Recurrence Based Performance Prediction and Prognostics inComplex Manufacturing Systems

摘要

This paper presents a local recurrence modeling approach for state and performance predictions in complexnonlinear and nonstationary systems. Nonstationarity is treated as the switching force between different stationarysystems, which is shown as a series of finite time detours of system dynamics from the vicinity of an attractor of anonlinear process. Recurrence characteristics of the attractor are used to partition the system trajectory into multiplenear-stationary segments. Consequently, piecewise eigen analysis of ensembles in each near-stationary segment cancapture both nonlinear stochastic dynamics and nonstationarities. The simulation experiment study revealssignificant prediction accuracy improvements over other alternative methods.