Abstract
Flaw profile estimation from measurements is a typical inverse problem in electromagnetic nondestructive evaluation (NDE). The application of recursive Bayesian nonlinear filters based on sequential Monte Carlo methods, in conjunction with measurement process models and a Markovian crack growth model, is a new approach for solving such inverse problems. The approach resembles the classical discrete-time tracking problem and is robust to the noisy measurement data. This paper reports a comparative study of the results of employing different measurement models in this Bayesian inversion framework. The results are evaluated on the basis of accuracy and computational cost.
Original language | English |
---|---|
Article number | 4787458 |
Pages (from-to) | 1566-1569 |
Number of pages | 4 |
Journal | IEEE Transactions on Magnetics |
Volume | 45 |
Issue number | 3 |
DOIs | |
State | Published - Mar 2009 |
Externally published | Yes |
Keywords
- Neural networks
- Nondestructive testing
- Particle filters
- Response surface methodology (RSM)