Validation of Data-Driven Labeling Approaches Using a Novel Deep Network Structure for Remaining Useful Life Predictions

Abstract

Today, most research studies that aim to predict the remaining useful life (RUL) of industrial components based on deep learning techniques are using piecewise linear (PwL) run-to-failure targets to model the degradation process. However, this PwL degradation model assumes a constant initial RUL value in which only time is needed to model normal operating conditions. Thus, it ignores the entire diagnostics aspect. To provide high and reliable RUL prediction accuracy, a prognostics algorithm must incorporate diagnostics information. This paper will provide the Prognostics and Health Management Community an empirical study that validates the PwL degradation model against other, more recent data-driven labeling approaches. We compare three different data-driven labeling approaches for RUL predictions. First, an unsupervised reconstruction-based fault detection algorithm is used to provide valuable diagnostics information. Then, optimized initial RUL values are calculated based on this information. Finally, these values are used to construct PwL, descriptive statistics, and anomaly score function run-to-failure targets for subset FD001 in the popular and publicly available C-MAPSS data set. A deep network structure is proposed and trained on the three different run-to-failure targets in order to predict the RUL. During the training process, a genetic algorithm approach is used to tune a selected search space of hyper-parameters. The results suggest that the network trained on PwL run-to-failure targets with the optimized initial RUL values performs the best and provides the most reliable RUL prediction accuracy. This network also outperforms the most robust results in the literature. INDEX TERMS Data-driven labeling approaches, deep learning, fault detection, prognostics and health management, remaining useful life.

A Review on Deep Learning Applications in Prognostics and Health Management

“…The time point segmenting the piece-wise function can be set according to prior knowledge, as in [175]- [179]. It can also be determined via a fault detection procedure, using, for example, statistical process control [180], SVM [181], variational AE [182] or a singular value decomposition (SVD) normalized correlation coefficient [183]. As an alternative to the linearly decreasing function, researchers investigated power functions [181] and low-order polynomials [182] with the hope of better capturing the degradation pattern.…”

“…Late prediction may lead to unplanned breakdown, or even catastrophic damage, whereas early prediction only causes extra maintenance cost. To cope with this problem, the following asymmetric scoring function for evaluating model performance was proposed by [189], adopted by [182], [190], [191] and modified by [192]- [194]:…”

“…In a method resembling the principle of random forest, Zhang et al constructed a multiple DBN ensemble to maximize two conflicting objectives: accuracy and diversity. Composite models using LSTM with RBM [190] and 1D CNN [176], [182] were also investigated recently, and quite competitive performance was reported.…”

State-of-Health Assessment for Aero-Engine Based on Density-Distance Clustering and Fuzzy Bayesian Risk

“…Regarding the SOH assessment of aero-engine, some scholars consider that SOH assessment is an auxiliary link to help other functions of PHM [19]- [21]. Among them, the combination of health assessment and RUL for an engine is the most common research issue, and researchers prefer to use the publicly available Commercial Modular Aero-Propulsion System Simulation (C-MAPSS) dataset as a typical representative to carry out relevant study [22]- [24]. In the C-MAPSS dataset, each engine starts with different degrees of initial wear and manufacturing variation and begins to degrade at some time, all these details are unknown to the public.…”

Remaining Useful Life Prediction of Rolling Bearings Based on RMS-MAVE and Dynamic Exponential Regression Model

“…Data-driven approaches build the degradation model from the historical data using machine learning [12] or statistical approaches [13]. Gebraeel et al [14] applied an ANN-based model to predict the RUL of bearings.…”