Stochastic process-based degradation modeling and RUL prediction: from Brownian motion to fractional Brownian motion

Abstract

Brownian motion (BM) has been widely used for degradation modeling and remaining useful life (RUL) prediction, but it is essentially Markovian. This implies that the future state in a BM-based degradation process relies only on its current state, independent of the past states. However, some practical industrial devices such as Li-ion batteries, ball bearings, turbofans, and blast furnace walls show degradations with long-range dependence (LRD), where the future degradation states depend on both the current and past degradation states. This type of degradation naturally brings two interesting problems, that is, how to model the degradations and how to predict their RULs. Recently, in contrast to the work that uses only BM, fractional Brownian motion (FBM) is introduced to model practical degradations. The most important feature of the FBM-based degradation models is the ability to characterize the non-Markovian degradations with LRD. Although FBM is an extension of BM, it is neither a Markovian process nor a semimartingale. Therefore, how to obtain the first passage time of an FBM-based degradation process has become a challenging task. In this paper, a review of the transition of RUL prediction from BM to FBM is provided. The peculiarities of FBM when addressing the LRD inherent in some practical degradations are discussed. We first review the BM-based degradation models of the past few decades and then give details regarding the evolution of FBM-based research. Interestingly, the existing BM-based models scarcely consider the effect of LRD on the prediction of RULs. Two practical cases illustrate that the newly developed FBMbased models are more generalized and suitable for predicting RULs than the BM-based models, especially for degradations with LRD. Along with the direction of FBM-based RUL prediction, we also introduce some important and interesting problems that require further study.

Scalability, Explainability and Performance of Data-Driven Algorithms in Predicting the Remaining Useful Life: A Comprehensive Review

“…As mentioned before, the Wiener process, also referred to as the Brownian motion process, is one of the most important and well-studied stochastic processes used for modeling the degradation processes. A Wiener process's ability to replicate the random behavior of systems makes it suitable for modeling degradation, thus serving as the diffusion term of a degradation model [176]. The stochastic degradation in timedependent systems is due to four main components, including variations in time, variations from one system to the next, uncertainty in measurements, and the non-linear variations in the system [116,138,176,214,248,249].…”

Multi-Fractal Weibull Adaptive Model for the Remaining Useful Life Prediction of Electric Vehicle Lithium Batteries

“…The Brownian motion is suitable to describe the non-monotonous degradation process [ 24 ]. However, the Brownian motion is Markovian, whereas the degradation data often show non-Markovian characteristics, e.g., the capacity data of the lithium battery possesses long-range dependence [ 25 ]. Long-range dependence means that the present value is influenced by the previous values of the time series.…”

Reliability evaluation based on stochastic degradation models with a new model selection criterion

“…Due to the inevitable uncertainties within the degradation process, the deterioration of the product performance characteristics can be regarded as the gradual accumulation of minor damages, which can be modelled by random degradation increments in stochastic degradation models. This property allows stochastic degradation models to capture the inherent temporal variability of the degradation process over time, and makes them a popular topic of reliability evaluation for complex systems [6] . Among several candidate stochastic degradation models, the Wiener process is a basic and widely used model with a simple structure and close analytical solution, especially for non-monotone degradation processes [7] .…”