An approach for assessment of level of degradation and RUL estimation for a dynamic systeme

In this paper, we propose an approach for failure prognosis by assessing the level of degradation and estimating the residual life before failure of a dynamic system. During its operation, an industrial process as a dynamic system evolves physically. This evolution can have an impact on the modes of operation: nominal, degraded or failure. The proposed approach is based on knowledge of the behavior model of the system. It deals with systems where available measurements are not able to provide information on the level of degradation reached by the system. Since the available data are not usable, and in order to generate indicator signals of degradation, the method based on the generation of residues by an observer was selected. Due to the model uncertainties and disturbances, the evolution of the characteristics of residues is not necessarily correlated with the evolution of the level of degradation. Therefore, the residues generated were exploited to model the degradation behavior. We selected a Hidden Markov Model or its multi-stream extension in the case of a stream of residuals. The proposed approach was conducted in two phases: a learning phase to model the degradation behavior and a prognosis phase to assess the level of degradation and estimate the Remaining Useful Life. To illustrate the approach, a simulated induction motor was used.

[1]  Van Nostrand,et al.  Error Bounds for Convolutional Codes and an Asymptotically Optimum Decoding Algorithm , 1967 .

[2]  L. Baum,et al.  An inequality and associated maximization technique in statistical estimation of probabilistic functions of a Markov process , 1972 .

[3]  Slim Tnani,et al.  Diagnosis by parameter estimation of stator and rotor faults occurring in induction machines , 2006, IEEE Transactions on Industrial Electronics.

[4]  Robert M. Gray,et al.  An Algorithm for Vector Quantizer Design , 1980, IEEE Trans. Commun..

[5]  G. G. Stokes "J." , 1890, The New Yale Book of Quotations.

[6]  P. Frank,et al.  Survey of robust residual generation and evaluation methods in observer-based fault detection systems , 1997 .

[7]  Ren-Hua Wang,et al.  Integrating Articulatory Features Into HMM-Based Parametric Speech Synthesis , 2009, IEEE Transactions on Audio, Speech, and Language Processing.

[8]  Toufik Aggab Pronostic des systèmes complexes par l’utilisation conjointe de modèle de Markov caché et d’observateur , 2016 .

[9]  G. McLachlan,et al.  Assessing the Number of Components in Mixture Models , 2005 .

[10]  Thomas Welte,et al.  Deterioration and Maintenance Models for Components in Hydropower Plants , 2008 .

[11]  Gilles Celeux,et al.  Selecting hidden Markov model state number with cross-validated likelihood , 2008, Comput. Stat..

[12]  Noureddine Zerhouni,et al.  Remaining Useful Life Estimation of Critical Components With Application to Bearings , 2012, IEEE Transactions on Reliability.

[13]  Charles M. Grinstead,et al.  Introduction to probability , 1999, Statistics for the Behavioural Sciences.

[14]  Bimal K. Bose,et al.  Modern Power Electronics and AC Drives , 2001 .

[15]  Frédéric Kratz,et al.  Failure Event Prediction Using Hidden Markov Model Approaches , 2015, IEEE Transactions on Reliability.

[16]  S. Kotsiantis,et al.  Discretization Techniques: A recent survey , 2006 .