Abstract
Accurate and timely prediction of defect dimensions from magnetic flux leakage signals requires one to solve an inverse problem efficiently. This paper proposes a new inversing approach to such a problem. It combines cuckoo search (CS) and particle filter (PF) to estimate the defect profile from measured signals and adopts a radial-basis function neural network as a forward model as well as the observation equation in PF. As one of the latest nature-inspired heuristic optimization algorithms, CS can solve high-dimensional optimization problems. As an effective estimator for a nonlinear filtering problem, PF is applied to the proposed inversing approach in order to improve the latter's robustness to the noise. The resulting algorithm enjoys the advantages of both CS and PF where CS produces the optimized state sequence for PF while PF processes the state sequence and estimates the desired profile. The simulation and experimental results have demonstrated that the proposed approach is significantly better than the inversing approach based on CS alone in a noisy environment.