The sixth UQSay seminar on Uncertainty Quantification and related topics, organized by L2S and MSSMAT, will take place on Thursday afternoon, November 28, 2019, at CentraleSupelec Paris-Saclay (Eiffel building, amphi V).
We will have two talks:
Remaining useful life prediction with imprecise and partial
information:
An evidential hidden Markov model-based approach
Industrial systems often degrade before failure occurs. True
degradation states, however, are often hidden and cannot be
observed directly. Condition-monitoring data, e.g., noise,
vibration, on the other hand, are available from sensors and
contain information on the degradation process. Apart from
condition-monitoring data, sometimes inspections can be made to
directly observe the system degradation state. The inspection
data, however, are often imprecise and contain only partial
information on the true degradation states.
In this talk, we present an evidential hidden Markov model-based
approach to integrate condition-monitoring data with inspection
data for degradation state estimation and remaining useful life
prediction. The degradation process is modeled by a discrete state
continuous time Markov chain. The condition-monitoring data are
modeled by a Gaussian mixture model given the true degradation
state. Inspection data are modeled using evidence
theory. Condition-monitoring data and inspection data are
integrated in the framework of evidence theory and the parameters
in the degradation model are estimated through expectation
maximization algorithm. Remaining useful life of the system is,
then, predicted based on the estimated parameters. The developed
methods are tested through some numerical experiments and applied
on a real case study of bearing degradation data from literature.
Joint work with Tangfan Xiahou and Yu Liu.
Surrogate modeling based on resampled polynomial chaos expansions
In surrogate modeling, polynomial chaos expansion (PCE) is popularly utilized to represent the random model responses. Recently, efforts have been made on improving the prediction performance of the PCE-based model and building efficiency by only selecting the influential basis polynomials (e.g., via the approach of least angle regression). An approach, named as resampled PCE (rPCE), is proposed to further optimize the selection by making use of the knowledge that the true model is fixed despite the statistical uncertainty inherent to sampling in the training. By simulating data variation via resampling (k-fold division utilized here) and collecting the selected polynomials with respect to all resamples, polynomials are ranked mainly according to the selection frequency. The resampling scheme (the value of k here) matters much and various configurations are considered and compared. The proposed resampled PCE is implemented with two popular selection techniques, namely least angle regression and orthogonal matching pursuit, and a combination thereof. The performance of the proposed algorithm is demonstrated on two analytical examples, a benchmark problem in structural mechanics, as well as a realistic case study in computational dosimetry.
Joint work with Dominique Lesselier, Bruno Sudret and Joe Wiart.
Reference: hal-01889651.
Organizers: Julien Bect (L2S), Fernando Lopez Caballero (MSSMAT) and Emmanuel Vazquez (L2S).
No registration is needed, but a simple email would be appreciated if you intend to come.
