Enabling Health Monitoring Approach Based on Vibration Data for Accurate Prognostics

TLDR

Data‑driven prognostics performance depends heavily on the form and trend of extracted features. The paper aims to develop features that clearly reflect machine degradation to achieve accurate prognostics. It extracts and selects monotonic, early‑trend features from vibration data using trigonometric functions, cumulative transformation, and discrete wavelet transform, then trains a summation wavelet‑extreme learning machine model validated on cutting‑tool and bearing datasets. The proposed method outperforms classical approaches in feature fitness, cutting‑tool wear estimation, and bearing long‑term prediction, confirming its effectiveness.

Abstract

The performance of data-driven prognostics approaches is closely dependent on the form and trend of extracted features. Indeed, features that clearly reflect the machine degradation should lead to accurate prognostics, which is the global objective of this paper. This paper contributes a new approach for feature extraction/selection: The extraction is based on trigonometric functions and cumulative transformation, and the selection is performed by evaluating feature fitness using monotonicity and trendability characteristics. The proposition is applied to the time-frequency analysis of nonstationary signals using a discrete wavelet transform. The main idea is to map raw vibration data into monotonic features with early trends, which can be easily predicted. To show that, selected features are used to build a model with a data-driven approach, namely, the summation wavelet-extreme learning machine, that enables good balance between model accuracy and complexity. For validation and generalization purposes, the vibration data from two real applications of prognostics and health management challenges are used: (1) cutting tools from a computer numerical control machine (2010); and (2) bearings from the platform PRONOSTIA (2012). The performance of the proposed approach is thoroughly compared with the classical approach by performing feature fitness analysis, cutting-tool wear "estimation", and bearings' "long-term prediction" tasks, which validates the proposition.