Particle filter based hybrid prognostics for health monitoring of uncertain systems in bond graph framework

• Published in: Mechanical systems and signal processing Vol. 75; pp. 301 - 329
• Main Authors: Jha, Mayank Shekhar, Dauphin-Tanguy, G., Ould-Bouamama, B.
• Format: Journal Article
• Language: English
• Published: Elsevier Ltd 15.06.2016; Elsevier
• Subjects: Automatic; Bond Graph; Bond graphs; Computer Science; Confidence intervals; Construction; Degradation; Engineering Sciences; Intervals; Machine Learning; Mathematical analysis; Mathematical models; Modeling and Simulation; Particle Filter; Prognostics; Remaining Useful Life; Robust Fault Detection; Signal and Image processing; Statistics; Uncertainty; Intervals; ARR; EOL; Particle Filter; BG; DPP; Robust Fault Detection; I-ARR; RUL; DM; RMSE; Prognostics; BG-LFT; RA; Bond Graph; PF; Remaining Useful Life; RMAD; SIR
• ISSN: 0888-3270; 1096-1216
• DOI: 10.1016/j.ymssp.2016.01.010

The paper’s main objective is to address the problem of health monitoring of system parameters in Bond Graph (BG) modeling framework, by exploiting its structural and causal properties. The system in feedback control loop is considered uncertain globally. Parametric uncertainty is modeled in interval form. The system parameter is undergoing degradation (prognostic candidate) and its degradation model is assumed to be known a priori. The detection of degradation commencement is done in a passive manner which involves interval valued robust adaptive thresholds over the nominal part of the uncertain BG-derived interval valued analytical redundancy relations (I-ARRs). The latter forms an efficient diagnostic module. The prognostics problem is cast as joint state-parameter estimation problem, a hybrid prognostic approach, wherein the fault model is constructed by considering the statistical degradation model of the system parameter (prognostic candidate). The observation equation is constructed from nominal part of the I-ARR. Using particle filter (PF) algorithms; the estimation of state of health (state of prognostic candidate) and associated hidden time-varying degradation progression parameters is achieved in probabilistic terms. A simplified variance adaptation scheme is proposed. Associated uncertainties which arise out of noisy measurements, parametric degradation process, environmental conditions etc. are effectively managed by PF. This allows the production of effective predictions of the remaining useful life of the prognostic candidate with suitable confidence bounds. The effectiveness of the novel methodology is demonstrated through simulations and experiments on a mechatronic system. •Novel integration of Bond Graph modeling framework and Particle Filter for estimation of state of health of system parameter and prediction of remaining useful life.•Novel algorithm for interval valued Thresholds for robust detection of degradation beginning.•Novel method proposed to obtain observation equation from nominal residual.•Novel simplified variance adaptation scheme for particle filters proposed, results in efficient estimation and prediction.•Approach demonstration by simulation and real time experiment on mechatronic Torsion bar system.