Multi-Resolution Markov-Chain-Monte-Carlo Approach for System Identification with an Application to Finite-Element Models
Estimating unknown system configurations/parameters by combining system knowledge gained from a computer simulation model on one hand and from observed data on the other hand is challenging. An example of such inverse problem is detecting and localizing potential flaws or changes in a structure by using a finite-element model and measured vibration/displacement data. We propose a probabilistic approach based on Bayesian methodology. This approach does not only yield a single best-guess solution, but a posterior probability distribution over the parameter space. In addition, the Bayesian approach provides a natural framework to accommodate prior knowledge. A Markov chain Monte Carlo (MCMC) procedure is proposed to generate samples from the posterior distribution (an ensemble of likely system configurations given the data). The MCMC procedure proposed explores the parameter space at different resolutions (scales), resulting in a more robust and efficient procedure. The large-scale exploration steps are carried out using coarser-resolution finite-element models, yielding a considerable decrease in computational time, which can be a crucial for large finite-element models. An application is given using synthetic displacement data from a simple cantilever beam with MCMC exploration carried out at three different resolutions.
- Research Organization:
- Lawrence Livermore National Lab. (LLNL), Livermore, CA (United States)
- Sponsoring Organization:
- US Department of Energy (US)
- DOE Contract Number:
- W-7405-ENG-48
- OSTI ID:
- 15014635
- Report Number(s):
- UCRL-TR-209485; TRN: US200802%%1355
- Resource Relation:
- Other Information: PBD: 7 Feb 2005
- Country of Publication:
- United States
- Language:
- English
Similar Records
Final Report: Detection and Characterization of Underground Facilities by Stochastic Inversion and Modeling of Data from the New Generation of Synthetic Aperture Satellites
Markov chain Monte Carlo posterior sampling with the Hamiltonian method.