Search

Flaw profile estimation from measurements is a typical inverse problem in electromagnetic nondestructive evaluation (NDE). The increasing availability of multiple measurement modes in NDE requires the development of multisensor data fusion algorithms to solve the NDE inverse problem. This paper proposes a multisensor data fusion algorithm for flaw profiling, based on a recursive state space approach. The problem of flaw profile estimation from given multisensor data is formulated using multiple measurement process models and a state transition model. This formulation enables the application of Bayesian non-linear filters based on sequential Monte Carlo methods. The new approach is computationally efficient if computationally simple measurement models are employed. Moreover, the technique is robust to noisy measurement data. The initial results indicate significant improvement in the accuracy of inversion results when more than one type of measurement data is used for flaw profile estimation. [ABSTRACT FROM AUTHOR]

Flaw profile estimation from measurements is a typical inverse problem in electromagnetic nondestructive evaluation (NDE). This paper proposes a novel state-space approach to combat the ill-posedness in the solution to the inverse problem, particularly in the presence of measurement noise. The inversion approach formulates the inverse problem as a tracking problem with state and measurement equations. This state-space model resembles the classical discrete-time tracking problem, and enables the application of Bayesian non-linear filters based on sequential Monte Carlo methods in conjunction with numerical models that represent the measurement process (i.e. solution of forward problem). The proposed approach is applied to simulated NDE measurements from known flaw shapes. Functional models of the NDE inspection process are used as the measurement model and a probabilistic state transition model is defined for the proposed technique. Initial results of applying the proposed approach to eddy current NDE inverse problems indicate the feasibility of the proposed method. [ABSTRACT FROM AUTHOR]