In magnetic flux leakage (MFL) nondestructive evaluation, a crucial problem was signal inversion, i. e., the defect profile and its parameters are recovered from measured signals. An inverse algorithm based on genetic algorithm was prcposed for reconstructing 2-D defect frcm MFL signals. In the algorithm, radial-basis function (RBF) neural network was utilized as forward model, and genetic algorithm was used to solve the cptimizaticxi problem in the inverse problem, which can avoid the local minimum solution possibly encountered in the iterative inverse algorithm based on gradient descent and consequently achieve the global cptimized solution of the inverse problem. Experimental results verify the validity of the proposed inversiai algorithm. |