely used approach is estimating the Remaining Useful Life (RUL) based on the behavior of degradation trends. In practice, these trends are very noisy because of the measurement process and environments. This paper proposes a method to extract profiles of trends based on a percentile calculation on several levels. This allows one to have a probability density function (pdf) of RUL with a Confidence Interval (CI) that ensures the safety margins for industrial applications. The proposed method is illustrated using a simulation example which highlights its effectiveness, comparing to the filtering methods based on discrete wavelet transform (DWT) and empirical mode decomposition (EMD) algorithms.